The Inference Report

Modern video games provide a measurable testbed for AI models, combining abilities of visual understanding, instruction decomposition, goal planning, and precise action control over multiple temporal horizons. Existing datasets and benchmarks, however, either cover a narrow range of games, lack language instructions, or rely on high-variance online rollouts. To address these challenges, we introduce GameHorizon, a unified data and evaluation suite that measures gameplay capabilities at different horizons for diverse model families. GameHorizon Suite consists of three components. First, GameHorizon-Annotator is a scalable and automated annotation pipeline for multi-horizon instructions. Second, utilizing the pipeline, we construct GameHorizon-Data, the first large-scale AAA gameplay dataset with temporally aligned videos, player actions, and multi-horizon instructions. It comprises 5,000 hours of recordings from 21 games, collected by 100 human expert players. Third, we build GameHorizon-Bench with reproducible offline and stepwise online testing. The offline track enables reproducible evaluation using thousands of standardized questions organized into three primary tasks and a series of diagnostic variants, while the online track tests whether offline scores reflect actual gameplay capabilities and localizes failures to specific steps within long-horizon gameplay. Based on our GameHorizon Suite, we evaluate 47 models through more than one million model invocations, revealing a meaningful hierarchy of task difficulty and pronounced differences in model capabilities. Our work can provide a standardized yardstick for evaluating gameplay capabilities across horizons and model families. We will release our dataset, annotator, and benchmark to facilitate future research.

Multi-turn tool-use failures can hinge on a single model call, yet reward variation alone does not reveal which call would benefit from training. When rewards depend on later interactions, their variation can reflect downstream randomness rather than differences between the current actions. We introduce Critical-State RL to identify trainable states in multi-turn interactions. Given task-defined candidate calls and local rewards, the method assesses whether each reward captures the action's effect on task success and whether improvement over a reference policy is possible. It then uses nested sampling to separate action-dependent reward variation from continuation noise and optimizes the policy at the selected states using contextual-bandit training. Experiments on the Berkeley Function Calling Leaderboard (BFCL) v4 compare training at diagnostic-selected states with training at alternative states. For missing-function tasks, the diagnostic selects the response after the tool becomes available; for missing-argument tasks, it selects the response before the missing argument is supplied. Training the selected responses improves performance, including about 14 percentage points on the missing-function task, while training the alternatives leaves performance flat or worse. We further apply the recipe across models and tasks, including logged repeat-call avoidance and memory management.

Video world models enable interactive exploration of dynamic environments, yet struggle to respect prior observations over long horizons and across viewpoints. We present WorldCrafter, a video world model that learns a camera-queryable implicit 3D-aware memory for this purpose. The key insight is to let the requested viewpoint shape how multi-view evidence is compressed into the video generator's limited token budget. Trained jointly with the video generator, a memory encoder and pose-conditioned readout module integrate historical observations into a fixed set of target view-specific tokens before denoising, without explicit depth-based correspondences. By combining this memory with recent temporal context and few-step distillation, WorldCrafter enables streaming scene exploration from a single input image or text prompt. Experiments across static and dynamic scenes show substantial gains in long-horizon consistency and camera-control accuracy while preserving visual quality during minute-scale exploration.

We present onPanda, an interactive tool for efficiently annotating LLM alignment data and agent trajectories. onPanda adopts token-level correction as its core interaction: while reading a model response, the annotator locates the first inappropriate token and either picks a substitute from the model's candidate tokens or types the correct text via free-form editing. The system then truncates everything after that position and continues generation from the corrected prefix, repeating this locate-correct-continue loop until a satisfactory response is obtained. This mechanism lets annotators precisely steer model outputs at low cost: a small controlled study suggests that onPanda reduces median annotation time by 52% over manual post-editing. Since the vast majority of tokens in the final response are generated by the model itself, the resulting data largely preserves the model's sampling distribution and is well suited for constructing on-policy SFT and preference data. Furthermore, the token-level corrections recorded during annotation provide fine-grained supervision with precise positions and naturally paired positive--negative samples. onPanda also connects to external tools and harnesses, enabling interactive trajectory annotation in realistic environments. In addition, we release Panda-CVL, a dataset annotated with onPanda, together with a benchmark for token-level correction.

On-device large language models (`LLMs'), e.g. running on mobile phones, are ripe for improvement via personalization. The limited compute resources of mobile devices impose limits on model scale and thus model quality, making any realizable quality gains highly impactful. At the same time, their personal nature (i.e., the close coupling to a particular user) means that a given on-device LLM tends to be used in similar, predictable patterns over the course of time. This paper presents a novel method for personalizing on-device LLMs. It trains a hypernetwork to map a user's context tokens to a low-rank adaptation (`LoRA') well-suited to that user. Once the trained common artifacts are deployed to users' devices, each user uses the hypernetwork to synthesize (entirely on device) a personalized LoRA. This approach blends the benefits while avoiding the drawbacks of two existing approaches to LLM customization: in-context learning (`ICL') and parameter-efficient fine-tuning (`PEFT'). Like ICL (and unlike PEFT), the on-device phase of our approach is computationally feasible, requiring only forward passes through neural networks. Like PEFT (and unlike ICL), our approach modifies the `target' base LLM via weights (the LoRA), avoiding negative consequences (e.g. increased latency) associated with extending the input sequence. Our approach is particularly well-suited to the mobile device regime. Apart from the on-device compute and latency benefits mentioned, it also requires minimal additional storage, as internally its architecture partly leverages the same LLM weights as belong to the target LLM to be personalized. We demonstrate the benefits of LoRA-generating hypernetworks on several representative personalization datasets, comparing against baselines like ICL and PEFT. Of note, our personalization experiments focus on more challenging and less studied long-form text generation tasks.

Dexterous manipulation depends on contact dynamics that are often only partially observable from vision. Recent World-Action Models (WAMs) couple predictive video world modeling with action generation, but remain largely vision-centric and therefore cannot directly model these contact dynamics. We present DexTacWAM, a visuo-tactile WAM that encodes each fingertip independently, aggregates the resulting features through a finger- and pose-aware tactile compressor, and injects the tactile latent into a video diffusion world model for joint visuo-tactile world modeling. Across six contact-rich dexterous manipulation tasks on a 22-DoF bimanual platform, DexTacWAM achieves the highest score on every task, averaging 70.6 versus 38.0 for the strongest baseline. Ablations attribute the gain to modeling contact evolution as part of the predicted world state rather than tactile conditioning alone: removing tactile world modeling reduces the four-task mean from 74.7 to 26.6 while keeping the same tactile features and action expert. After four hours of tactile-encoder adaptation with a frozen pretrained vision VAE, our continual vision-to-touch learning extends the pretrained video model to touch using roughly 100 demonstrations per task without tactile midtraining, while retaining visual prediction quality within 0.5 dB of vision-only counterparts. The compressor retains 89.4% of pre-fusion contact recall while enabling 2.26x faster training and 1.29x faster inference. Together, these results show that pretrained video priors can be extended to distributed multi-finger contact dynamics in a data- and compute-efficient manner.

Agent harnesses, the external systems that mediate model-environment interaction, can substantially improve agent performance, but their gains remain tied to the harness at deployment. Because the best harness varies across domains, instances, and models, a general-purpose agent must either settle for a suboptimal shared harness or route among an ever-growing set of specialized ones. We therefore study agent harness distillation: using a domain- or instance-optimized harness as training-time guidance and transferring the behaviors it induces into model weights, so that its gains survive under a single fixed target harness. The challenge is that the two harnesses differ in action space and available information, so guidance from the optimized harness cannot serve directly as supervision for the target one. We introduce Harness-Zero, which enables harness distillation through agent-as-harness. Guided by the optimized harness, a harnessing agent corrects student responses before execution in the target harness's action space, turning harness guidance into training demonstrations. Fine-tuning on the resulting trajectories internalizes harness-induced behavior into the model, so the specialized harness can be removed at deployment. Our experiments spanning knowledge work, tool use, and science domains show that: (1) For frontier LLMs using the same evolved harness, agent-as-harness outperforms code-as-harness. (2) With the specialized harness removed at deployment, Harness-Zero improves the base model's macro-average task success from 23.3% to 44.3%, even exceeding the 41.7% it reaches with that harness still attached. (3) Harness-Zero recovers harness-induced behaviors absent from the base model, with 82.3% average recovery across 28 patterns in the three domains.

An LLM agent's capability is largely magnified by its harness, namely the prompts, control flow, tooling, memory, and context management surrounding the frozen backbone model. Recent methods increasingly automate this process by iteratively proposing and selecting component-wise edits of an agent harness, practically establishing a form of recursive self-improvement (RSI) at the agent-system level. However, such recursive evolution may overfit by memorizing the training tasks, showing large in-distribution gains that shrink or even vanish on out-of-distribution benchmarks. We introduce Regularized Recursive Self-Improvement of Agent Harnesses (RRSI), which incorporates the principles of regularizations into harness self-improvement by constraining the evolution candidate proposal and selection. The proposer operates with a temporally annealed budget, limiting how many edits a candidate can bundle, and it encourages unexplored trajectories based on evolution history. The selector is equipped with a critic and a pruner: the critic screens benchmark-specific proposals, while the pruner, removes changes that are too small, too expensive, or no longer useful. Together these constraints favor reusable agent mechanisms over benchmark-specific ones or even noises. Across eight benchmarks spanning coding, agentic workspace and engineering design tasks, RRSI gains up to 14.1 points on the split it evolves against and up to 4.7 points on the five out-of-distribution benchmarks, while producing a harness that runs on 30% fewer policy tokens than the unregularized evolution. Code is available at https://github.com/google-research/rrsi and project page is https://regularized-rsi.com/.

Agents today often take real-world actions that depend on long-term memory and context recall over time. However, most current memory benchmarks are built for a conversational question-answer format, where the question itself signals that some fact must be retrieved, and often which one. Moreover, benchmarks rarely require anything beyond accuracy from submissions, allowing memory systems to make unreasonable cost/time tradeoffs to achieve higher scores. We present DolphinBench, a benchmark that evaluates memory directly through an agent's task completion. DolphinBench includes three knowledge-work personas with roughly 500k tokens of user messages per persona and evaluates agents on tasks that depend on information from that history. We verify all 200 tasks per persona by running an agent with and without the relevant history, requiring success with it and failure without it. Finally, we require all evaluations to report total cost and latency alongside accuracy, which enables us to evaluate agent memory systems holistically. No existing memory benchmark combines all three. The dataset and evaluation code are available at https://dolphinbench.ai.

As agents are deployed with increased autonomy, even extremely rare events along their stochastic output trajectories can occur and prove catastrophic. Safe deployment therefore does not depend on whether these events can occur, but on how often they might. We study the problem of estimating the probability of rare events that arise from stochastic variation in the agent's own actions. Estimating this type of risk requires searching over the combinatorially vast space of trajectories. Naive Monte Carlo is computationally prohibitive in this regime, and constructing effective importance sampling (IS) proposals requires coordinated changes to a context-dependent chain of conditional distributions. We develop a new IS method that perturbs the original model's weights to construct the proposal. The proposal is itself a differentiably parameterized language model, enabling gradient-based search over weight space. We formulate an objective that combines a differentiable surrogate for event amplification and an adaptive regularization scheme that dynamically balances amplification against estimator stability. We evaluate our approach on $\sim$120M and $\sim$2.6B models across three event families spanning 300+ rare events as rare as $10^{-9}$, with reference probabilities computed with $<10\%$ relative standard error. In our most verifiable settings, we observe that our IS estimator achieves over $800\times$ compute-weighted efficiency gains over naive Monte Carlo for events with probabilities lower than $10^{-7}$. Our implementation is available at https://github.com/namkoong-lab/iterative-unalignment.

LLM agents are increasingly deployed in collaborative settings, yet long-term interaction may give rise to undesirable coordination. We study the emergence of collusion in a long-horizon multi-agent environment: two agents repeatedly complete individual tasks, share task logs, verify each other's work, and receive rewards. We introduce realistic constraints that make compliance with the verification protocol incompatible with reward maximization, and find that agents increasingly deviate from the protocol over repeated interactions. Collusion emerges in 94% of trajectories across 10 models, and more capable models within the same family reach it earlier. Controlled peer interventions show that collusion is shaped by peer behavior, while ablations reveal additional effects of reward structure, the verification feedback agents receive, and their interaction history. In particular, restricting the amount and scope of interaction history available to agents reduces collusion. Overall, our findings show that long-horizon interaction can reshape how agents coordinate in ways that create safety risks.

Scientific workflows often require choosing among known relations before a deterministic calculation can proceed. Whether observations share a culture, treatment or reference standard can change the scientific meaning of the resulting count or comparison. We evaluate Jev as a semantic decision component using a harness that follows its documented guidance and assigns arithmetic to code. The study compares twelve model configurations on twenty source-grounded Choices across ten scientific cases, each repeated five times. We measure semantic selections, downstream outputs and final claim labels separately. Jev matched five other configurations at complete semantic correctness and achieved the lowest observed median latency among successful responses. Across three comparison models, seven wrong selections on one culture-history question changed downstream counts while preserving the correct final label. These results identify a useful role for Jev in prepared scientific decision tasks and show why evaluating that role requires checking the relations and quantities that a workflow will reuse.

Visual instructions for physical tasks are typically authored in one context and followed in another, requiring users to translate demonstrated tools, materials, and spatial relationships into their own environment. We introduce Generative Tutorial, a conceptual framework for live visual instruction that depicts intended outcomes and actions within the user's environment and task flow. A formative evaluation of state-of-the-art image and video generation identifies failures and potential benefits across 15 physical tasks. Drawing on these findings, we build an augmented-reality prototype system that proactively generates goal images and demonstration videos using observed workspace context and predicted visual outcomes of preceding actions. A 24-participant lab study found higher task performance quality, greater perceived workspace correspondence, and shorter step-confirmation intervals with the system than with pre-authored guidance. Qualitative findings highlighted how contextual resemblance shapes trust, how generation errors affect interpretation, and how guidance delivery should adapt to users' needs, informing future designs.

Pharmaceutical process characterization is central to Quality by Design because it defines how variations in process parameters affect the ability to meet product quality specifications, thereby supporting proven acceptable ranges and robust manufacturing. In practice, however, characterization still relies largely on factorial design of experiments (DOE) approaches, which are inefficient for resolving multivariate pass/fail boundaries in higher-dimensional spaces. While Bayesian optimization has transformed process optimization, adaptive methods for multi-objective process characterization remain lacking. Here, we introduce JAREX (Joint Acceptable Region EXploration), a Bayesian active-learning acquisition function for multi-objective process characterization. JAREX formulates characterization as a joint boundary-learning problem and adaptively selects experiments to recover the joint pass region defined by simultaneous satisfaction of threshold criteria across multiple objectives. JAREX combines an optimistic joint-feasibility mask with a multi-objective extension of randomized straddle, focusing sampling on the joint edge of failure. Our benchmark study suggests that JAREX provides more accurate and sample-efficient recovery of the joint pass region than factorial DOE, space-filling designs, and greedy objective-wise strategies over the full experimental budget range. For batched experimentation, it reduces the number of iterative process characterization experiments by more than half while preserving high accuracy for the boundary-identification task. Implemented in the open-source obsidian package, JAREX provides a modular framework for adaptive, data-efficient multi-objective algorithmic process characterization, supporting sample-efficient range finding in high-dimensional spaces.

Neural operators evaluate parametric partial differential equations cheaply but degrade sharply outside their training distribution. Physics-informed neural networks avoid dependence on labeled data, yet their optimization can be basin-fragile: when the governing residual admits multiple solutions, a PINN trained from scratch may converge to a physically incorrect state despite achieving a small residual. We show that these failure modes can be addressed jointly: an imperfect NO provides the structural prior needed to place a PINN in the correct solution basin, while the PDE residual refines the solution beyond the operator's accuracy. We introduce a three-stage framework that freezes the spatial basis of a physics-informed NO, extrapolates its solution branch to an out-of-distribution parameter using a polynomial continuation prior, and distills the resulting field into a fresh PINN. The NO need not be accurate at the target; it transfers solution-branch information, while PDE residual minimization in the PINN governs convergence. We evaluate the framework on three nonlinear PDEs: 1D viscous Burgers, 2D steady Allen-Cahn near a pitchfork bifurcation, and 2D steady lid-driven cavity flow. For Allen-Cahn, where the trivial solution satisfies the PDE residual exactly, a standard PINN collapses to the trivial zero branch, whereas distillation from the crude extrapolated operator recovers the non-trivial branch that matches the finite-difference reference. For the lid-driven cavity, extrapolating to a Reynolds number of Re = 3200 accelerates convergence to the correct physical state, achieving competitive accuracy using fewer parameters and optimization steps than recent literature baselines. These results establish a simple principle: an NO need not accurately predict the solution to be useful; it only needs to identify the correct basin from which PINN optimization can recover it.

A model generalizes outside its training distribution only when it computes a representation structurally equivalent to the generating mechanism, not an approximation fitted to it. Such equivalence is necessary for exactness in and out of distribution, and extrapolation is governed by this exactness at inference, whatever its realization. Tensor Logic shows this: a zero-temperature contraction is equivalent to discrete logic, deducing in place with no artefact extracted, its tensors Boolean, its embeddings orthonormal, only its arithmetic continuous. Lacking infinite recursion it reaches Datalog, not Prolog, and though exact over closed domains it needs external memory to bind a novel entity. The criterion needs neither a discrete representation nor an extracted expression, and constrains inference, not training: an exact marginal in $[0,1]$ passes, a Neural Network thresholded to a hard label does not. Logic Tensor Networks fail it, while differentiable ILP and Tensor Logic at $T=0$ pass. Piecewise-affine extrapolation divergence and an inability to bind novel entities are two faces of a shortfall in exact representability. For hybrid architectures, a propagation rule follows: the output inherits the bounds of every fitted estimator on its path, explaining which axes fail in equivariant models and the ARC-AGI induction/transduction split. Only an exact hypothesis class certifies what the training data leave underdetermined: on a law-derived partition it finds the $56.3\%$ of distant queries that are answerable, which ensembles meet with false confidence and distance metrics rank backwards. Common inductive biases, from symmetries to memory, reach exactness only because humans inject them, an argument for inducing exact representations rather than fitting surrogates whose residuals, even at the arithmetic floor in training, diverge outside the data and compound under composition.

Despite the success of neural models in natural language processing, their black-box nature limits interpretability and conceals the linguistic phenomena underlying their predictions. We present SLITE, an explainable hybrid model for Recognizing Textual Entailment that integrates two complementary layers of semantic analysis: a structural-relational layer, based on semantic compatibility and incompatibility between compositional entities, and a distributional-informational layer, based on structured patterns of information change between embedding-based representations of the premise and the hypothesis. We propose 17 features that combine entity-level semantic relations, polarity-sensitive lexical matching, and alignment measures over semantic sub-representations of the similarity matrix, including measures based on entropy and transfer entropy. A logistic regression trained on these features achieves an accuracy of 83% on three-class SICK and 96% on SICK-CE, outperforming IsoLex by 4 percentage points and falling within 2 percentage points of RoBERTa with a fraction of its computational complexity. Ablation studies and SHAP analysis confirm that structural-relational features are the primary drivers of classification, while distributional-informational features provide essential complementary contributions, particularly for detecting neutrality and contradiction. Our results demonstrate that further exploration of hybrid approaches is a viable and scientifically productive alternative to massive neural architectures, and we hope they will strengthen the dialogue between linguistic theory and computational modeling of inference

In this paper, we study nonasymptotic $L^p$ error bounds for interval length and conditional coverage in split conformalized quantile regression (CQR). Our bounds rely on local regularity conditions and accuracy guarantees for the estimated quantiles. We further instantiate our bounds for quantile regression with sparse ReLU neural networks. We also consider covariate shift, where the calibration and test covariates have different distributions, and derive nonasymptotic bounds for this setting. We obtain matching minimax upper and lower bounds in expectation for two constructed fixed-score calibration benchmarks under known covariate shift. The bounds match for every $p\in[1,\infty]$ in the scalar problem and for finite $p$ in the $K$-threshold problem; for the latter, a high-probability minimax lower bound holds for every $p\in[1,\infty]$.

Autonomous software engineering agents (SWE-agents) automate coding tasks. Each agent update may require re-running the full benchmark to detect regressions and improvements, at a cost of hundreds of millions of LLM tokens per run, which makes evaluation a bottleneck. One solution is to evaluate only a subset of benchmark instances. Yet, simple approaches, such as random sampling or stratified random sampling based on past pass/fail outcomes, risk producing high variance and unrepresentative subsets. We turn to agent trajectories, the step-by-step record of the actions an agent took. We propose a trajectory-aware subset selection approach that replaces random sampling with deterministic selection based on trajectory embeddings. We first group test set instances by their test outcome in a recent full test run to preserve the historical pass/fail rate, then select the subset using the trajectory's embedding space. We evaluate 76 subset selection configurations, including random sampling, embedding-based selection, clustering-based selection, and hybrid shortlist-then-subsample strategies, across three regression scenarios: same-configuration reruns, model and configuration changes, and agent framework changes. Our best trajectory-aware method is the one selecting benchmark instances closest to the centroid of each outcome group in the embedding space. It achieves the lowest estimation error among all methods we evaluate. For instance, when evaluating a given agent version on a selected subset of 5% or 10% of the test instances, our approach reduces the average estimation error by 3--11% and the worst-case error by 4--11% relative to the typical draw and 38--46% relative to the 95th-percentile draw of the strongest baseline. Our results show that a 10% trajectory-aware subset keeps the median estimation error below 5% while cutting token cost by roughly 90%.

Personal AI agents make recommendations and take actions on people's behalf in high-stakes economic contexts, e.g., buying a flight, choosing health insurance, or selecting a graduate program. The agent is given access to the user's personal context, e.g., their email inbox and a structured profile of personal attributes, with the intention of making an optimal, personalized decision for the user. We show that by simply providing this personal context, the agent steers recommendations based on inferred wealth, without being explicitly instructed to do so. In a suite of 325K experiments on 13 agents across three types of economic decisions (flights, health insurance, and graduate programs), we find that 8 models systematically choose more expensive options for wealthier users when requests are identical. This steering continues even when it directly goes against the user's stated objective: when explicitly instructed to find the cheapest option, some agents still act on the wealth profile they have inferred. It also occurs when wealth is inferred from ambient data, such as emails unrelated to the task. And it persists under privacy controls that block specific attributes: blocking financial attributes largely removes the disparity, but blocking other attributes leaves it unchanged and can increase it by up to 40% for insurance, as agents rely on the remaining signals to infer wealth. Larger and more capable models are no better; Claude Opus 4.8 shows the largest effect. We term this misalignment "adversarial delegation", in which the very conditions that make a personal AI agent useful - access to personal information - enable it to act against the user's interests.

Scientific weak signals are early, low-visibility research directions that later become central to mature scientific topics, yet existing resources such as trend tracking, citation forecasting, and foresight reports rarely provide validated reference sets that link concrete early precursors to later paradigms. We introduce BackTrend, a retrospective benchmark in which, given a mature target topic and a temporal evidence constraint, systems must recover two types of precursors: problem-space signals, underrecognized research problems, and solution-space signals, emerging methods for known problems. BackTrend contains 25 mature target topics in artificial intelligence and machine learning and 66 human-validated weak signals, reconstructed from large-scale literature by grounding each candidate in its 2019-2024 publication-frequency trajectory. We evaluate frontier LLMs, RAG systems, and agentic research systems using semantic matching and coverage-based metrics. Current systems often generate plausible but misaligned precursors, exhibiting topic drift, granularity mismatch, near-miss matching, and incomplete coverage; the strongest system achieves only 10.1% F1, while Coverage10 reaches at most 18.5% of the reference signals. Our budget analyses show that additional retrieval and web-search evidence can improve performance up to a moderate budget, but does not by itself close the substantial performance gap.

Social simulation offers the social sciences an experimental instrument that the real world cannot supply, and generative agents have transformed it by acting as silicon samples that unite agent-based modeling with real behavioral data. Existing platforms verify collective behavior, align simulated populations with real societies in cross-sections, and employ autonomous agents for the research process. However, two social science requirements remain without systematic support: intervention in the content of a simulation and the researcher's control over the process that produces it. We present SocioVerse2, which extends SocioVerse 1.0 into a human-AI co-evolutionary paradigm built from two loops and one infrastructure. The longitudinal simulation loop simulates the target population with evolving environments and forks counterfactual branches via interventions. The controllable research loop takes the study itself as an editable state and updates state versions via controllable editing. The social science agentic infrastructure carries both loops through composable skills with researcher checkpoints, a population service over five persona pools, and an environment service over 21 real-world signal sources with point-in-time guarantees. We validate SocioVerse2 across three case families and seven case studies, from reproducing canonical agent-based models to modeling policy processes on real records and nowcasting macro-economic indices beyond the response model's knowledge cutoff. With the human-AI co-evolutionary paradigm, these cases go beyond system demonstrations to become substantive studies that investigate frontier questions in their respective disciplines. Code, data services, and a workbench are released as open-source resources.

Dormant tree pruning is labor-intensive yet essential for maintaining modern high-productivity fruit orchards. In this work, we focus on pruning of modern planar tree training systems - V-Trellis apples and UFO cherries - where trunks and primary branches are trained into approximately planar walls. We introduce an end-to-end pipeline to learn a closed-loop visuomotor controller for robotic pruning. This controller is trained entirely using simulation and synthetically generated data and deployed in real orchards in a zero-shot manner. The pipeline comprises synthetic generation of planar orchard tree meshes, construction of a physics-based orchard simulator, automated collection of successful pruning trajectories via motion planning, and policy learning with a novel hybrid reinforcement-learning algorithm that combines offline demonstrations with online simulated rollouts. The controller uses optical-flow inputs from a wrist-mounted camera - avoiding the need for full 3D-reconstruction - and continuously guides the cutter through cluttered branch environments to a specified cutpoint with correct tool orientation. In exhaustive simulated task-space evaluations over 3,000 pruning points, the policy attains 49.9% success on V-Trellis apples and 46.0% on UFO cherries. We validate the learned controller across 38 physical trials - comprising 28 outdoor field trials in commercial and experimental orchards and 10 indoor laboratory tests - demonstrating zero-shot sim-to-real transfer. The learned policy also outperforms a classical RRT-Connect baseline on physical hardware in laboratory trials.

Tone languages constitute over 50-70% of the world's languages, but the vast majority are low-resource, lacking the large transcribed corpora needed for automatic tone classification. Existing datasets are typically collected at the sentence level, whereas field linguists require fine-grained syllable-level annotations. We propose ToneCL, a lightweight contrastive learning framework for few-shot syllable-level tone classification. We simulate low-resource conditions on Mandarin and Vietnamese, limiting labeled data to tens of examples per tone class. ToneCL is pretrained on unlabeled speech with augmentations that preserve tonal identity, then fine-tuned on few-shot examples. Experiments show our method consistently outperforms baselines, achieving 91.6% on six-speaker Mandarin at 10 shots. Cross-lingual transfer is also effective: pretraining on Vietnamese and fine-tuning on Mandarin reaches 91.0\% accuracy at 10 shots. Ablation confirms that frequency band rejection is the most critical augmentation.

When an LLM supplies an argument that a user could not readily construct, how can the user decide whether to accept its claim? Inspired by interactive proofs, we model human-LLM deliberation as an interaction between a prover with unrestricted internal search and a resource-bounded human verifier. The verifier requests and checks supporting details without access to the LLM's internal state. Passed checks accumulate evidence toward an acceptance threshold. We prove anytime-valid soundness against adaptive provers: the probability of ever accepting a false claim is at most a chosen error level, provided the task supplies bounds on false passes and human checking errors that remain valid after every relevant history. A finite-horizon completeness bound additionally requires bounds on the adequacy of honest responses and sufficient diagnostic progress. Further checks can strengthen the evidence for acceptance, but each requires another adequate response and reliable human effort. Whether this tradeoff permits certification depends on the verifier's effort budget, cognitive load, expertise, and fatigue. We identify conditions under which the supplied bounds certify a specified sequence of local checks but not a specified global check under the same resource budgets.

Whole-slide pathology images (WSIs) contain gigapixel-scale visual content, creating a major scalability challenge for slide-level multimodal large language models (MLLMs). Existing approaches process thousands of patch tokens and typically apply compression only after slide encoding, leaving multimodal attention computationally expensive. We introduce SLICEChat, a slide-level MLLM that integrates progressive token pruning within a hybrid Mamba--Transformer slide encoder. Mamba layers enable efficient long-range propagation, while Transformer layers preserve global interactions as the sequence is progressively shortened. Between stages, language-supervised, region-aware pruning removes spatially coherent low-utility regions under a controlled keep-rate schedule, producing compact slide representations before multimodal fusion. On SlideBench VQA, SLICEChat achieves 79.84% accuracy on TCGA and 59.09% on BCNB cohorts, outperforming prior slide-level pathology MLLMs, and achieves the highest overall WSI-Bench metrics. It also provides competitive memory usage and the inference latency among the evaluated models. These results demonstrate accurate and computationally efficient multimodal reasoning over gigapixel WSIs.

Evaluation of Computer-Use Agents (CUAs) is often limited to the final deliverables they create (at the end of hundreds of steps) and assessed with functional verifiers, as seen in OSWorld. However, such evaluation of end-state performance lacks transparency into how and why agents fail in various tasks, obfuscating critical insight for subsequent improvement. For instance, agents that err during keyboard inputs would require a different mitigation strategy from those that fail to precisely provide click-based inputs on the graphical UI. We introduce OSWorld-Pro: a set of over 300 tasks containing over 2800 subgoals to enable the procedural evaluation of CUAs grounded in over 67,000 human annotations. We use robust human-aligned LLM-Judges to evaluate the fulfillment of OSWorld-Pro subgoals and thereby reveal the progress that models make throughout a series of sequentially dependent subgoals. Our findings reveal that OSWorld-Pro is challenging even for state-of-the-art LLMs, with top performers like Claude Opus 5 achieving only 75.7% vs. 83.4% on OSWorld. Furthermore, we identify critical process-focused failure modes of various models (e.g. subgoal-irrelevant actions and click-based mistakes) to provide insights to improve performance and efficiency of CUAs.

When a language model answers from a curated corpus via graph-based retrieval, a large grounding uplift does not establish reasoning over the retrieved structure: the context may already expose the gold answers. We propose exposure accounting, which classifies each gold item by whether the shown context exposes it and whether the answer recovers it. Its scalar reference is the copy ceiling, the recall a verbatim copy of the context achieves; signed gain over copy measures the model's recall relative to this deterministic, judge-free baseline. Across ten models, unaided recall averages 0.26 and grounded recall 0.92, yet gain over copy is uniformly negative (-0.067 to -0.022). Of 11,360 gold-item observations, representing 1,136 target instances evaluated under ten models, only three unexposed items receive lexical credit. A stratified model-judged audit of 423 observations, with a symmetric quotation-verification policy, estimates that 97.1% of credited items assert the requested relation; all three unexposed credits fail relational adjudication. On targets the scaffold does not expose, lexical recovery falls from 0.121 unaided to 0.004 grounded; adjudication validates 71 of the 92 unaided credits and none of the three grounded credits, without establishing full-frame relational recovery rates. Rephrasing questions outside the graph's title vocabulary reduces exposure from 0.964 to 0.328, while an absence-triggered fallback activates on only 2 of 506 questions. A paired production study improves judged quality by +0.27 pooled, but negative controls do not establish content specificity beyond a well-formed on-corpus block. These results support exposure accounting as a standing control for corpus-derived evaluations. The accounting distinguishes exposed-item omissions from beyond-exposure recoveries; it does not determine whether reasoning occurred.

Artificial intelligence (AI) registers and inventories aim to make governmental AI visible, but their institutional scope, schemas, and reporting practices construct different representations of public-sector AI. We compare 8,368 records from country-specific and transnational inventories covering 72 countries. Across 23 harmonized fields, registers shared a descriptive core but rarely requested information about appeals, risks, legal bases, or external evaluation. We found that broad schemas often contained substantial missingness, schema similarity showed no significant patterned convergence, and multiple sources covering the same jurisdictions overlapped only selectively. Based on these findings, we synthesize a layered visibility framework that shows how register records reflect disclosure arrangements and why interoperability requires shared concepts, clear definitions, and preserved provenance.

Hybrid AI-physics climate modeling aims to improve coarse (~100km-resolution) Earth system models by learning to parameterize subgrid processes from high-fidelity data. However, this so far mostly involves local-in-time, diagnostic parameterizations, in which the subgrid state depends only on the current coarse state with no memory of previous states, which is unrealistic for processes such as convection that have intrinsic persistence. To address this, we enhance local-in-time parameterizations by learning prognostic variables that compactly carry important, additional past information where no explicit sub-grid information is available. First we compress past information into a low-dimensional latent space using an autoencoder, which then informs a neural network trained to parameterize targeted subgrid-scale processes. We then replace the autoencoder with symbolic equations that govern the time evolution of the latent variables, yielding additional prognostic memory variables that can be integrated alongside the resolved atmospheric state. We evaluate this approach on two systems: the Lorenz-96 model (online) and surface precipitation from high-resolution atmospheric simulations (offline). A forced multivariate linear ordinary differential equation recovers most of the added value achieved by the autoencoder-based approach in both experiments. Benchmarked against diagnostic parameterizations without memory, our memory-informed approach improves climate statistics and temporal structure, including a realistic diurnal cycle of tropical land precipitation.

In high-stakes decision-making applications of large language models (LLMs), practitioners require not only accurate LLMs but also uncertainty estimates for their predictions. Existing approaches to uncertainty estimation for LLMs require access to log-probabilities output by the model or require fine-tuning access. However, many industrial LLM products use closed-source API models, and many such API models like GPT do not return log-probabilities and may not allow fine-tuning. We introduce Pinocchio, an external calibrator that estimates the correctness of responses from black-box API models. Trained jointly on responses from seven LLMs, it achieves 0.862 AUROC predicting the correctness of held-out responses from those same models, and shows zero-shot transfer to thirteen unseen models across eight organizations. Our model needs only a single forward pass to generate an uncertainty estimate and requires no access to the target model's logits, weights, or internal states. A lightweight text only 0.8B checkpoint matches our largest model's AUROC. We release code for adding uncertainty estimation to existing repos in only two additional lines of code.

In automated clinical coding, where the label space spans tens of thousands of diagnosis and procedure codes, models are currently evaluated against a single gold annotation, treating any deviation as error. But we find when two teams code the same 110 ACI-Bench encounters, they agree on only 73% of codes (Jaccard similarity) for the same note; even after an independent clinical audit removes erroneous codes, agreement rises only to 77%. Is that gap error or something systematic? We model the systematic component as coding style $ψ$, a coder- or site-specific policy over what to code and how much to document, and recast coding as $p(\mathrm{code}\mid\mathrm{note},ψ)$, estimating $ψ$ with a 10-dimension rubric. If style were noise, conditioning on it would do nothing. Instead, across five datasets a model conditioned with a data-matching style raises ICD F1 by up to 26 points and an extreme mismatched one lowers it by up to 21. Four prompt based coding methods spanning 39-49 F1 converge to 52-56 once style is supplied (All p<0.05). Much of what single-gold evaluation charges to model error is recoverable, unmodeled style.

In multi-agent social settings, model reliability varies across relationships. Beyond inferring what others will do, an agent must calibrate how confidently those inferences should guide policy selection for each relationship. An agent may maintain a well-validated model of one partner, a fragile model of another, and a model under revision for a third; collapsing these into a single confidence estimate loses information relevant to policy selection. We therefore formalize affective precision as a relationship-specific metacognitive estimate of confidence in the current partner model. Each partner's behavioral evidence updates a local confidence estimate that modulates policy precision during selection, regulating how strongly current beliefs are expressed in policy rather than changing the content of those beliefs. Simulations in a multi-partner graded trust game show that partner-local affective precision influences behavior primarily through policy commitment rather than direct improvement of partner-state inference. Because the mechanism tracks partner-response predictability rather than realized payoff, greater confidence produces sharper policy commitment without necessarily producing higher rewards. Under abrupt shifts in social behavior, confidence accumulated from previously reliable predictions can remain behaviorally active after the relationship changes, showing that confidence revision can lag behind social change. Finally, varying precision gain and priors produce distinct trust-calibration dynamics, showing how confidence accumulation and revision depend on model parameters. Together, these results show how relationship-specific affective precision can distinguish social prediction from social policy commitment.

Reaching a 6-DoF grasp pose in clutter requires a collision-free trajectory, conventionally obtained by reconstructing the scene in 3D and planning inside that reconstruction, at the cost of its accuracy and compute. Potential fields learned directly from images remove that dependency but inherit the classical weakness of artificial potential fields: where attractive and repulsive gradients cancel, the descent grazes the obstacle instead of going around it, and can stall short of the goal. We present an SE(3) neural potential field learned from posed RGB images and supervised with a navigation function, the geodesic distance to the grasp through free space recovered from those same images during training, which removes both failures. On two tabletop scenes, from obstacle-blocked starts executed on a UR10, the field converges within 3 cm of the grasp from every start and every path it executes is collision-free against the ground-truth geometry, against 25% and 0% under image supervision alone; mean clearance rises from under a centimeter to 8.6-8.8 cm and arm-link contacts fall from 20.6-50.4% to 2.7-5.5% of executed configurations. Executed grasp success is 90.0% and 40.0% on the two scenes, the residual failures being refusals of the Cartesian executor rather than of the field. Planning takes about 2 s against 67-133 s for RRT* on a reconstruction of the same images, though under a common offline harness the two are comparable: the deployed margin is the cost of collision-checking a dense reconstruction, not planner complexity.

Agentic time series forecasting concerns systems whose underlying mechanisms evolve, making the relative effectiveness of numerical models, reasoning strategies, and intervention rules inherently time-varying. Consequently, a time series agent must adapt the forecasts it produces and the orchestration policy that determines which components to trust and how to coordinate them. The deployment process naturally provides supervision for this adaptation as forecast horizons elapse and realized targets reveal the effectiveness of earlier decisions. Committing all numerical expert forecasts and candidate agent paths before target observation allows each realized outcome to evaluate the entire alternative set, providing delayed feedback without additional annotation. However, existing time series agents primarily incorporate prior experience through forecast refinement, reflection, or retrieval, without systematically converting realized outcomes into persistent updates to the joint orchestration policy governing later origins. To exploit this delayed feedback systematically, we introduce TimEvolve, a frozen-backbone time series agent that converts each realized outcome into persistent joint updates of expert trust, agent path selection, and intervention strength. A temporally ordered predict, reveal, and update protocol applies this feedback to subsequent forecasts. Experiments across eight Time-MMD domains show that TimEvolve achieves the best average MSE and MAE ranks among fifteen methods and the lowest errors on both metrics in seven domains. These results demonstrate the value of learning forecasting policies from the futures encountered during deployment.

The ideation phase of participatory AI risk assessment often starts with a blank slate or a limited list of predefined risks, making it difficult to surface indirect or systemic harms. To address this limitation, we propose a three-stage ideation support tool. The tool complements participatory AI, rather than replacing it, and helps focus later engagement with affected communities. First, it dynamically discovers stakeholders depending on the given AI use and recursively expanding outward, allowing overlooked or indirect stakeholders to emerge. Second, it simulates these stakeholders with LLMs, connecting them into a network of a given topology, and having them ideate about risks. Third, it prioritizes risks using network centrality measures. In an initial evaluation, we found that betweenness centrality run through agents connected in a small-world network works best as it elevates risks raised by stakeholders who bridge disconnected groups, surfacing novel, systemic harms that traditional methods often miss. On an AI chatbot companion use case, this approach increased the novelty of the identified risks by approximately 1.1 points over single LLM brainstorming, and by 0.5 points over agentic LLM brainstorming, measured on a normalized five-point Likert scale, without reducing the plausibility or severity of the identified risks. To test whether our framework helps a human-led ideation session using the Futures Wheel approach, we divided 11 teams of non-western young chatbot users into two types: control (team) and treatment (team) in a participatory AI risk assessment. The control teams started from a list of risks generated by the 45 AI practitioners in the initial evaluation; the treatment teams started from a list generated by our framework. The treatment teams identified more risks overall, and more systemic, human-computer interaction, and environmental risks.

Argumentative component detection (ACD) is a core subtask of Argument(ation) Mining (AM) and one of its most challenging aspects, as it requires jointly delimiting argumentative spans and classifying them into components such as claims and premises. While research on this subtask remains relatively limited compared to other AM tasks, most existing approaches formulate it as a simplified sequence labeling problem, component classification, or a pipeline of component segmentation followed by classification. In this paper, we propose ITFACD, a novel approach based on instruction-tuned Large Language Models (LLMs) using compact instruction-based prompts, and reframe ACD as a language generation task, enabling arguments to be identified directly from plain text without relying on pre-segmented components. Experiments on standard benchmarks show that our approach achieves higher performance compared to state-of-the-art systems. To the best of our knowledge, this is one of the first attempts to fully model ACD as a generative task, highlighting the potential of instruction tuning for complex AM problems. Our code and the datasets used are openly available in the following GitHub repository.

LLM inference on edge devices is constrained by computational and memory resources, making efficient autoregressive decoding challenging. Speculative decoding alleviates this bottleneck by generating tokens with a smaller draft model and verifying multiple tokens in parallel with a batched target model pass. However, verification introduces a runtime-dependent intermediate regime between memory-bound general matrix-vector (GEMV) operations in decoding and compute-bound general matrix-matrix (GEMM) operations in prefill, as its arithmetic intensity varies with speculation length and acceptance rate. We present SPECTRA, a runtime-reconfigurable tiled architecture that sustains high utilization across the full speculative decoding pipeline. Within each tile, the compute engine switches between systolic execution for GEMMs and vector-lane execution for GEMVs. Across tiles, SPECTRA dynamically adapts computation parallelism by selecting tile count, kernel partitioning, and communication pattern. Both tile-level and system-level reconfiguration operate on a per-kernel basis, enabling efficient execution across these diverse regimes. Evaluated on a 20-tile FPGA prototype across the Pythia, SmolLM2, and GPT-2 families, SPECTRA achieves up to $2.09\times$ speedup from tile-level reconfiguration and a further $1.25\times$ gain from system-level adaptability over fixed designs.

Diffusion models offer flexible motion generation, but translating this flexibility into feedback-responsive humanoid control remains challenging. Hierarchical systems steer motion through references that may exceed a separate tracker's capabilities, leaving recovery and physical execution largely to the tracker. Action-only diffusion generates actions directly but lacks an explicit future-state trajectory for test-time motion objectives. Joint state-action diffusion provides this representation, yet representative controllers often depend on privileged full-body states, and support for learned behavior selection and test-time motion steering remains fragmented. We present PredActor, a predictive action diffusion policy that brings these complementary steering capabilities into one directly executed policy using proprioceptive observations. Conditioned on proprioceptive history and optional task context, PredActor jointly generates executable actions and an internal future-state trajectory. Classifier-free guidance strengthens text-conditioned behavior, while classifier guidance steers predicted states toward test-time objectives. Only actions are executed, without a separate motion-reference tracker or externally estimated full-body states as policy inputs. In simulation, PredActor reaches all 15 destination targets and achieves a text retrieval score of 0.580, compared with 0.373 for conditional action diffusion, with similar observed disturbance survival. To make this guided policy practical onboard, rolling denoising and computation-preserving runtime optimizations reduce the complete callback to 16.790 ms median and 19.383 ms p95 on a Jetson Orin NX, both below the 20 ms control period. We deploy PredActor on a Unitree G1; evaluations across simulation and physical hardware demonstrate text-conditioned motion, disturbance response, joystick control, and semantic interpolation.

Medical agents increasingly combine general reasoning models with specialized clinical tools, yet their capabilities remain largely fixed by what clinicians and engineers design before deployment. Recursive self-improvement (RSI) offers a different paradigm in which agents learn from their own failures and autonomously expand their capabilities, but directly applying RSI to medicine introduces fundamental safety challenges. We introduce MedRSI, the first recursive self-improvement framework for medicine, which continuously transforms diagnostic failures into new clinical capabilities through tool composition and task-specific model training. Inspired by clinical practice, MedRSI introduces two mechanisms for clinically aligned self-evolution. Clinical-cost-aware failure prioritization directs improvement toward errors according to their potential clinical consequences rather than frequency alone. Fast discovery with slow registration separates rapid capability invention from conservative adoption, allowing new tools to enter the persistent agent only after demonstrating sustained benefit across subsequent patient cohorts. Across public glaucoma and heart disease benchmarks and two private clinical tasks, MedRSI progressively develops segmentation, measurement, prediction, multimodal reasoning, and generative capabilities, surpasses manually engineered medical agents, and autonomously discovers solutions to clinical problems not anticipated by its original designers. Our results show that medical agents need not remain constrained by capabilities specified before deployment: with clinically grounded mechanisms governing what to improve and what to retain, they can continuously construct, validate, and accumulate new capabilities from diagnostic experience. Code is available at https://github.com/ImprintLab/MedRSI.

Agents have attracted considerably increasing attention due to the power of executing both Reasoning and Acting (ReAct) in open and dynamic environments. The ReAct process typically exhibits a multi-turn trajectory in which one drives Large Language Models (LLMs) to generate both reasoning chains and task-specific actions in an interleaved manner. However, agents often suffer from significant uncertainty, where identical tasks yield divergent trajectories; trajectories with higher uncertainty often produce incomprehensible behaviors, severely undermining agent credibility. This work conjectures that such trajectory-level uncertainty frequently stems from cumulative turn-level reasoning uncertainty induced by LLMs; the latter often exhibits a collection of branches of divergent reasoning chains and their resulting actions. Built upon this, we present the Graph-based Reasoning UncErtainty in Trajectories (GRUET) method for the uncertainty quantification of ReAct, comprising turn-level reasoning uncertainty quantification and trajectory-level uncertainty aggregation; the former precisely quantifies reasoning uncertainty via modeling the reasoning space spanned by potential reasoning branches as a graph and then approximating the reasoning space complexity with graph complexity, while the latter employs simple aggregation strategies for quantifying the overall trajectory credibility. Empirical evaluations across nine LLMs and five benchmarks validate the effectiveness of our proposed GRUET in terms of selective generation performance, measured by AUROC, AUPRC, and AUARC.

We introduce Graph Neural Automata Clustering (G-NAC), an unsupervised clustering method in which observations interact as cells on a fixed neighborhood graph. A shared recurrent graph-neural cellular rule evolves latent domain states through local interactions, which are converted into a rank-based spectral affinity for partitioning. Across 73 clustering tasks from 57 benchmark datasets, G-NAC achieved a mean adjusted Rand index (ARI) of 0.7951, comparable to Genie at 0.7941 and higher than the other evaluated baselines. Empirical training time and GPU memory scaled approximately linearly from 5,000 to 100,000 nodes. Learned transition rules also transferred from smaller source graphs to independent 100,000-node samples generated under matched conditions. These results demonstrate a recurrent graph-clustering formulation while identifying dependencies on graph quality, readout design, and source-target similarity.

The Linear Representation Hypothesis associates high-level concepts with directions in language models, but it remains unclear how these concept-related linear structures are organized within the model. We propose the Answer-Basin Representation Hypothesis: the probability measure induced over answers by the model's continuation distribution organizes these linear structures, with its statistics represented along linear directions shared across questions. All continuations yielding the same answer form an answer basin, whose mass is their total probability. These basin masses define the pushforward probability measure over answers. We posit that concept-related linear structure emerges from differences in the answer measure rather than being determined by changes in concept labels. Experiments across models and tasks link concept-consistent effects and their reversals in probing and steering to the alignment between concept labels and the answer measure.

Scalable simulation is essential for robot data generation, policy training, evaluation, and safe iteration, yet real-world interaction is costly and conventional simulators require labor-intensive construction. We present Uranus, a data-driven robot simulator built around a joint-trajectory-conditioned autoregressive diffusion model. Uranus offers three key capabilities: (1) streaming, open-ended rollout, which receives future joint-position trajectories online and autoregressively generates one latent frame per step, corresponding to four RGB frames, without a fixed horizon; (2) low-latency generation, achieving 24 FPS after inference optimization; and (3) scalable, extensible robot control, providing a unified interface for synchronized multi-view generation across diverse robot embodiments and camera configurations. We conduct comprehensive quantitative and qualitative evaluations on both in-distribution and out-of-distribution data, providing an objective assessment of Uranus and clearly identifying its current limitations. We release the code and model weights to empower the community with practical tools and insights.

Advances in processing power, camera technologies, and mobile image analysis have made smartphones and other mobile devices, such as laptops, increasingly suitable for medical diagnosis and healthcare applications. Researchers have developed low-cost solutions for the early detection and monitoring of various health conditions, including eye and ENT diseases, malnutrition, heart rate variability, skin and oral conditions, and injuries, using images captured by non-medical devices such as smartphones and webcams. This survey examines existing research on mobile image-based medical diagnosis, with an emphasis on its potential to enable low-cost and accessible healthcare. We comparatively analyze state-of-the-art solutions across different healthcare application categories, examining their advantages and limitations. Based on this analysis, we identify desirable characteristics of mobile image-based diagnostic tools and highlight areas where existing approaches have made progress as well as areas requiring further research. We also discuss application-specific and common challenges and outline directions for future research. Overall, this study provides a comprehensive overview of mobile image-based healthcare solutions and their potential to support low-cost disease diagnosis and monitoring, particularly for underserved populations in remote and resource-constrained settings.

Voice provides a natural and immediate interface for AI agents. Many settings in which voice agents could be useful, including meetings, households, and collaborative work, are inherently multi-speaker. Supporting these settings introduces challenges that are largely absent from one-on-one interaction. We introduce the Multi-Speaker Interaction Benchmark (MSI-Bench) for evaluating multi-speaker voice interaction. Each test case is a short multi-party multi-turn audio scene with participant context, expected tool calls, and atomic rubrics. The benchmark targets three capability families: multi-speaker memory, multi-speaker instruction following, and multi-speaker reasoning. It comprises 1,152 test cases, evenly split between Mandarin Chinese and English (576 each). The strongest configuration on each split passes all rubrics on only 66.8% of English and 54.5% of Mandarin cases, and the strongest open-weight configuration on 34.0% and 19.3%. Failure analysis separates perception from reasoning: open-weight models are bottlenecked by the multi-speaker audio front-end, while frontier systems still fail speaker-scoped decision making on clean transcripts---and models across the board often respond when no one has addressed them. These results identify speaker-grounded perception, speaker-scoped decision making, and conversational restraint as concrete targets for future voice agents.

Large language models (LLMs) are increasingly deployed in production systems, raising concerns about their exposure to adversarial manipulation through prompt injection and jailbreak attacks. Classifier-based guardrails, such as Prompt Guard 2, are widely used as a first line of defense against such attacks, but their internal decision logic is largely opaque to both defenders and attackers. This paper presents an exploratory case study that applies explainable artificial intelligence (XAI) techniques to analyze how Prompt Guard 2 distinguishes malicious from benign prompts. We conduct four experiments to probe this question empirically. Guided by Vanilla Gradient and SHAP attributions, we find that Prompt Guard 2's decisions rely on the cumulative contribution of many tokens rather than a few dominant ones, yet saliency-guided synonym substitution and sentence-level paraphrasing can flip its predictions while altering only a moderate fraction of the text, in some cases yielding a successful jailbreak against the underlying LLM. A dataset-scale saliency analysis further shows that undetected injection prompts systematically lack the lexical markers the classifier relies on. We discuss the implications of these findings for the design and evaluation of classifier-based guardrails, and argue that explanation methods intended to support transparency can simultaneously lower the cost of constructing successful adversarial bypasses.

Post-training quantization (PTQ) enables efficient deployment of large language models, and PTQ methods are usually optimized and evaluated with generic reconstruction, perplexity, or answer accuracy. But in explanation-critical domains, preserving only the final answer may be insufficient, since users may also inspect generated rationales to judge whether a prediction is trustworthy. We study this issue in medical multiple-choice question answering, where rationales should provide evidence that supports the selected answer. We propose an explanation-aware objective for transformation-based PTQ. Our method builds an offline faithfulness cache from full-precision teacher rationales and uses it during optimization to preserve answer-supporting evidence tokens and evidence-conditioned answer behavior. We instantiate it on OSTQuant under W4A4KV4 quantization and evaluate four 7B--8B medical and instruction-tuned LLMs on MedExQA, MedExpQA, and ChallengeClinicalQA. While a same-calibration OSTQuant baseline preserves task accuracy, it can substantially weaken answer-supporting rationales. Our objective is to preserve the full-precision model's answer-supporting behavior rather than improve gold-label accuracy, and our method better preserves the full-precision model's answer behavior and rationale-to-answer support. These results suggest that PTQ for explanation-critical settings should evaluate preservation of answer-supporting evidence, not only answer accuracy. Code and evaluation scripts are available at https://github.com/dut0817/EAQuant.

Linear RNNs based on the delta-rule enable efficient sequence modeling, but their linear updates with a low-rank correction constrain their expressivity. Prior work has shown that composing two delta-rule transitions in a single recurrent update can model a 2D rotation, but this increases the rank and the cost of the updates compared to a single transition. We show that Kimi Delta Attention (KDA) can realize 2D rotations by combining a single delta-rule transformation with a second reflection supplied by its channel-wise gate. This requires extending the parameter ranges of KDA by combining two existing range extensions: allowing gates in $[-1,1]$ and the delta-rule coefficient $β$ in $[0,2]$. We call the resulting model Complex KDA (CKDA). It preserves KDA's stability and efficiency, with transitions that remain diagonal-plus-rank-one and non-expansive, while reaching the state-tracking expressivity of DeltaProduct$_2$. We characterize the expressivity of CKDA and prove that every orthogonal diagonal-plus-rank-one matrix is exactly a CKDA transition matrix. A single CKDA layer can track every finite group isomorphic to a subgroup of $\mathrm{SO}(3)$, and many state-tracking results use one fewer layer for CKDA compared to other diagonal-plus-rank-one Linear RNNs. Empirically, combining both extensions yields the strongest length extrapolation among tested KDA range settings on $S_3$, $S_4$, and periodic audio continuation. In language modeling, CKDA outperforms Transformers and other linear RNNs, obtains similar results to a KDA baseline, and shows promising scaling behavior. Our code is open source at https://github.com/OpenEuroLLM/ComplexKDA and our models are available at https://huggingface.co/collections/openeurollm/complexkda.

Challenging behaviors including aggression, self-injury, and property destruction are observed in 68% of autistic youth and pose risks to youth and caregivers. These episodes are preceded by agitation, a rising state of distress expressed through movement, vocalization, and autonomic arousal. Its signs are subtle and individualized, and its autonomic components are invisible without instrumentation. We collected upper-body movement from inertial measurement units, physiology from a wrist-worn device, and vocalizations from lapel microphones across 30 clinician-led sessions with 15 autistic youth, paired with expert behavioral annotations. We adapt four pretrained foundation models, one per modality, project each to a shared 128-dimensional space, and fuse them into a single group model. The model detected agitation with an area under the ROC curve of 0.724 at the clinician-annotated onset (within-participant permutation p=0.0005), declining to 0.608 at 30,s before onset. Thirteen of fifteen participants were above chance. A from-scratch configuration reached only 0.58, while frozen and fine-tuned features performed comparably (0.71 and 0.72). Audio contributed most of the signal, and a watch-only configuration stayed near chance. Individualized agitation is therefore detectable, including in unannotated windows preceding the annotated onset, using foundation-model transfer with one shared model rather than one per child.

AI governance increasingly requires providers and public authorities to reason about multiple AI instantiations, alternative versions, and deployment configurations of multiple AI systems. Yet current regulation remains predominantly single-system-centric, acknowledging such multiplicity only sparsely without treating collections of related AI systems as governance objects. This creates an AI population governance problem: determining which instantiations can be meaningfully considered together and how their changing configurations can be represented and monitored. The first requirement has recently been addressed through trustworthiness-based accounts of AI identity. We address the second by introducing convex AI compositionality: a formal representation of the configurations generated by finite AI system populations that uses convex spaces. The core idea is that convex compositions of the operational states that a population of AI system instantiations may occupy over time are compatible with lifecycle reachability across the population and can preserve the formal identity relations between these systems. Well-known statistical and geometric constructions, such as weighted state distributions and convex hulls, become AI governance tools for distinguishing operational states, population weights, heterogeneity, and AI configuration change across different governance modes while remaining compatible, under stated conditions, with lifecycle reachability and AI identity. We illustrate our AI population governance framework through distributed healthcare deployments and controlled deployment of recruitment AI variants.

Localising artifacts in synthetic speech remains challenging, as most evaluation methods yield only global quality scores. This paper presents XSQ-AST, a framework that combines the SQ-AST speech quality model with WhisperX phoneme alignment and multiple saliency methods to produce temporally localised artifact diagnostics without model retraining. Saliency maps are projected onto continuous distributions via kernel density estimation and onto phoneme boundaries via phoneme-discretised saliency maps. A 40-participant listening test validated the framework across five perceptual dimensions. Attention Rollout, Attention Flow and an adapted GradCAM produced temporal distributions that correlated with listener highlights, with different methods best suited to different artifact types. An AUC-ROC analysis confirmed discrimination above chance.

Professional photo finishing relies on both global adjustments and region-specific local edits guided by semantic masks, yet current automated methods handle this workflow only partially. We present PrismGPT, a Vision-Language Model (VLM) framework that produces structured, region-aware editing plans from a single input image without relying on commercial black-box tools. Training a VLM to simultaneously diagnose aesthetic deficiencies at both global and local levels while predicting precise editing parameters is challenging due to the vast combinatorial decision space. We address this through proxy-guided learning: two simpler proxy tasks -- operation decomposition and region-aware aesthetic ranking -- teach the foundational skills the model needs, while a competence-based dynamic scheduler automatically rebalances the multi-task training ratio, progressively shifting emphasis from the proxy tasks to the primary editing task as each skill is mastered. Crucially, all reasoning traces used for supervised fine-tuning are self-synthesized by the same base model, eliminating the need for a stronger external teacher. Experiments on MIT-Adobe FiveK and SPIRE, a new professionally retouched benchmark we introduce, show that PrismGPT achieves state-of-the-art results while using only ~6% of the training data compared to the previous best method.

When facing complex problems, humans tend to try various ideas for different issues. Human thinking patterns exhibit remarkable flexibility in adapting to diverse scenarios. GPT-o1, GPT-o3, and DeepSeek-R1 adopt long chain-of-thought models to address complex problems by increasing reasoning depth, which default to a forward reasoning mode. We conducted statistical analysis on the accuracy of different mathematical problem datasets on models of different scales, and found five reasons for errors: Insufficient solution-space coverage, Computational mistakes, Unverified assumptions, Ignoring constraint conditions, Maximum response length limitation. To address the above issues, we proposed a backward reasoning pattern construction method aimed at enhancing the model's reverse thinking ability and dynamic adaptability. First, we constructed an easy-hard two-stage Math dataset for training large models and gradually improving their inference ability at different difficulty levels. The dataset contains forward reasoning paths as well as backward reasoning paths. And a two-stage supervised fine-tuning process is applied to progressively train the model's backward reasoning capability. Furthermore, a fine-grained reward mechanism is developed, employing smoothed reward signals to strengthen the model's ability to autonomously select thinking modes during the reasoning process, thereby avoiding reward hacking. A linear-decay balanced sampling strategy is designed to maintain a balance between forward and backward reasoning path samples during training, enabling the model to converge quickly and stably. Experimental results show that our method significantly improves reasoning efficiency and accuracy in tasks such as mathematical proofs, offering a flexible and efficient reasoning paradigm for solving complex problems.

This paper presents the design, optimization, implementation, and on-board validation of a neural processing unit (NPU) accelerator for real-time vehicle detection on the resource-constrained Xilinx Zynq XC7Z020 device of the PYNQ-Z1 board. The work follows a hardware/software co-design methodology that combines quantization-aware training (QAT), lightweight YOLO-derived detectors, Brevitas/QONNX model export, FINN dataflow compilation, Vivado implementation, and physical benchmarking on the target board. Four simultaneous engineering requirements define successful deployment: throughput above 30 frames/s (FPS), energy efficiency above 7 FPS/W, programmable-logic (PL) hardware latency below 50 ms, and Pascal VOC detection accuracy above 0.55 mAP@0.5. The design space includes LP-YOLO and LP-YOLO Slim variants, a custom YOLOv3-tiny reference, 4-bit and mixed low-bit quantization, 320$\times$320 and 256$\times$256 inputs, manual and automatic FIFO sizing, and programmable-logic clocks from 100 to 200 MHz. The final LP-YOLO Slim configuration uses a 256$\times$256 input, w2a4 quantization, and a 142.86 MHz PL clock. With batch 100 it reaches 35.66 FPS at 2.91 W, corresponding to 12.25 FPS/W, while measured PL latency is 45.11 ms and VOC mAP@0.5 is 0.594. This is the only evaluated configuration for which the supplied measurements satisfy all four requirements simultaneously. The results show that low-bit QAT, architectural slimming, FINN folding and FIFO optimization, and moderate clock scaling can jointly provide a practical real-time detector on a small Zynq FPGA.

Autonomous AI agents are increasingly deployed in areas where wrong decisions are hard to reverse. This paper examines schema mismatch: the condition in which an agent operates within an interpretive frame that no longer applies to the current context. Outputs produced under such a mismatch can appear internally consistent, linguistically plausible, and largely factually correct; output-quality metrics alone therefore capture the underlying loss of validity only partially. The paper introduces Epi-Logic, a conceptual framework for epistemic runtime control. It couples the detection of schema dissonance, a graduated reduction of autonomy, and the auditable switch to a validated schema. A schema is formalised as a tuple of variable space, expectation model, validity conditions, axioms, and metadata. The Epi-Score aggregates seven graded dimensions of epistemic dissonance; the temporal validity dimension D8, violations of the validity conditions G, and axiom violations are carried as separate categorical paths that are not offset against the aggregate. The architecture rests on a checking asymmetry: formalised validity conditions can be checked at runtime, whereas the correctness of many actions is established only ex post. The paper separates two architectural properties, a conditional result from sequential changepoint detection, and an empirical remainder. Eight falsifiable propositions with named baselines describe the transition to empirical validation. All propositions are empirically testable hypotheses, not established results.

We investigate next generation reservoir computing (NGRC) as a data-driven approach for inferring unseen components of dynamical systems. We compare NGRC with traditional reservoir computing (RC) using the Lorenz and Rössler system, where two unknown components are inferred from one given component. For both systems, NGRC achieves accurate results while requiring fewer training data and less computational time than RC. We identified an inverse proportional behavior between the number of time-delayed steps needed for NGRC and the temporal resolution, indicating that the physical time span covered by the delay interval is an important factor in determining the required number of delayed steps. Finally, we apply NGRC to the observational climate data of ENSO (El Niño--Southern Oscillation) and infer one observable from the remaining variables. Despite the noise and complexity of the real-world data, the NGRC shows promising results. Our findings demonstrate the potential of NGRC for efficient inference of unseen components in both controlled dynamical systems and real-world data.

The growing demand for real-time, data-driven decision-making in complex and dynamic systems is placing increasing pressure on traditional Operational Research (OR) methodologies. Reinforcement learning (RL) has emerged as a complementary approach, offering strong learning and computational capabilities for sequential decision-making in dynamic and uncertain environments. Recent research shows an increasing interest in integrating RL with OR to address dynamic decision-making problems, enhance heuristic and exact methods for combinatorial optimization, and support the development of digital replicas of operational systems. The overarching goal across these efforts is to leverage the learning capabilities of RL to strengthen traditional OR algorithms, improving solution quality, computational efficiency, and robustness. Given the diversity of integration approaches and application settings, there is a clear need for a systematic and technically detailed review of how RL empowers OR methods. To address this gap, this paper presents a structured review of three key roles that RL plays in empowering OR: (i) solving sequential decision-making problems in dynamic environments, (ii) serving as an end-to-end solution method or as a component integrated within heuristic and exact OR methods for combinatorial optimization problems, and (iii) facilitating extended reality analysis through integration with digital twin systems. We critically synthesize recent advances across these roles, highlighting their advantages, implementation requirements, limitations, and challenges. Finally, based on these insights, we outline a roadmap for future research to further advance the methodological and practical integration of RL and OR.

Latent world models predict the consequences of actions, but accurate prediction does not guarantee that latent distance reflects which candidate will execute successfully. We identify a decision-local prediction gap: among the few futures competing for execution, a candidate predicted closer to the goal can produce a worse realized outcome than an available alternative. We introduce D-JEPA, a decision-aligned latent world model that learns decision-relevant relations among candidate futures from executed outcomes. A bounded, permutation-equivariant operator jointly reasons over goal-relative predictive features and ordinal evidence, refining pretrained predictive geometry where action choices are most consequential. Restricted predictor adaptation and a shared ordinal interface extend this alignment across complementary predictive geometries. D-JEPA further realizes the learned decision structure in JEPA-compatible future representations, enabling deployment through native latent-distance planning. Evaluations across latent control, manipulation, pretrained action-producing models, physical robots and autonomous driving demonstrate improved action selection, including 87.89% success on PushT, a 15.04-point average gain on RoboTwin, and a 17-point gain on physical robot tasks. These results establish decision-relevant relational structure as a direct bridge between predictive world modeling and effective control.

Existing approaches to solving differential equations, such as symbolic regression, physics informed neural networks, and neural operators, typically focus on numerical approximations or blind symbolic search via fitting to numerical data. Less attention has been paid to learning structured representations of mathematical expressions that preserve commutative properties and could support mathematical reasoning in symbolic forms. Transformer models have shown strong capabilities in solving symbolic differential equations. However, standard positional embeddings in transformers are designed for sequence data. Symbolic differential equations are naturally represented by expression trees, so these positional embeddings may not efficiently capture their hierarchical structures. We investigate existing tree positional embeddings in symbolic ordinary differential equation (ODE) tasks. We systematically study their effectiveness under different settings. Our results show that tree positional embeddings aid learning in early epochs and continue to improve performance throughout, ultimately yielding consistent advantages across various data sizes and tasks. Based on learned structural representations, we apply contrastive learning to support the commutative property in mathematics. Ablation studies provide insight into how these methods interact in modelling symbolic mathematical structures.

Language agents solve complex tasks through plans and actions. A single step the world refuses puts the goal out of reach, and what the agent does next decides the task. Prompted planners fail at exactly this point, rewriting the refused step in new words, meeting the same refusal, and burning the attempt budget without moving. They fail because the plan was never tied to the world, so a refusal has nothing in the plan to attach to. A world is where a task runs, and it has its own rules, its own admissible actions, and its own constraints. We build synthetic worlds across 7 domains and extract training data from them. A program enforces each world's rules and grades its goal, and every world is admitted only if its goal is reachable from its initial state. Agents run inside and leave verified failures paired with repairs that carried the run to a state the world certified, a record of about 226K trajectories. On this record we train the World State Generator, a model that writes a plan as checkable states of the world and keeps that plan aligned with the world it runs in. That alignment is what a plan written in language lacks, since the world it runs in has physical limits, logical dependencies, and required orders the language never states, and the plan encounters these rules only when a state fails. WSG takes that failure as the rule the world has stated and rewrites the remaining states to obey it, so the plan bends to the world as the run goes on. Across 7 public benchmarks, WSG raises end-to-end success for two open models near 30B parameters over prompting and brings to the level of proprietary model.

Humanoid robots are increasingly being popular and developed for human-centered applications, yet their ability to provide intelligent conversations and natural interactive knowledge assistance remains constrained by traditional rule-based dialogue systems, pre-defined responses and limited knowledge repositories. Large language models (LLMs) have emerged as a powerful foundation for enabling natural, adaptive, and context-aware Human-Robot Interaction (HRI), which provides a significant opportunity to address such limitations by enabling robots to understand natural speech language, reason over complicated queries, maintain high-quality conversational context, and generate knowledge-rich responses. In this work, we originally present and implement an LLM-based versatile Conversational AI Knowledge Assistant for the Raspberry-Pi-powered 13-Axis MyBuddy humanoid robot, which integrates LLM-driven language understanding and AI reasoning with real-time speech recognition, knowledge retrieval via extensible access of internet engines (e.g., Wikipedia, arXiv), flexible dialogue management, and natural speech synthesis to enable much more intelligent multi-turn continuous conversations and advanced emotional-support Human-Robot Interaction.

We develop an exact linear programming (LP) formulation for bounded-depth classification trees with binary features, using a junction-tree representation. The formulation is integral and supports recursive subtree optimization. Exact reductions make the model smaller while preserving the optimal value and recovery of an optimal tree. The reduced model supports two solution methods: column generation and message passing. Column generation solves integral restricted LPs and uses bounds over the full feasible domain to certify optimality. Message passing recursively combines optimal subtree costs. Both methods solve common subtree problems that, once the preceding tree decisions are fixed, can be evaluated independently and in parallel. Computational experiments show that the exact reductions substantially reduce the size of the junction-tree formulation. The resulting linear programming formulation certifies instances for which the tested mixed-integer formulation does not establish optimality within the same computational budget, while the column-generation and message-passing methods certify more instances and achieve an order-of-magnitude reduction in geometric-mean runtime relative to an existing state-of-the-art exact method for optimal classification trees.

Mitotic figure (MF) analysis supports tumor grading and prognostic assessment, but automated models remain sensitive to differences in tissue type and image acquisition. We present MiTHras, a task-specific pretraining framework that combines pseudo-label-guided image- and token-level contrastive learning with masked reconstruction. We construct TCGA-MF-Pseudo, a corpus of 1.8 million cell-centered images from 14 TCGA cohorts spanning 11 organ sites. Comprehensive evaluation on MF classification, detection, count-based survival prediction, and subtype classification demonstrates the efficacy of MiTHras. It achieves the highest mean F1 on all three MF classification benchmarks and both subtype benchmarks. MiTHras also outperforms general-purpose and pathology foundation encoders by a larger margin under frozen-encoder linear probing than under full fine-tuning. Although detection gains are modest due to a shared candidate-detection stage, ablations confirm that token-level supervision improves typical-versus-atypical classification and linear probing. These findings establish that MiTHras yields robust, transferable representations for automated mitotic activity assessment.

Head-and-neck anatomy changes over a six-to-seven-week proton course, and the anatomy of a later week cannot be imaged when the plan is made. We present a digital-twin framework that forecasts a patient's treatment-day anatomy as an ensemble of predicted CTs with propagated contours and quantifies the uncertainty of the forecast contours. The twin is a library of previously treated patients with planning and weekly quality-assurance CTs (QACTs), made patient-specific by a two-step foundation-model deformable registration: a cross-patient field carries each library patient onto the current patient, and a longitudinal field, estimated in the current patient's frame, carries that patient's planning-to-QACT change onto the current patient's own planning CT. A library of 302 observations from 88 patients yields about 300 replicates per patient, each a deformation that occurred in a treated patient. The dispersion of the propagated contours, resolved by outward normal, is six-direction contour uncertainty in millimeters. This is uncertainty in the input to the forecast, which library patient the current patient follows, rather than in model parameters, and it is unchanged when the registration engine is exchanged. On ten patients with clinician contours on two QACTs, the library alone fixes the anisotropic shape of the uncertainty (4.5 to 6.2 mm); the first QACT narrows it by a factor of 3.2 to 3.6 without a contour being drawn; an approved contour improves the center but not the width. The estimate orders directions correctly but is not Gaussian-calibrated. A clinical target volume expansion is worked out as one application.

While a centralized approach involving patient consent to collect and analyze data centrally would theoretically offer the best data quality and predictive performance, it is not always feasible in practice. Federated Learning (FL) architectures have shown to be a very promising approach to use and access distributed disease related resources within the GDPR boundaries. In a previous case report, we described the preconditions at the participating sites and necessary administrative and process related steps to prepare data, people and infrastructure for improving subtype identification and assessing treatment options in pancreatic cancer. We update this report sharing our experience in tackling the challenges and show preliminary results of the actual federated learning AI pipelines. At the participating sites, we have to identify and annotate the data being accessible after extraction and transformation in a local FL hub - in our case a centrally developed and distributively deployed Docker container. This container comprises the FL scripts generating local models. We apply a newly developed FL algorithm considering all local features, including partial overlapping features specific to the local sites. Theoretically, an annotation in a cancer setting should succeed using the German oncology core data set (oBDS), which is already utilized for mandatory reporting to cancer registries, and can be sustained in the FL setting. The FL algorithms deal robustly with partially overlapping features as we showed with public data sets. Major roadblocks including straightening operational concepts for the infrastructures, ethics approval for such novel architectures and support for every site have been addressed. However, scaling up this approach in the future faces hurdles; while including broader multi-modal data sets should be feasible, large-scale deployment to more sites remains challenging.

Large Language Models (LLMs) are increasingly used in cybersecurity, where accurate analysis often requires multi-step and context-dependent reasoning over complex and heterogeneous data. However, existing prompting approaches typically focus on eliciting reasoning without explicitly considering how intermediate reasoning steps are structurally organized. We introduce Security Reasoning Topology, which models reasoning through three representative structures: Linear, Branching, and Graph. To evaluate their effects, we conduct controlled experiments on three cybersecurity datasets covering MITRE ATT&CK network traffic, cyber threat intelligence (CTI), and CVE vulnerability analysis. We evaluate multiple LLMs, including Llama 2 (7B, 13B, 70B), GPT-5.1, and Mistral Large 3, while keeping task inputs consistent and controlling reasoning structure through system-level prompting. Results show that reasoning topology substantially affects performance: Graph reasoning achieves the highest overall accuracy, improving over few-shot prompting by 9.8-12.2 percentage points across datasets, while Branching provides a strong intermediate solution. The results further show that the effect of reasoning topology remains consistent across model families and scales, highlighting reasoning topology as an important design factor for LLM-based cybersecurity analysis.

Autobiographical remembering supports identity, well-being, and social connection in later life, yet voice-based memory technologies largely rely on isolated prompts. We designed and built MeBo, a fully functional relational voice-based memory companion, through participatory design with 11 older adults. Their accounts shaped four Design Strategies that guided MeBo's interaction design and multi-agent implementation. In a mixed-methods evaluation with 20 older adults, participants found MeBo exceptionally usable (SUS = 87.75), enjoyable, sociable, emotionally responsive, and trustworthy. Participants reported higher positive affect and momentary social connection and lower negative affect after the session than before. Participants described how MeBo followed their stories, returned to earlier memories, adapted to their preferences, and made its growing memory visible and controllable. MeBo's relational framing surfaces tensions around what it should remember, who may access memories produced through interaction, and what becomes of them when the user or MeBo is no longer present.

Data-driven protection may complement conventional relays in distribution grids whose operating conditions vary with distributed generation, switching events, and changing short-circuit levels. We study line-selective tripping from static trajectories of a realistically simulated CIGRE medium-voltage network using offline reinforcement learning. A convolutional Q-network receives causal voltage-current phasor and apparent-impedance features, optionally together with raw waveforms, and is trained with conservative Q-learning (CQL). A controlled sensitivity study evaluates two observation windows, reward variants, and three CQL weights under a common split and training protocol; one exploratory post-hoc run additionally increases the discount factor from $γ$=0.95 to 0.99. On 225 held-out episodes, the best per-timestep result is obtained with combined input and CQL weight $α$=0.9, reaching precision 0.9993, recall 0.9496, and F1-score 0.9738. Because dense per-timestep scores do not encode the terminal semantics of relay operation, we also evaluate the first non-wait action in each episode. The default combined-input agent selects the correct line-trip action first in 98.13% of 214 fault episodes, but trips in 72.73% of the 11 non-fault episodes. In the post-hoc run, the corresponding rates are 98.60% and 54.55%, respectively. The results show that dense predictive performance and terminal protection behavior can lead to different model rankings. Offline CQL therefore demonstrates strong faulted-line selection on the simulated fault episodes, while the static trajectories, small non-fault set, and single-seed post-hoc design preclude conclusions about practical relay security or deployment readiness.

Repeated execution of the target model during autoregressive decoding is a major source of LLM inference latency. Unlike linear speculation, which follows a single candidate chain, tree-structured speculation retains multiple branches from shared prefixes; under the same budget, this broader coverage can improve acceptance and efficiency. Adapting it to DeepSeek-V4 is nontrivial: its CSA/HCA online compressed attention concentrates the difficulty on the target-verify side, where branches diverging from a shared prefix compress into different states, breaking cross-branch state consistency. We integrate tree-structured speculative decoding into the DeepSeek-V4-Flash pipeline via branch-aware causal verification, temporary state isolation, and accepted-path state refresh, keeping verification and compressed-state updates consistent across branches. Across budgets D=5 to D=8, batch sizes 1 to 64, and three datasets (GSM8K, MBPP, ShareGPT), tree speculation achieves a higher accepted length than the matched linear configurations in all settings (e.g., at D=8 about 2.83--3.41 versus 2.39--2.84) and improves throughput in nearly all configurations---marginal only at the smallest budget---by up to about 18.5%. More importantly, the gains follow stable, transferable regularities: the relative gain grows with the budget and is most pronounced for less predictable workloads at small-to-medium batch sizes, while beyond a certain budget throughput plateaus and decouples from the still-rising accepted length. These results show that retaining multiple candidate paths under the same budget can effectively improve DeepSeek-V4 decoding efficiency, and offer experience for adapting speculative decoding to future models with compressed, sparse, or structured context representations.

Latent diffusion models now dominate medical image generation, and every such pipeline rests on a \emph{tokenizer} that compresses images into the latent codes for image generation to operate on. Thereby, the tokenizer choice bounds every downstream task from reconstruction fidelity and generation quality to the representations available for downstream analysis. Yet, medical imaging pipelines routinely utilize tokenizers from natural imaging on the hypothesis that their behavior carries over. However, this is an assumption never tested in the medical imaging regime, where datasets are orders of magnitude smaller and images exhibit far lower inter-sample variance. We present a systematic evaluation of medical image tokenizers evaluating thirty configurations across ten model families on twelve datasets at three compression factors, spanning reconstruction, generation, latent geometry, downstream classification, and memorization. We find that (1) performance on image reconstruction and generation strongly correlate, unlike prior reports on natural images; (2) modern tokenizers use nearly all of their codebook entries, but still leave most of the latent space unused; (3) training-set memorization is mild and is further suppressed by stronger latent space compression; and (4) discrete quantization can largely preserve downstream classification, with lookup-free schemes being the main exception.

Angular-margin speaker encoders are widely used in voice conversion, yet the geometry of their classifier prototypes remains poorly understood. We analyze ECAPA-TDNN classifier prototypes as points on the unit hypersphere and characterize their organization using rotation-invariant angular statistics together with global and local effective dimensionality measures. Our analysis shows that standard training can induce angular concentration and a substantial reduction in effective dimensionality. To address this, we investigate two geometric regularization strategies (hinged Riesz log-energy and effective-dimension maximization) applied to classifier prototypes to encourage more uniform hyperspherical coverage. The resulting prototype sets exhibit higher effective dimensionality and improved isotropy, with configuration-dependent effects on speaker-recognition performance. When the corresponding ECAPA-TDNN models are used as speaker encoders for Fast-VGAN, the regularized systems also exhibit improved robustness in zero-shot voice conversion, particularly for previously unseen speakers.

We consider the recovery of low-multilinear-rank tensors from linear measurements and propose an adaptive block-weighted modewise Riemannian gradient descent method. The method combines memory-efficient modewise measurements with a normalized adaptive weighting strategy for the core and factor components of the Riemannian gradient. The weighting improves convergence without increasing the multilinear-rank bound of the search direction or the size of the reduced core used for retraction. Under the tensor restricted isometry property and a suitable initialization, we establish local linear convergence and derive sampling guarantees for sub-Gaussian and subsampled orthogonal with random sign (SORS) measurements. Numerical experiments on synthetic low-Tucker-rank tensors show that the proposed method reduces iteration counts and computational time while maintaining reliable recovery performance, especially near the recovery threshold and for structured SORS measurements.

Continual learning with Low-Rank Adapters (LoRA) typically mitigates forgetting by penalizing the overlap between a new update and the accumulated past weights, which discourages certain update directions without controlling how an update distributes its energy over the ones that remain. We ask whether that restriction has to be task-aware, or whether a generic one supplied by the optimizer is enough. We train a plain incremental LoRA (IncLoRA) with Muon, which orthogonalizes each update, and compare it against O-LoRA and ELLA over five seeds and three task orders on the Standard CL Benchmark and three seeds on TRACE. IncLoRA+Muon reaches the accuracy band of the dedicated methods on Standard CL and improves on every AdamW configuration on TRACE. One update-constraining mechanism is enough, whether it comes from the loss or from the optimizer; on Standard CL a second one does not help, and for the most restrictive method it costs 8.4 points of accuracy and the plasticity to fit each task. What separates the two optimizers is not the size of the update, which under Muon is 0.91 to 2.06 times that under AdamW, but how it is distributed. AdamW confines it to between 1.4 and 1.8 effective singular directions, Muon spreads it over 7.0, and the two do not overlap in any tracked run. Part of the advantage usually attributed to dedicated CL methods may therefore be explained by the geometry of the optimizer's updates.

Large language models (LLMs) are increasingly used to make predictions from numerical time-series histories and textual events. Yet accuracy alone cannot reveal whether correct answers reflect effective integration of the two inputs or instead arise from event polarity, unimodal priors, or superficial cues. Likewise, plausible explanations may rationalize predictions without faithfully reflecting the evidence that drives model behavior. We introduce TimeLitmus, a diagnostic benchmark for cross-modal understanding and explanation faithfulness in event-conditioned time-series prediction. TimeLitmus contains 4,856 evaluation records across Finance and Traffic, combining natural prediction with controlled counterfactual and contrastive interventions, explanation-targeted faithfulness tests, and systematic shortcut controls. Across ten representative LLMs, standard prediction accuracy substantially overstates reliable cross-modal understanding: Hard Paired Contrast (HPC) pair correctness peaks at only 19.2% in Finance and 11.7% in Traffic, and all ten models show lower-than-expected consistency on Finance series-side controls. Models often recognize scenario relations explicitly yet fail to apply them during independent prediction. Explanation faithfulness shows a similar gap: in Traffic, most models cite the manipulated temporal factor in over 90% of cases, while behavioral support remains below 22%. Human annotators outperform LLMs on matched controlled and hard-pair diagnostics, confirming that these distinctions are recoverable from the inputs. Natural-only adaptation yields selective gains in evidence selection and input sensitivity, but not consistent gains in controlled or hard-pair behavior. The benchmark, evaluation suite, and supervised adaptation data will be released publicly.

Providing autonomous intelligence, pervasive connectivity and usability to human life and industry has led to the emergence of the Internet of Things (IoT). To support time-sensitive and resource-constrained applications, IoT systems nowadays increasingly rely on edge computing. This brings computation and decision-making closer to end devices. In edge-enabled IoT architecture, latency and communication overhead are reduced, but interactions among a larger and more diverse set of devices, edge nodes, services, and data sources are introduced as well. In such environments, security and privacy mechanisms provide the foundation for protection, while trust management can assess the reliability of interacting entities and adapting secure decisions. In this paper, we systematically review the current state of trust management in edge-enabled IoT. To this end, we propose a comprehensive taxonomy that maps physical, network, and application architectural IoT layers against the consumer, commercial, industrial, and infrastructure IoT domains. We further investigate state-of-art research based on their trust design, how trust integrated into secure IoT operations, the attacks that effect trust management process. Based on these findings, we identify key gaps in current research and outline future directions for context-aware and adaptive trust management in edge-enabled IoT.

Self-evolving agents convert interaction feedback into persistent artifacts, such as memories or skills, which in turn guide subsequent decisions. As these artifacts are iteratively updated throughout an experience stream, the capabilities they support may evolve. Consequently, endpoint performance alone offers an incomplete view of self-evolution. Process-level evaluation is therefore essential to identify when a target capability emerges and whether later updates strengthen, preserve, or weaken it. Motivated by this, we propose \textsc{EvoPathBench}, a benchmark that tracks individual capabilities during artifact-level self-evolution. EvoPathBench fixes the base model, tools, freezes evolving artifacts at successive checkpoints, and evaluates the target capability on held-out episodes. This benchmark evaluates agent self-evolution using public trading data and calibrated trajectories. It tests three capabilities: generalization to unseen tasks, retention after unrelated learning, and rule adaptation to new evidence. Experimental results show that gains on similar unseen tasks often weaken under distribution shift, retention losses are concentrated in a minority of evolution paths, and no method achieves reliable rule adaptation. Moreover, while self-evolution enables agents to generate candidate artifacts with substantial held-out gains, the selected updates consistently fall short of realizing this potential. Together, these findings establish capability-level process evaluation as a foundation for analyzing self-evolution, identifying candidate evaluation and selection as key targets for improvement.

LLM-based agents increasingly operate in environments where they interact with users, tools, and external systems. Yet most security evaluations assume passive users and static control, ignoring the interactive dynamics that shape real agent behavior. We introduce \textbf{DUMA-Bench}, a benchmark and evaluation protocol for measuring agent security under \emph{dual-control} interaction, where both the agent and the user can influence the shared environment state. DUMA-Bench extends $τ^2$-bench ~\cite{barres2025tau} with adversarial environments covering eight vulnerability classes, including RAG poisoning, cross-agent manipulation, and unsafe output handling. We evaluate \textbf{14 models from five model families} (OpenAI, Anthropic, DeepSeek, Qwen, and Z.ai) across eight domains and multiple user-behavior regimes. Across our experiments, introducing dual-control interaction increases the attack success rate from \textbf{26.9\%} to \textbf{41.1\%}. These results show that agent security is not solely a property of the model but emerges from the interaction between the model, the user, and the environment. DUMA-Bench provides a missing evaluation layer for studying security in realistic agent deployments.

Human demonstrations offer a scalable way to collect manipulation data, but their contacts may be unstable or infeasible when transferred to a robot hand. Collecting demonstrations directly on the target robot avoids this mismatch, but substantially increases the cost of data collection. To address this trade-off, we present \textbf{Touch2Robot}, a framework that lets humans collect demonstrations while seeing how the target robot hand would contact the object. We capture human hand motion, tactile-glove measurements, and object motion during human manipulation. These recordings guide object-specific RL policies to reproduce the demonstrated object motion while favoring contacts consistent with the recorded human touch. We distill the learned behaviors into a unified real-time retargeter that maps incoming human observations and object geometry to robot hand configurations. During collection, the predicted robot configuration is synchronized with the tracked object pose in simulation to reconstruct robot-object contacts, which are visualized to help the demonstrator adapt subsequent interactions to the target hand. Across four real-world tasks, Touch2Robot improves average real-robot replay completion from 37.9\% to 72.1\% over visual-only feedback, while reducing the collection time per replay-successful demonstration from 58.6~s to 18.2~s. Reconstructed target-hand contacts achieve 44.2\% F1 against real-robot tactile measurements, and policies trained on Touch2Robot demonstrations improve downstream Diffusion Policy performance by 29.1 percentage points over visual-only feedback. These results show that bringing robot touch into the human demonstration loop improves both the quality and efficiency of scalable dexterous data collection. \textit{Project webpage: \href{https://Touch2Robot.github.io/}{https://Touch2Robot.github.io/}.}

Language models can be adapted by changing the computations applied to individual tokens. Quantum circuits offer one such approach, but evaluating wider circuits inside a large model can be computationally demanding. Here we introduce HyperQ, which adds token-conditioned quantum residual branches to a frozen masked-diffusion language model. A quantum residual branch is a module in each transformer block that reads a token's hidden state, emits the coordinates of that token's circuit, executes it, and adds the measured values back through a residual connection. The backbone remains frozen, and only the added branches are trained. Within each branch, a lightweight circuit hypernetwork emits token-specific rotation angles, coupling strengths, and measurement axes in a shared sparse circuit structure. The required expectation values have an exact classical expression whose evaluation cost grows linearly with the qubit count, enabling circuits from 16 to 64 qubits to be trained within a 1.1-billion-parameter backbone. Across downstream benchmarks, increasing circuit width raises the average score from 47.65 to 54.30. At 64 qubits, HyperQ exceeds the backbone and its low-rank-adapted counterpart by 4.71 and 3.67 points, respectively. HyperQ is fine-tuned on 20,000 prompt-response pairs, compared with 200,000 for the classical baselines. These findings support token-conditioned circuit emission as a tractable architectural approach to quantum-augmented language modelling.

Diffusion and flow models, as promising generative paradigms for speech enhancement, face a training--inference mismatch: training uses analytical path states, whereas inference recursively evaluates models on self-generated rollout states along discretized sampling trajectories. This mismatch causes prediction and discretization errors to accumulate. To address it, we introduce Corrective Forcing (CoF), a post-training paradigm that forces diffusion and flow models to learn from self-generated rollouts and correct their predictions. CoF corrects clean-speech predictions on rollout states toward the ground truth under dynamic sampling schedules, exposing the model to varying inference conditions. It further regularizes local evolution using locally corrected counterfactual transitions as references for factual transitions. By expressing model outputs through a shared clean-speech prediction parameterization, CoF applies the same post-training objective across diffusion and flow formulations. Experiments with SB-VE and OT-CFM demonstrate improvements in perceptual quality and reconstruction fidelity, together with robust performance across different numbers of sampling steps.

Readability assessment is essential for tailoring texts to intended audiences across educational, healthcare, and information retrieval domains. However, traditional readability formulas struggle to generalize across genres and languages, while supervised machine learning models rely on scarce, domain-specific annotated corpora, limiting their applicability--particularly for less-resourced languages. Large Language Models (LLMs) offer a highly scalable, multilingual alternative that requires no task-specific training, yet the impact of advanced prompting strategies on their performance remains underexplored. In this paper, we conduct a systematic benchmark of diverse open-source LLMs for multilingual readability assessment, focusing on the prediction of discrete readability levels required by educational frameworks. In addition to English, we evaluate our approach on a less-resourced language, Slovenian, to establish whether LLMs remain effective in low-resource settings. Specifically, we investigate the influence of explicit reasoning, demonstrating that Chain-of-Thought (CoT) prompting and reasoning-oriented models yield significant improvements over direct answering. Furthermore, our exploration of few-shot in-context learning reveals that providing just one labelled example per category (1-shot) substantially enhances prediction quality compared to zero-shot settings, with additional examples offering diminishing returns. By comprehensively comparing these approaches against traditional unsupervised metrics and state-of-the-art supervised baselines, we establish the viability of out-of-the-box LLMs as robust, cross-lingual readability assessors.

On-policy self-distillation fine-tuning (SDFT) learns new skills from demonstrations while reducing forgetting, but it always distils toward the full demonstration-conditioned teacher. This fixes teacher influence at the full-teacher endpoint, providing no control over how much demonstration information should be transferred at each prediction state. We introduce Information-Proximal SDFT (iSDFT), which instead treats the teacher as a budgeted source of information. At each token, iSDFT selects the distribution closest to the current student that satisfies a prescribed teacher-information constraint, yielding a closed-form exponential target with a locally determined tilt. To control cumulative drift, we further anchor the student to its frozen base policy. Across four heterogeneous LLM backbones and two specialisation tasks, iSDFT improves vanilla SDFT in 7 of 8 model-task settings and matches it in the remaining one. It also provides tighter retention on the original SDFT benchmark suite, with 73% of evaluations remaining within 0.5 points of the base model versus 52% for the strongest baseline, while achieving the largest mean improvement on all ten additional mathematics, coding, and competition-mathematics benchmarks. These results show that controlling how much and when teacher information is introduced improves specialisation while preserving broader capability.

As people turn to generative AI for financial advice, these systems can personalize how they communicate and what they say. Whether these forms of personalization shape decisions differently remains unclear. We conducted a preregistered 2 x 2 between-subjects factorial experiment (N=240): participants ranked three comparably viable stocks, discussed them with an AI, and reranked them. Participants perceived both forms of personalization, but only context-based personalization reliably changed ranking behavior: it increased reconsideration and moved rankings toward the AI's assigned recommendation. Participants felt more influenced without judging the AI as more correct, trustworthy, intelligent, likeable, or high-quality. Those initially farther from its recommendation moved more toward it while judging its advice less correct; exploratory analyses suggest greater susceptibility among lower-expertise participants. These findings show how personalized AI can steer decisions among defensible options with only a minimal evaluative trace, raising concerns for the design and governance of personalized decision support.

When a language model reads an operation such as "Swap the contents of Box F and Box B", its forward pass writes keys and values for those tokens into the KV cache. Prior work on entity tracking establishes what models use: bindings are resolved at query time rather than stored as explicit latent state. We ask what they write at the operation span and how it is accessed. We split a forward pass into a frozen writer and a reader: the writer's cache is recomputed without gradients, while the reader sees only the instruction and operation tokens, with all state descriptions hidden, and is trained in isolation. Anything the reader recovers was therefore already present in the unmodified cache. On a synthetic boxes task, a base reader recovers $\leq 0.06$ of queried bindings against $0.75$--$1.00$ after training, and recoverability tracks the operation's read/write footprint. We find two modes of access. Across Llama-3.1-8B and Mistral-7B, operation-span transplants causally redirect which visible state is read even when the two worlds hold identical values, revealing a routing record. Isolation training preserves routing and adds direct access to the payload, the value the operation read, from the single operand-name token in a narrow mid-depth band (layers 12--15 of 32 in Llama-3.1-8B, 14--17 in Mistral-7B) --- the same site that holds the routing record. The same recipe extends to further operations, ToMi and GSM8K, but is bounded by training coverage and costs open-book accuracy. Operation tokens thus leave localized, causally recoverable records that support both routing and direct payload access, though the model that writes them reads mainly the address they carry and not the value.

Autonomous parking in nonconvex and narrow environments remains challenging. Although optimal-control methods can explicitly enforce vehicle dynamics and collision constraints, nonconvexity compromises solver robustness and can cause failures. Large language models (LLMs) exhibit strong semantic reasoning capabilities, but directly generating dense trajectories makes it difficult to guarantee physical feasibility. We introduce SE-LLM-OCP, a unified framework in which LLMs make high-level discrete maneuver decisions, while an optimal-control module enforces low-level vehicle dynamics and collision constraints. Online, the LLM proposes sparse maneuver plans, decomposing the parking task into a sequence of short-horizon trajectory-optimization problems. A low-level solver then sequentially solves optimal-control problems. If the solver fails, the LLM aggregates failure evidence from the solver and validation stages to guide replanning. Offline, SE-LLM-OCP automatically evolves a structured decision-making knowledge base from scratch, driven by accumulated online failures. We validate our proposed framework in simulation on a car-like vehicle model and on a differential-drive robot. Our experimental results show that SE-LLM-OCP enables safer autonomous parking in narrow scenarios and demonstrates transfer of the same maneuver representation to a different kinematic platform.

A/B testing requires large sample sizes, long timelines, and significant costs. When auxiliary predictions of experimental outcomes are available from machine learning models, uncertain prediction quality precludes replacing human experiments entirely, yet these predictions may still contain useful signal. We propose a principled framework for learning-augmented hypothesis testing that leverages predictions of unknown quality to reduce sample sizes while maintaining statistical validity. Predictions naturally vary in granularity, from coarse aggregate signals to fine-grained individual-level estimates, and our framework addresses both ends of this spectrum: (1) for population-level directional predictions, where only a binary signal on the treatment effect sign is available, we use an asymmetric test and prove consistency and robustness bounds within the learning-augmented algorithms paradigm; (2) for individual-level predictions, we introduce Generalized PPI++ (GPPI), extending Prediction-Powered Inference to handle nonlinear prediction errors through higher-dimensional transformations. Both methods benefit from accurate predictions while remaining robust to inaccurate or adversarial ones. We validate our framework using persona-based LLM simulations, where AI agents equipped with user personas predict individual behavior, as a natural prediction source spanning both granularity levels. Experiments on four real-world datasets demonstrate that our methods, combined with persona-based predictions, substantially reduce experimental costs while preserving rigorous statistical validity.

Multi-organ segmentation using deep learning requires large amounts of annotated patient data; however, institutions often lack sufficiently large and diverse annotated datasets. Privacy constraints further prevent institutions from sharing patient data to overcome this limitation. Moreover, due to the labor-intensive nature of annotation and the scarcity of diverse expertise, institutions typically have labels for only a small portion of their local data, leaving the larger unlabeled portion unused. In this work, we propose a flexible semi-supervised federated multi-task student-teacher framework that leverages federated learning (FL) to improve multi-organ segmentation using both labeled and unlabeled data across participating sites. At each communication round, the proposed framework initiates local training, where clients with labels for the same task form a federation to produce an aggregated teacher model. The resulting teachers generate task-specific features for all data at each client. Subsequently, all clients form a second federation to train a multi-task student model with a shared encoder and task-specific decoders that replicate the teacher-generated features across all segmentation tasks. The aggregated student model is then used to update the local teachers and initiate the next training round. Extensive experiments demonstrated the effectiveness of the proposed method compared with local and federated single-organ models, yielding an average performance gain of 13 percent across clients. The experiments also demonstrated the impact of multi-task learning and unlabeled data and the applicability of the framework in relaxing labeled-data requirements for client participation. The code is available at https://github.com/AshknMrd/FedMust.

How should natural language processing models be selected and adapted for global health literature in environments where annotated data and computational resources are limited? This thesis investigates these challenges through experiments on semantic tag discovery, named entity recognition (NER), and multi-label topic classification. First, skip-gram word2vec models trained on progressively larger specialized corpora are compared with BioWordVec to assess how corpus size and domain context influence tag discovery. Vocabulary coverage and qualitative evaluation indicate that broader coverage does not necessarily yield more useful domain-specific associations. The analysis then turns to entity extraction, comparing convolutional spaCy models with a RoBERTa-based transformer on 1,000 annotated sentences. Under a lenient scoring protocol, the transformer achieves 0.80 micro-F1 versus 0.65-0.69 for convolutional models, but takes 82 seconds rather than 5-6 seconds. This trade-off motivates fine-tuning convolutional models and integrating a disease recognizer that achieves 81.33% test F1 on the NCBI Disease Corpus. Combined with PDF preprocessing, entity filtering, and MeSH enrichment, the resulting pipeline supports document-level indexing. To complement entity extraction with thematic annotation, MiniLM-based few-shot classification is compared with BART-MNLI zero-shot inference across 50 topics and 1,000 handcrafted test sentences. BART-MNLI achieves 95.2% single-label accuracy versus 59%; reported multi-label accuracies are 88% and 32% under partly manual assessment. However, its higher inference cost limits practical integration. The results show where domain specialization and lightweight adaptation offer practical value, and where transformer accuracy justifies higher inference costs, providing an empirical basis for building knowledge systems under resource constraints.

Tactile sensing is increasingly being incorporated into learning-based robotic manipulation, yet many existing approaches rely on spatially distributed sensors. Here we introduce SpectRobot, a framework that transforms single-point tactile signals into compact time-frequency spectrograms. These spectrograms encode high-bandwidth tactile histories as fixed-size image-like representations. They can be processed by standard vision encoders and integrated into learning pipelines originally developed for vision, while preserving temporal and frequency information unavailable to conventional cameras. Rather than increasing spatial density through arrays of tactile elements, SpectRobot exploits the rich dynamics contained in sparse, high-bandwidth single-point measurements. In our implementation, the sensors are mounted away from the contact surface while remaining mechanically coupled to it, reducing direct exposure to wear and potentially improving robustness in harsh environments and for long-term deployment on dexterous robots. Our experiments demonstrate that: (1) a robot can exploit single-point vibration signals to solve a visually occluded manipulation task; (2) temporal history strongly influences policy performance, while sensing bandwidth controls the spectral information available, with measurements extending to 100~kHz; and (3) the same representation can be used across different tactile sensing technologies mediated by acceleration, force, or strain. We further show that capabilities previously associated with research-grade instrumentation can be accessed using readily available, off-the-shelf hardware. We believe that broader access to high-bandwidth tactile sensing could facilitate the integration of contact dynamics into embodied learning systems and, for some tasks, offer an alternative or complement to increasing the spatial density of tactile sensing.

Answering epidemiological questions from real-world clinical data requires medical coding, schema-aware SQL, and validation of implicit choices about populations, denominators, and time. We present Ascent, an agentic system that exposes medical coding, question answering, and cohort analysis through a shared Model Context Protocol tool surface for standardized and native schemas. We introduce EpiTrap, a dataset testing whether systems avoid recognized pharmacoepidemiological errors, and compare a fixed pipeline with agents across models and orchestrators. With capable models, agents improve accuracy over the fixed pipeline by an average of 27 and 20 percentage points on native and standardized schemas, respectively. These gains require more tool calls and longer runtimes. Experience from real projects highlights the system's value for feasibility assessment, diagnostic iteration, and expert-guided analysis.

Prior case retrieval (PCR) aims to identify precedent cases relevant to a given query case. Existing PCR benchmarks and methods predominantly operate at the document level, treating entire judgments as the unit of relevance. This formulation is suboptimal for legal practitioners, as judgments address multiple legal issues and only a small subset of paragraphs is relevant to a particular query. Addressing this gap, we introduce UK-PRBench, a benchmark for paragraph-level precedent retrieval in UK case law, constructed from judgments obtained from the UK National Archives and covering a broad range of UK courts and tribunals. Furthermore, we evaluate state-of-the-art retrieval models and establish baseline results. Our experiments show that paragraph-level precedent retrieval remains challenging for current retrieval approaches, highlighting substantial room for improvement. UK-PRBench provides a standardised benchmark for evaluating fine-grained precedent retrieval and advancing retrieval systems for the UK legal domain.

Electricity forecasting often involves spatially related signals observed over regions, substations, and feeders, and Graph Neural Networks (GNNs) provide a natural way to represent these relations. Building a complete GNN forecasting experiment is nonetheless laborious, because graph construction, model selection, training, aggregation, and interpretation sit in incompatible tools. We present GraphToolbox, an open-source Python framework that unifies these stages in one configurationdriven pipeline built on PyTorch Geometric. It offers data-driven graph construction, an adapter that instantiates and trains 51 of the 65 PyTorch Geometric convolutions together with the recurrent cells of PyTorch Geometric Temporal, online expert aggregation, forecasting interpretability, and significance testing on cached forecasts. We evaluate the pipeline in two case studies. On French regional load, the 48 convolutions included in the complete forecasting sweep fall in a band from 1.14% to 1.60% error, online aggregation lowers this to 0.98%, and the graph models improve on classical additive and boosting baselines. On net-load, direct graph models are less accurate than a classical additive model, while forecasting each physical component separately improves them without closing that gap. Both comparisons use the same experimental interface, illustrating the role of GraphToolbox in systematic architectural evaluation.

Sea ice forecasts are issued several days ahead, allowing errors to accumulate while new, often sparse sea ice concentration (SIC) observations become available. We find that fixed-propagation errors concentrate near structured, high-gradient ice edges, whereas homogeneous interiors require limited propagation, suggesting that propagation distance should be state dependent. We therefore introduce ECHO (Evidence-guided Correction with Heterogeneous prOpagation), where ECHO-Scale adapts propagation distance while preserving correction geometry, and ECHO-Delta learns a bounded residual around fixed propagation. Across all 96 standard evaluation settings spanning diverse priors, observation times, sparsity levels, geometries, and noise conditions, both outperform fixed propagation. ECHO-Delta achieves the best average accuracy, while ECHO-Scale is more robust to geometry shifts. Code is available at https://github.com/yingtian22/TAKING-A-SECOND-LOOK.

Traditional evaluation metrics provides numerical values but often lack comprehensibility, hindering effective differentiation of model performances. Our work addresses this challenge by introducing overlay\_dx, a novel evaluation metric measuring the performance of time series prediction models. Overlay\_dx is a visual metric that represents the percentage of predictions falling within a confidence interval around actual values. Additionally, once evaluation results are plotted, overlay\_dx computes the area under the overlay curve, providing a quantitative measure of alignment between predicted and actual values across different thresholds and predictions. Through extensive experiments, we demonstrate that our approach offers a unified evaluation framework that combines both visual and numerical assessments, enabling improved model comparison and providing valuable insights for further research and optimization efforts in time series prediction.

Event logs arise in a wide range of real-world processes, capturing not only event activities and timestamps but also multi-modal contextual information. Existing event-sequence models, including many temporal point process approaches, primarily model event activities and timestamps while overlooking heterogeneous context, such as numerical measurements, categorical attributes, textual descriptions, and metadata associated with individual events and entire traces. In this paper, we propose Universal Multi-Modal Traceformer (UMT), a unified framework for incorporating heterogeneous process context into next-event prediction. Built on a Transformer backbone, UMT introduces a universal feature encoder that maps diverse feature types into a shared representation space and handles contextual information at both the event and trace levels. UMT further develops a per-event Perceiver module that dynamically weights contextual features and adaptively integrates them into event-token representations. To accommodate the heavy-tailed and potentially multi-modal distribution of inter-arrival times, UMT represents each interval at multiple temporal scales and jointly predicts the corresponding scale-specific quantities. Experiments on 13 real-world event logs show that UMT improves both next-event activity and time prediction over existing approaches.

Computational social science increasingly relies on large language models for text annotation, and the validity of published findings now rests on the labels generated by such models. Decision models, a new model class built for categorical question answering, answer typed questions with a choice, a probability distribution over the label set, and a confidence score rather than free text, at a small fraction of frontier inference prices. Whether their answers are accurate, and whether that stated confidence can be trusted on social science constructs, are unknown. Here, we mirror the evaluation of Ziems et al. (2024) on 18 computational social science classification tasks (7,977 items), comparing the first commercial decision model and two open-weight counterparts against 19 frontier and open-weight language models under the same zero-shot protocol. The decision model trails the per-task best LLM on 14 of 15 evaluation tasks, with a median deficit of 11.6 macro-F1 points, at a median 44 times lower measured cost. Its confidence is better calibrated than the verbalized confidence of 16 of the 19 LLMs, yet three frontier models show lower median calibration error (0.157 against 0.066). While items above 0.9 confidence are typically labeled accurately (median accuracy 0.815), on one task, empathy in peer-support dialogues, the model reports high confidence while performing near chance. Nonetheless, our results suggest that decision models are useful as a first step in the annotation pipeline: routing low-confidence items to an LLM matches or exceeds the LLM alone at a quarter to half of its cost.

Over the past decade, a growing body of research has shown that $γ$-weak submodularity broadly arises in numerous subset selection tasks, including feature selection, neural network pruning, and video summarization. Despite its prevalence, maximizing a $γ$-weakly submodular function subject to a general matroid constraint remains challenging. To date, the only known approximation guarantee is the conservative $(1+1/γ)^{-2}$ factor established by \citet{chen2018weakly}. To improve upon this result, this paper proposes a novel algorithm called \MGPE, which repeatedly performs maximum-gain local exchanges through careful control of a non-homogeneous Poisson clock, and proves that this \MGPE\ can attain an approximation ratio arbitrarily close to $ρ_γ=1-\left(γ/(2-γ)\right)^{ \frac{γ^2}{2(1-γ)} }$. In sharp contrast to the previous guarantee, our obtained factor $ρ_γ$ not only strictly improves upon $(1+1/γ)^{-2}$ for every $γ\in(0,1]$, but also can asymptotically approach the optimal $(1-1/e)$-approximation for submodular maximization as $γ\to1$. Furthermore, we surprisingly find that when the matroid constraint reduces to a cardinality or the objective satisfies the stronger notion of $α$-weak DR-submodularity, \MGPE\ can automatically recover the tight approximation ratios of $1-e^{-γ}$ and $1-e^{-α}$, respectively. Here, $α\in(0,1]$ denotes the DR ratio.

We present $t_0$, a family of open-weights foundation models for forecasting with multivariate context. We release its first two members: $\texttt{t0-alpha}$ and $\texttt{t0-beta}$, respectively 102M and 256M parameters. Both condition their forecasts on target history, past covariates, and known-future covariates, without task-specific retraining. Their transformer layers alternate attention along time and across variates. They produce probabilistic forecasts through quantile predictions. Pretraining combines curated public data with synthetic generator families constructed to contain covariate-to-target dependencies. On GIFT-Eval, $\texttt{t0-alpha}$ reaches an aggregate CRPS of 0.4941, and $\texttt{t0-beta}$ a CRPS of 0.4738 and a MASE of 0.6865, third on both and within 4.0% of the best zero-shot TSFM. On fev-bench they score 42.2 and 46.7 in skill, the latter third again and 2.0 points behind the leader. We analyze $\texttt{t0-alpha}$ in depth. Known-future covariates raise its skill by 6.3 percentage points across 30 tasks. The report also examines its calibration, its rollout strategy on long horizons, and its robustness to missing data. On the Victoria electricity-demand benchmark, $\texttt{t0-beta}$ is among the most accurate models with a context of nearly a year. In an independent Macrocosm evaluation of hourly ERCOT prices over 29 months, both cut the MAE of the lagged-price baseline by 38%.

Principal Component Analysis (PCA) minimises aggregate reconstruction error, which can inadvertently represent majority subgroups with substantially higher fidelity than minority subgroups. Fairness-aware extensions of PCA correct this disparity but require group labels as input. We address the logically prior question: given only a data matrix, which binary partition of the data suffers the greatest representational disparity under a shared PCA projection? We formalise this as the max-disparity partition problem and propose a greedy local-search algorithm, grounded in the Fiduccia-Mattheyses bipartitioning framework, that discovers the disparity-maximising partition without any predefined group labels. Two benchmark algorithms, a fixed-projection sorting baseline and a simulated-annealing variant, confirm that the greedy solution is empirically near-optimal. Having identified the partition, we attribute the disparity to specific features via PCA loading scores and association rule mining, enabling a practitioner to assess whether the disadvantaged group corresponds to a human-meaningful minority. On the Predict Students' Dropout and Academic Success dataset, representational disparity is driven predominantly by institutional and programmatic proxies for socioeconomic disadvantage, with gender emerging as a secondary but consistent contributor within the disadvantaged group. The discovered partition is then passed directly to Fair PCA, completing a detect-explain-mitigate pipeline.

We introduce the Endless Exam, a benchmark for measuring mathematical progress from today's models toward artificial superintelligence through fourteen parameterised construction families. Each submitted object is checked automatically for validity and assigned a relative quality score against a published frontier or construction baseline, without capping improvements at $1$. The families draw on open mathematical problems for long-term targets and generate new instances at larger parameters, where compact certificates keep large constructions verifiable. Across eight models evaluated on 69 distinct instances, continuous quality scores distinguish performance even though no evaluated system surpasses a published frontier. Size-quality curves show how construction quality changes as problem size increases. We release the generators, verifiers, references, model responses and analysis to support continued measurement before and beyond human frontiers.

Concepts are commonly defined as abstract, compact representations of knowledge and treated as basic units of intelligent behavior. Yet, cognition, psychology, and AI lack a shared mathematical language for them. Modern systems represent concepts as vectors, distributions, symbols, graphs, and other structures, but these formalisms are typically treated as competing rather than as solutions to a common problem. We propose an operation-based view that evaluates mathematical frameworks by the conceptual operations they support, identifying thirteen operations (including similarity, composition, generalization, and grounding) that recur across cognition, psychology, and AI. We show that ten frameworks embody distinct commitments to concepts as self-contained content, relational structure, or evolving process, and that these commitments determine which operations each supports naturally. For example, vector-based models facilitate graded similarity and generalization but struggle with explicit composition, whereas symbolic models support composition but offer but generalize poorly. No single framework we examined naturally supports all operations without extension. We test this account empirically using categorization as a case study, operationalizing nine theories on the same items against human judgments. Despite addressing the same conceptual question, the theories produce different procedures and results, demonstrating that mathematical commitment shapes what a theory can explain. We call for hybrid formalisms that treat content, relation, and process as jointly primary.

Functional annotation of newly sequenced proteins remains a bottleneck in molecular biology: the number of sequences in public repositories grows far faster than the capacity for manual curation. Most computational approaches consider annotation as multi-label classification over a fixed ontology, which constrains predictions to a predefined label set. In this work we study the the protein annotation as a sequence-to-text generation problem. We fine-tune the 3B-parameter Ministral 3 base model with QLoRA (4-bit NF4 quantization with low-rank adapters) on sequence annotation pairs. We assess predictions with an LLM-as-expert protocol: a GPT model prompted as a senior molecular-biology curator scores organism identification as binary and function annotation quality. We conclude that QLoRA-fine-tuned compact LLMs can generate curator-style annotations with genuine biological value for a substantial subset of proteins. We also discuss future directions in data quality, model scaling, and evidence grounding that are needed to make the approach sufficiently reliable for practical use.

Random forests (RFs) predict well but are opaque, whereas single decision trees are interpretable but unstable. Artificial representative trees (ARTs) were developed as interpretable surrogate models for RFs, but their use as standalone prediction models with uncertainty quantification has not been systematically investigated. We combine ARTs with leaf-wise Mondrian conformal predictive systems (CPS), enabling a single tree to provide continuous predictions, prediction intervals, and probabilities of exceeding arbitrary thresholds. We compared ARTs with CPS against decision trees with CPS and separate regression and probability trees across five simulation scenarios, 21 benchmark datasets, and a cross-sectional NHANES example data set. Repeated cross-validation assessed predictive performance, interpretability, and stability. ARTs with CPS yield compact, structurally stable trees with substantially more reproducible split-variable selection than decision trees across benchmark datasets and NHANES. Decision trees showed slightly better predictive performance and narrower prediction intervals, while coverage was broadly comparable. CPS-based trees generally achieved lower and less variable Brier scores than multi-model approaches. Combining ARTs with CPS therefore provides a single, interpretable, and stable model for continuous predictions and calibrated probabilities, balancing predictive performance with reproducibility and transparency in settings where stability and interpretability are essential.

Model merging aims to combine multiple fine-tuned models derived from a common pretrained model into a single multi-task model without additional joint training. Recent spectral merging methods improve over simple weight averaging by exploiting low-rank structures of task-specific updates, but they commonly assign the same rank capacity to every task. This uniform allocation ignores that task vectors can have heterogeneous spectral complexity, causing the shared merging space to be used suboptimally. In this paper, we propose Spectral Energy-proportional Rank Allocation (SERA), a simple task-adaptive strategy that allocates ranks according to the singular-value energy structure of each task vector. By assigning richer spectral capacity to complex or isolated tasks and fewer directions to compact tasks, SERA extends SVD-based model merging from uniform-capacity merging to task-dependent capacity allocation. Experiments under standard vision model merging protocols show that SERA improves multi-task merging performance while preserving the same total rank budget as existing spectral merging methods. Further analysis demonstrates that task-level spectral concentration is closely related to the per-task effect of adaptive rank allocation, providing insight into when and why SERA is effective.

In recent years, large language models (LLMs) have emerged as a popular alternative for evaluation. Often referred to as LLMs as judges (LLJs), these systems have been widely adopted by researchers and practitioners across a broad range of measurement tasks, driven by their strong performance, scalability, and cost-effectiveness relative to human judgment. However, a growing body of work has shown that the use of LLJs raise concerns about their validity and reliability as evaluators. Existing efforts to address these challenges have largely focused on developing bias-mitigation techniques and refining prompting strategies. While these approaches represent an important step forward, they primarily offer technical fixes and leave a more fundamental challenge unaddressed: the lack of standardized, transparent, and reproducible evaluation practices. In this paper, we introduce LLJ Cards, a framework that synthesizes best practices from measurement theory, natural language generation, and machine learning literature into practical guidelines for LLJ-based evaluations. While LLJs offer a promising path toward scalable evaluation, their effective use requires grounding in rigorous evaluation principles to ensure validity, reliability, and reproducibility. LLJ Cards addresses this need by providing a structured framework for applying these principles in the design and reporting of automated evaluations.

AI agents are being deployed rapidly, accompanied by a growing number of AI-specific attacks and corresponding incidents. As incident reporting becomes increasingly important for legal compliance, governance, accountability, and security; current frameworks must be adapted to the unique characteristics of AI agents. In this paper, two editorial authors compare AI systems and AI agents and, drawing on input from 23 experts in academia and industry, identify the information required for reporting incidents where the security of AI agents is harmed. %involving AI agents. Potential reporting elements include, for example, agent memory and memory accesses, actual and potential levels of autonomy, and tool usage. Based on these findings, we identify several open research questions, including how to efficiently record incidents and how to determine whether vulnerabilities and incidents generalize. Expert feedback also highlighted potential reporting weaknesses, such as risks of data leakage and attacks targeting the reporting infrastructure itself, creating additional research needs. Lastly, we summarize privacy requirements and outline research directions for the secure and trustworthy deployment of AI agents.

Fine-tuned checkpoints and adapters now fill public repositories, and the most common operation applied to these artifacts is model merging: arithmetic on their weights that assembles capabilities cheaply. We ask what this operation does to emergent capabilities: behaviors an artifact carries that were never an explicit training target. Studying two independent testbeds (activation oracles and emergent-misaligned models) across three model families, we find that the answer is threefold. First, merging preserves an emergent capability that both parents carry: merging two misaligned checkpoints retains most of their broad misalignment across the whole mixing range. Second, merging cannot create an emergent capability that is superadditive in its parents: no weighted merge of two single-task oracles reaches the jointly-trained oracle's auditing ability. Third, when only one parent carries the capability, merging dilutes it faster than the trained capability that accompanies it: the gap is significant in most settings. In short, emergent behaviors of an artifact do not compose the way its trained capability does.

Offline reinforcement learning (RL) typically trains a critic by minimizing a regression loss against bootstrapped value targets stabilized by target networks with exponential moving average (EMA) updates. Multi-step targets incorporate behavior-policy actions and therefore require off-policy correction. We instead impose in-sample Bellman optimality on the critic through inequality constraints. We formulate the Lifted Bellman Linear Program (LBLP), which lifts the linear programming characterization of Bellman optimality to the joint $(Q,V)$ space so that every constraint involves only state-action pairs in the dataset. Its unique minimizer is the in-sample optimal pair, and constraints along $K$-step segments of dataset trajectories leave this minimizer unchanged for any rollout policy and horizon. Under deterministic dynamics, this minimizer lies between the best dataset return and the optimal value. Relaxing the constraints into hinge penalties recovers the same solution above a finite penalty coefficient in the tabular case. Approximate Lifted Bellman Unconstrained Minimization (ALBUM) implements this relaxation with neural networks and detaches the $K$-step rollout targets by stop gradient. Its objective contains no squared regression onto bootstrapped targets, so it can be trained without target networks or EMA updates. Under deterministic dynamics, the LBLP solution is a stationary point of the detached update under a coefficient condition independent of $γ$ and $K$, and the inequality constraints allow discounted returns along dataset trajectories to serve as lower bounds without off-policy correction or action chunking. On OGBench, ALBUM uses a single critic with a Gaussian policy, matches the average performance of FQL, and is comparable to recent action-chunking methods, while using the fewest parameters and the least peak GPU memory among all compared methods.

In this work, we present a method for shape reconstruction and tracking from video via agentic analysis-by-synthesis. Unlike prior methods which first estimate dense pixel correspondences and then recover object motion from them, our method infers a structured 3D object model, including its geometry and kinematic structure, and uses this model to optimise object track estimates over time. In our optimisation loop, a Vision-Language Model (VLM) agent iteratively refines shape or generalised pose through a render-and-compare loop, combining coarse visual reasoning with numerical pose optimisation for precise state estimation. This structured formulation enables our method to track through large motion, articulation, and severe occlusion without relying on pixel-matching objectives. Quantitatively, on ARCTIC, our method substantially outperforms state-of-the-art 3D point-tracking baselines for articulated objects, and on HOT3D it outperforms all evaluated rigid-object tracking baselines.

Visual token pruning is a promising approach to reducing the inference cost of large vision-language models (LVLMs), yet aggressive token reduction often causes substantial performance degradation. We identify three key factors behind this degradation: text-guided selection bias, information loss from discarded tokens, and positional distortion caused by sequence compaction. Based on these observations, we propose \textbf{VPRune}, a training-free pre-LLM pruning framework consisting of visual-only diversity selection, similarity-guided token recycling, and position-preserving restoration. Experiments on FastVLM-1.5B across multiple vision-language benchmarks demonstrate that VPRune achieves a favorable accuracy--compression trade-off, with particularly pronounced advantages under aggressive compression. Furthermore, evaluations on edge-device show that VPRune effectively reduces end-to-end inference latency while maintaining superior task performance, demonstrating its practicality for resource-constrained LVLM deployment.

Deploying Large Language Models (LLMs) in healthcare requires robust performance across two complementary dimensions - diagnostic reasoning: the convergent, evidence-driven task of inferring a patient's condition from clinical data to produce a diagnosis, and clinical healthcare reasoning: the broader, navigational judgment required to communicate, plan, and adapt across multi-turn clinical interactions where a single correct answer may not exist. Recent benchmarks such as HealthBench and MedXpertQA reveal persistent weaknesses in both areas, exposing failures in complex diagnostic scenarios and limitations in contextual, patient-centered dialogue. We introduce a sequential training framework that targets these facets using synthetic data and rubric-based reinforcement learning. First, we improve diagnostic reasoning using MedBullets-derived questions with rule- and rubric-guided Reinforcement Learning (RL). We then shift to clinical reasoning by generating 5.3k synthetic multi-turn scenarios, each paired with multi-dimensional rubrics to comprehensively assess the response. This approach yields over 10% improvement on MedXpertQA, and our 30B model achieves 50.1% accuracy on HealthBench-Hard, surpassing proprietary baselines including GPT-5 (thinking). Our results show that targeted synthetic datasets and rubric-based training can systematically improve both diagnostic and interactive clinical reasoning in medical LLMs.

Music festival lineups emerge from complex relationships among artists, genres, releases, labels, and past performances, making the prediction of future lineups a natural fit for temporal knowledge graph (TKG) forecasting. In this work, we present a TKG covering 380 festivals over 55 years, comprising more than 90K festival performance quadruples along with information on festivals, artist tours, and artist metadata, and release it as a resource for TKG forecasting evaluation. We formalize festival lineup forecasting as temporal link prediction between artists and festivals at future timestamps. We evaluate six TKG forecasting models on this task, analyze their capabilities and limitations, and compare them against Large Language Models applied zero-shot. Our resource complements existing TKG benchmarks by grounding evaluation in a concrete, real-world application domain.

Using a Large Language Model (LLM) as the clusterer at production scale is hard: prompts cannot hold the entire label space, and per-document serial processing does not deliver the throughput real workloads require. We present RAILS, a retrieval-augmented incremental LLM clusterer that turns clustering into a simple loop over a growing label pool and scales through document batching with bounded concurrency. On six public benchmarks RAILS exceeds the strongest prior LLM-clustering method on average, lifting accuracy from 51.2% to 59.3%, NMI from 67.2% to 74.8%, and ARI from 45.4% to 54.7%. We further report production-deployment evidence from a SaaS ticket-topic-discovery pipeline, where RAILS has replaced a traditional HDBSCAN stage with higher clustering quality, transparent prompt-driven control, and stateful incremental operation.

Distributed reinforcement learning (RL) scales training by parallelizing actors and learners around an Experience Buffer. As RL workloads grow, however, the buffer becomes more than a replay queue: it is the storage substrate of a large-capacity, latency-critical experience path that every iteration traverses to move, transform, sample, and batch experiences before learner updates can begin. Existing RL systems embed this path inside framework control flow or expose it as a request-driven buffer service, leaving experience placement fixed and experience-path work difficult to schedule independently as a runtime-level optimization target. We present Conduit, a framework-agnostic runtime that exposes RL experience management as an explicit systems optimization problem. At its core is the Experience Data Plane (EDP), a runtime abstraction that separates RL experience-handling semantics from framework-specific execution logic by exposing experience ingestion, experience placement, and experience delivery as explicit control points. Built on EDP, Conduit introduces capacity-constrained, bandwidth-aware placement, which distributes experience state across CPU/GPU memory tiers and nodes under heterogeneous interconnect and device-memory constraints, and latency-aware scheduling, which controls when experience-path handling runs to reduce exposed experience-path latency while preserving RL semantics. Integrated with RLlib without changing its framework execution logic, Conduit reduces exposed experience-path latency by up to 97% and end-to-end iteration latency by up to 38%, scales to 1,024 GPUs, and preserves convergence.

Intelligent transportation systems require Incremental Learning (IL) to continually improve their overall performance in dynamic environments. However, most edge devices lack the computational resources to support on-device IL, requiring updates to be transmitted from centralized servers. We propose using this setup to obtain dense, specialized module coverage that adapts a fixed base model to specific spatiotemporal contexts, such as parking lots, gas stations, ferries, or construction sites. However, in order to reliably transmit these modules to the edge device, using TCP, UDP, and BTP over V2X, Wi-Fi, and 2G-5G hardware, we establish a strict limit of 14.6 KB per module to fit within the first TCP window and to minimize UDP/BTP fragmentation. We further introduce Mixture-of-Experts for Communication-Aware Incremental Learning (MECAIL), the first method that meets this strict requirement, in which each new domain or environment is served by a small expert network that adapts the base model. We validate MECAIL on D-RICO and ODinW-13, where it largely matches the performance of parameter-heavy approaches while enabling practical, bandwidth-efficient large-scale deployment. This allows comprehensive coverage by experts for highly specific, focused, and temporary situations.

Predicting postprandial glycemic response (PPGR) is fundamental to personalized nutrition and type 2 diabetes management, yet existing approaches typically rely on manually reported dietary intake, limiting their scalability in free-living settings. We propose a multimodal framework that replaces manual dietary logging with image-derived macronutrient estimates and integrates them with clinical variables and gut microbiome information for personalized PPGR prediction. The framework jointly performs image-based macronutrient estimation and glucose prediction, while an attention-based prediction module models interactions between dietary and host-specific information. We evaluate the proposed approach on a real-world dataset comprising meal images, continuous glucose monitoring, clinical variables, and gut microbiome profiles. The proposed model outperforms existing PPGR baselines using image-derived nutritional inputs and approaches the performance of methods that rely on manually reported macronutrients despite using automatically estimated nutritional information. These results demonstrate that combining image-derived nutrition with complementary clinical and gut microbiome information provides a practical foundation for scalable personalized PPGR prediction.

Recent 4D LiDAR language models aim to reason about objects and their evolving spatial relationships. Yet, in our evaluation, always selecting the same option nearly matches the multiple-choice accuracy of two B4DL-derived configurations. We introduce LiDAR-Hallu, a geometry-referenced benchmark and diagnostic protocol with 10,000 questions across 150 nuScenes scenes. It covers object existence, ego-relative position, distance ordering, relative motion, and temporal localization, with explicit rules for selecting objects, comparing times, and determining reference answers. Our protocol combines fixed-answer and candidate-content controls, cross-scene pairs with identical prompts but opposite reference answers, and relation-specific recall. Analysis of 100,000 recorded responses reveals failures hidden by aggregate accuracy. Candidate duration alone makes temporal answers predictable without observing LiDAR. On paired questions, the models frequently give the same answer to scenes requiring opposite answers. Relation-specific analysis further shows that both configurations miss every positive lateral-motion case across all tested conditions. Temporal-shuffle contrastive decoding provides little net improvement, as repairs are largely offset by new errors and the main failures persist. These results show that evaluating spatio-temporal reasoning requires testing whether models distinguish the queried physical relationships, rather than relying on individual-answer accuracy alone. The source code, checkpoints, and data are released at https://github.com/Awesome4D/4DMLLM_Hallucination_Bench.

Large language model (LLM) agents increasingly execute long-horizon workflows through external tools, allowing untrusted outputs to influence subsequent actions and exceed user authorization. Existing defenses isolate injected content or constrain execution with predefined plans and static policies, but these approaches are brittle under dynamic workflows and scale poorly across extensible tool ecosystems. In this work, we present ActGov, a runtime enforcement framework that validates each LLM-proposed tool action before it causes external effects. Built on a unified semantic model of authorization, actions, runtime context, and security constraints, the ActGov-Policy component iteratively constructs a policy set from tool specifications, benign tasks, and observed failure traces, with each update verified through SMT-based counterexample checking. At runtime, ActGov-Runtime abstracts each tool call into finite policy records and permits it only if it remains within the task-scoped authorization boundary and satisfies all applicable policies. This per-action enforcement preserves authorization throughout long-horizon, dynamically branching workflows. We evaluate ActGov on the AgentDojo and AgentDyn benchmarks across multiple models and attack configurations. It shows that ActGov consistently reduces the success rate of indirect prompt-injection attacks while preserving task utility, significantly outperforming existing defenses. These results demonstrate that ActGov can enforce fine-grained authorization over dynamic agent executions without relying on the underlying LLM to correctly identify malicious instructions.

Accurate, reliable, and deployable wind power forecasting is critical for power system dispatch, renewable energy integration, and electricity market operations. Progress in this field hinges on the ability to empirically and comprehensively benchmark forecasting methods. Yet existing benchmarks fall short of supporting systematic evaluation in four key aspects: 1) limited coverage of wind power scenarios across turbine scale, variable composition, and spatial structure; 2) incomplete coverage of forecasting model families; 3) evaluation metrics misaligned with wind power requirements; and 4) limited structure-aware diagnostics beyond individual temporal patterns. To address these limitations, we propose WPBench, a comprehensive, fair, and extensible benchmark for wind power forecasting. WPBench integrates 26 public datasets organized by turbine scale and variable composition, spanning single-turbine, multi-turbine, univariate, and multivariate settings. Under unified processing, training, and evaluation protocols, it benchmarks 19 representative models covering traditional methods, deep temporal models, spatio-temporal models, and foundation models. Beyond point-wise errors, WPBench assesses forecast-curve fidelity and computational efficiency, and delivers structure-aware diagnostics across temporal, variable-dependency, and spatial-dependency perspectives. Together, these capabilities enable systematic model comparison across diverse wind scenarios and provide a reusable platform for future research.

Reliable disruption prediction is essential for the safe operation of future tokamaks. Existing full-distribution survival methods model the complete residual time-to-disruption distribution, whereas operational decisions primarily depend on disruption risk within a finite prediction horizon. This mismatch motivates introducing Early Event Prediction (EEP) objectives into survival-based disruption prediction. We take Deep Survival Machines (DSM) as the full-distribution baseline and propose applying two established EEP methods to tokamak disruption prediction: Temporal Label Smoothing (TLS), which directly predicts disruption probability within a finite horizon, and survTLS, which additionally models the event-time distribution within that horizon. Using a common causal encoder, we compare these methods on DIII-D, Alcator C-Mod, and EAST. We distinguish threshold-free deadline ranking from validation-selected fixed-policy alarm performance and evaluate prediction horizons and encoder architectures. TLS achieves the best mean alarm performance on DIII-D and EAST, whereas all methods perform poorly on Alcator C-Mod. survTLS does not consistently outperform DSM, suggesting that directly learning horizon-level event probability is more effective than modeling detailed within-horizon event-time distributions in the present setting. Finally, the selected prediction horizons and encoder-ablation results vary across devices, reflecting differences in disruption characteristics.

Multivariate time-series forecasting is essential to many real-world applications. Recent large vision models (LVMs) offer a promising paradigm by transferring cross-domain visual priors to time-series forecasting. However, existing LVM-based methods face two key challenges: balancing independent visual representation spaces with cross-variable dependency modeling, and adapting vision backbones pretrained on natural images to the distinct temporal semantics of time-series images. To address these challenges, we propose MUSE, a dependency-aware adaptation framework built on a fully frozen pretrained MAE. First, the Variable Context Refinement Module (VCR) aggregates shared temporal information within each variable and models cross-variable contextual dependencies while preserving independent visual spaces. Second, the Temporal-Periodic Refinement Module (TPR) performs lightweight refinement at different encoder depths and explicitly models across-period temporal dependencies and within-period periodic dependencies. The two modules independently produce forecasts, which are fused through a learnable prediction-level gate. Experiments on 10 real-world datasets demonstrate that MUSE achieves state-of-the-art performance.

The quadratic scaling of Transformer self-attention has driven the adoption of sub-quadratic Selective State Space Models (SSMs) like Mamba, which compress past context into a fixed-size recurrent hidden state. This strict informational bottleneck raises a foundational question for mechanistic interpretability: do SSMs and Transformers learn fundamentally distinct latent representations? In this work, we employ Sparse Autoencoders (SAEs) to conduct a large-scale, feature-level correspondence analysis between Mamba-130m and Pythia-70m over a 10-million token corpus. Contrary to hypotheses predicting widespread architectural divergence, we find no evidence of systematic representational divergence between architectures: across the observed Jaccard distribution, 99.98% of Mamba features cluster toward the upper alignment boundary, providing preliminary feature-level support for the Universality Hypothesis. We further identify and qualitatively characterize this microscopic fraction (0.02%) of diverging features, finding patterns consistent with the hypothesis that the recurrent bottleneck selectively limits the parsing of rigid syntax rather than broad semantic ontology. We demonstrate that while Pythia's unconstrained attention permits the monosemantic decomposition of distinct formatting edge-cases, Mamba is forced to compress unrelated syntactical anomalies into polysemantic "junk drawer" neurons to preserve state capacity. Collectively, these results suggest that architectural routing mechanisms may have negligible impact on core semantic understanding, with representational divergence confined to extreme structural margins.

Low-bit vision-language-action inference must reduce observation-to-action latency while preserving robot behavior. We present FoldQuantVLA, a post-training quantization framework that carries a consistent activation representation through calibration, weight rounding, and native integer execution. It combines channel scaling and block Hadamard transforms with dynamic per-token quantization, without policy retraining. Custom TensorRT plugins execute projections in both the language backbone and iterative action expert with four-bit weights and activations (W4A4) on Ada GPUs and Jetson AGX Orin. Evaluation spans LIBERO, SimplerEnv, and two robot platforms. Across three GR00T checkpoints and $π_{0.5}$, W4A4 achieves $1.20$ to $1.33\times$ speedups over floating-point TensorRT on Orin and $1.25$ to $1.52\times$ on desktop. Retaining language attention-output and feed-forward down projections at eight bits (W8A8) improves held-out action fidelity on all four checkpoints. Across four real-robot tasks, this configuration raises observed GR00T N1.7 success from $80.0\%$ with uniform W4A4 to $92.5\%$ over 80 trials per configuration, with a measured additional Orin latency of 1 ms.

Sparse on-policy distillation (OPD) allocates teacher supervision to a small subset of tokens in student-generated trajectories. However, useful teacher guidance can yield a noisy update when its gradient is estimated from a sampled next token. We study this estimation problem at a fixed prefix in information geometry and propose an information-efficiency ratio (IER) based on a signal-to-noise decomposition. IER characterizes relative gradient estimation error under an optimal scalar baseline. A candidate-set approximation enables token selection based on IER and its combination with existing usefulness scores, while retaining the sampled reverse-KL training objective. On mathematical and medical reasoning tasks, adding IER improves existing selectors in multiple settings, with sparse configurations matching or exceeding full OPD without token selection at small token budgets of 0.1\%--1\%. These results support accounting for both usefulness and gradient-estimation reliability when allocating sparse supervision. Our code is available at https://github.com/BruceSheng1202/IER-OPD.

Monocular RGB-based hand pose estimation has emerged as a critical research frontier in computer vision. The local hand pose estimation methods predict hand poses relative to the wrist, while global hand pose estimation also requires estimating the wrist's position in the camera coordinate system. However, this camera-space estimation confronts two fundamental challenges: (1) depth ambiguity in monocular settings, and (2) the coupling effect of hand local poses and global wrist positions in the perspective projections. In particular, this coupling reflects that the projections are jointly determined by local hand poses, wrist positions, and camera intrinsics. To overcome these challenges, our framework proposes two key innovations: Transformation-Isomorphism Supervision for hand-depth information extraction and Perspective Information Embedding for resolving above coupling effect of local pose and wrist position, both integrated within the mainstream encoder-decoder architecture. Besides, we propose a novel framerate-aware multi-dataset training strategy for sequential pose refinement. Our fully integrated approach achieves at most 37.1\% superiority in CS-MJE over SOTA on HO3D. Project page: https://github.com/Mine268/CS-ViT.

We consider a standard convex composite optimization problem with either smooth or nonsmooth objective function, and under quadratic growth. In recent years, several works gave algorithms based on a \textit{weak proximal oracle} (WPO) that essentially match in oracle complexities proximal (sub)gradient methods relying on exact prox operations. Importantly, such WPOs, which relax the strong optimality condition of the standard prox operator, may admit much more efficient implementation in terms of runtime when optimal solutions have some sparse structure. A question remained if such WPO-based methods can be accelerated (in the sense of Nesterov's accelerated gradient). In this work we provide a negative answer by establishing lower bounds against both deterministic and randomized methods. Thus, while WPOs can substantially reduce the cost of individual oracle calls, this comes with an inherent loss in oracle complexity. We also provide a new upper-bound for WPO-based nonsmooth convex composite optimization, nearly matching the proximal subgradient method.

Hierarchical data is ubiquitous in the empirical sciences and is most commonly analyzed with generalized linear mixed-effects models (GLMMs). Bayesian inference for GLMMs yields calibrated uncertainty but requires MCMC; the No-U-Turn Sampler (NUTS) is the gold standard but is slow and must restart from scratch for every new dataset, model and prior. We introduce metabeta, a pretrained neural network for prior-amortized in-context Bayesian inference over GLMMs. Unlike previous neural posterior estimators that fix the prior at training time, metabeta accepts prior families and hyperparameters as inputs at test time, enabling zero-shot generalization. Two set transformers and conditional normalizing flows mirror the posterior's two-level structure (global parameters shared across groups, local parameters per group). The model is trained on millions of realistic simulated datasets spanning continuous, binary, and count outcomes. By default, the flow posterior is refined by Independence Metropolis-Hastings against the unnormalized posterior, so its correctness rests on the sampler rather than the network; this yields tuning-free inference two to three orders of magnitude faster than NUTS. Alternatively, the flow can warm-start NUTS, giving nearly identical inference with substantially increased speed and stability. On controlled benchmarks with ground-truth parameters, metabeta matches NUTS in parameter recovery, calibration and out-of-sample prediction. On out-of-distribution real datasets, its posteriors closely match those of NUTS across all parameter types, and they remain faithful under misspecified likelihoods and priors, out-of-distribution predictors, collinear designs, and data-poor regimes. The model is open-source and open-weights and thus immediately deployable.

Despite advances in long-context inference, large language models (LLMs) remain fundamentally limited by the key-value (KV) caching mechanisms that are necessary for stable computation. Techniques such as selective token eviction and pruning have vastly mitigated these issues, but often discard core information to manage the growing cache. In this paper, we propose Attention with Routed Memory (ARM) a novel KV caching structure that introduces a fully differentiable, fixed-size memory system organized as a hierarchical router. Via a Gumbel-Softmax, ARM learns to select memory slots and perform sigmoid-gated updates that softly combine new and stored information, avoiding hard eviction and reducing information loss. By further training a policy to dynamically select varying amounts of memory at inference, ARM adapts its accesses for both simple contexts and inputs that require deeper reasoning, enabling more scalable and effective retrieval on both short- and long-contexts. Experimental results on standard commonsense and long-context reasoning benchmarks demonstrate that ARM achieves superior performance and efficiency compared to fixed KV-caching approaches, while remaining efficient and scalable in terms of both memory and generation latency.

Speech-to-text engines are extremely needed nowadays for different applications, representing an essential enabler in human-robot interaction. Still, some languages suffer from the lack of labeled speech data, especially in the Arabic dialects or any low-resource languages. The need for a self-supervised training process and self-training using noisy training is proven to be one of the up-and-coming feasible solutions. This article proposes an end-to-end, transformers-based model with a framework for low-resource languages. In addition, the framework incorporates customized audio-to-text processing algorithms to achieve a highly efficient Jordanian Arabic dialect speech-to-text system. The proposed framework enables ingesting data from many sources, making the ground truth from external sources possible by speeding up the manual annotation process. The framework allows the training process using noisy student training and self-supervised learning to utilize the unlabeled data in both pre- and post-training stages and incorporate multiple types of data augmentation. The proposed self-training approach outperforms the fine-tuned Wav2Vec model by 5% in terms of word error rate reduction. The outcome of this work provides the research community with a Jordanian-spoken data set along with an end-to-end approach to deal with low-resource languages. This is done by utilizing the power of the pretraining, post-training, and injecting noisy labeled and augmented data with minimal human intervention. It enables the development of new applications in the field of Arabic language speech-to-text area like the question-answering systems and intelligent control systems, and it will add human-like perception and hearing sensors to intelligent robots.

Structured pruning is a model compression technique that is used to reduce the computational cost of deploying deep neural networks on resource-constrained devices. Popular methods of pruning rely on opaque heuristics or weight-based criteria that give no indication as to the structural dependencies in the network. To address these limitations we present Artificial Structure Function Search (ASF-S): a novel structured pruning framework. ASF-S utilizes Principle Gradient Importance (PGI): a novel prune-candidate selection criteria that is inspired by structure-function relationships in the brain. By ensuring the pruned structure of the model respects topographical organization of the output layer, we define Artificial Functional Connectivity (AFC) for artificial neural networks. AFC provides evidence to demonstrate that accurate smaller networks can be found using careful prune candidate selection criteria. We present results for PGI as a selection criterion and for ASF-S as a pruning framework against recent benchmarks, demonstrating that our method yields model variants with 70\% parameter reduction, that can recover baseline accuracy without re-training the pruned layers.

The conventional testing process quickly fails when applied to ML-based systems such as obstacle detection in vehicles: if an obstacle is not detected in a test, classical bug fixing is impossible and an AI system will always retain shortcomings. Test results can therefore only be interpreted statistically, which in turn requires test sets that are not only complete with respect to the operational design domain (ODD) of the system, but also representative of it. To this end, we introduce probabilistically extended ontologies (PEONs): ontologies describing the ODD, augmented with a probability distribution over the partitioning they induce. Instead of unmaintainable conditional probability tables, only marginal distributions and functionally described dependencies need to be specified; algorithms based on couplings and optimal transport complete this specification to a Bayesian network. From a PEON we derive the sampling of representative test cases, rigorous end-of-test criteria for given quality targets and significance levels, and methods for re-evaluating existing test results and for assessing the balance of training data. We demonstrate the practical modelling of a complex ODD using the example of automatic train operation.

Climate variability influences whether a market disruption escalates into a food crisis, yet broad climate patterns like El Niño, tracked months before they alter hydro-climatic conditions, are still not incorporated as an early-warning component in food-security responses. We address this gap by introducing sensitivity regimes, a stratification of regions by the direction and strength of their vegetation response to the El Niño Southern Oscillation, and using them to estimate how food price spikes affect acute food insecurity across sub-Saharan Africa. Integrating remote sensing, socioeconomic data, and causal machine learning, we find that in regions where ENSO systematically suppresses vegetation, a price spike raises the share of the population at acute risk by 5.4 percentage points in the following month. In regions where vegetation is unaffected by or positively linked to ENSO, the estimated effect is smaller (around 2 percentage points) and statistically insignificant. These results demonstrate that climate context is critical for understanding food security vulnerabilities. Sensitivity regimes can be combined with operational price-spike triggers to stage anticipatory action: the ENSO state flags vulnerable regions months ahead, and a pre-positioned response in those regions to a price spike would avert the largest jump in acute food insecurity.

Voice-controlled interaction in industrial settings is hampered by acoustic noise, which severely degrades audio-only speech recognition. Audio-visual speech recognition (AVSR) addresses this by fusing lip-motion cues with the audio stream, but state-of-the-art pipelines rely on three-dimensional convolutions, recurrent units, and attention modules that exceed the budget of typical edge devices. We present NAVIR, an end-to-end AVSR system targeting the BrainChip Akida neuromorphic processor, which natively supports only sequential two-dimensional convolutional inference. The pipeline factorises spatial and temporal encoding into separate AkidaNet-based modules: a per-frame visual encoder, a temporal video encoder, and a spectrogram audio encoder, fused by a lightweight predictor head and decoded by constrained beam search. Models are trained with connectionist temporal classification on noise-augmented audio and then fine-tuned with quantization-aware training. On the GRID benchmark, the quantized audio-visual model reaches 14.0% word error rate (WER) under noise on the unseen-speaker split and 3.3% WER on the overlapped-speaker split, against 22.5% and 11.8% for audio-only baselines, and it attains 98.6% command accuracy at 1.5% WER on a task-specific industrial-command corpus. Operation-count analysis indicates a 13-fold energy advantage of the spiking formulation over its artificial neural network counterpart at 27.6% mean firing rate. On-board measurements show roughly 5-fold lower energy per inference than a Raspberry Pi central processing unit on the lip-reading model, and over 100-fold lower than a laptop graphics processing unit, while sustaining 14.5 inferences per second. To the best of our knowledge, this is the first complete multimodal AVSR pipeline running on neuromorphic hardware of this class.

Childhood stunting remains a major public health concern in Bangladesh and reflects long-term growth failure influenced by child, maternal, household, socioeconomic, and health-service factors. This study used nationally representative Bangladesh Demographic and Health Survey data from 2007 to 2022 to develop machine learning models for population-level prediction of childhood stunting and to assess temporal robustness and subgroup fairness. Children aged 0-59 months with complete anthropometric and predictor data were included. Data from the 2007, 2011, and 2014 survey rounds were used for model development, while the 2018 and 2022 rounds were retained as temporal test datasets. Twelve feature-selection approaches were assessed, and the KNN permutation importance-selected predictor set was used for final model evaluation. Eleven machine learning models were evaluated: ten conventional algorithms and one pretrained tabular foundation model, TabPFN. Performance was assessed using balanced accuracy, AUROC, F1-score, Brier score, and expected calibration error. Subgroup fairness was examined by child sex, place of residence, and socioeconomic status. The final analytic sample included 18,844 children, of whom 35.05% were stunted. In the development hold-out test dataset, TabPFN showed the highest observed balanced accuracy overall at 67.58%, while AdaBoost showed the highest observed balanced accuracy among conventional models at 67.51%. In temporal testing, the highest observed balanced accuracy was found for Gradient Boosting in BDHS 2018 and XGBoost in BDHS 2022. Model performance varied across survey rounds and subgroups, highlighting the importance of temporal validation, subgroup fairness assessment, and transparent interpretation in public health prediction modeling.

Tactile sensing is an essential modality for robots performing contact-rich, dexterous manipulation, particularly under visual occlusion. While pre-trained image encoders are standard in robot learning pipelines, tactile encoders are still commonly trained from scratch from raw, noisy signals, which might limit their expressivity. Existing self-supervised learning (SSL) approaches focus predominantly on vision-based tactile sensors, leaving distributed electronic skins largely unaddressed. These sensors, however, have a distinctive property: their sensing elements are sparse and irregularly arranged over the surface they cover, which makes direct reuse of visual SSL methods suboptimal. We present Tactile-JEPA, an efficient self-supervised pre-training method that uses the spatial arrangement of tactile sensors to learn topology-aware representations. Specifically, it is trained to predict the embeddings of masked sensing elements from the unmasked remainder, using the sensor connectivity graph to guide spatial masking. Our analysis shows that effective tactile representations require capturing both local contact details and the global state of the tactile surface, which we achieve through dual-scale masking. Across three diverse datasets spanning magnetic and piezoresistive sensors, different robot embodiments, and single- and paired-sensor configurations, Tactile-JEPA reduces force estimation error by 6.3% and in-hand orientation error by 20.8% over the prior state-of-the-art, with consistent gains in other downstream applications, including policy learning. Overall, our results demonstrate that the benefit of tactile sensing depends critically on the quality of encoder pre-training, a problem which Tactile-JEPA addresses directly. Code is available at https://github.com/E-Kovtun/tactile.

Recognizing hand-drawn geometric shapes is a foundational sub-problem of sketch recognition, with applications in education, human-computer interaction, and diagram digitization. This paper presents the design, implementation, and evaluation of a desktop application that recognizes four basic hand-drawn geometric shapes, circle, square, rectangle, and triangle using a compact Convolutional Neural Network (CNN). A dataset of 2,000 labeled 28x28-pixel shape images was collected independently and released publicly. The classifier consists of three convolutional blocks (16, 32, and 64 filters) with max-pooling, an in-model data-augmentation stage (random horizontal flip, rotation, and zoom), a dropout-regularized dense layer of 128 units, and a 4-way linear output layer, totaling 97{,}956 trainable parameters. The network is trained with the Adam optimizer on a sparse categorical cross-entropy objective computed directly on logits. On an 80/20 train-validation split, the model achieves 94.80% training accuracy and 96.01% validation accuracy with a validation loss of 0.1437. A Tkinter-based graphical interface allows a user to draw a shape with the mouse and receive an immediate class prediction with a confidence score. We situate this system within the broader sketch and shape-recognition literature, compare its accuracy against related hand-drawn shape classification studies, and discuss the limitations inherent to a small, single-contributor dataset. The complete source code, trained model, and per-class datasets are released publicly to support reproducibility.

The intensification of climate change poses a growing threat to food security, especially in vulnerable communities. This study employs an observational machine-learning framework to estimate the causal association between access to credit and acute food insecurity in Somalia and across the Horn of Africa, drawing on a harmonized dataset spanning key environmental, socioeconomic, and conflict-related factors from 2015 to 2022. Results indicate that greater credit access is associated with a 2% reduction in acute food insecurity at the population level over the study period. Given that, on average, 16% of the population is in crisis, this effect represents a meaningful shift within the at-risk group. We interpret these estimates under explicit identification assumptions and complement them with robustness and refutation tests. The results provide context-specific evidence on how financial access correlates with food security outcomes in data-scarce, crisis-affected settings, and offer a transparent framework for integrating heterogeneous data sources when randomized evaluations are infeasible.

RLVR has substantially improved the reasoning capabilities of LLMs. However, existing methods typically parameterize temporal progression in the Markov Decision Process by token-by-token generation, despite the highly non-uniform information flow along autoregressive trajectories. In this paper, we propose InfoPPO, which reparameterizes temporal progression using information density rather than raw token count. This reparameterization induces a common state-dependent structure for both temporal credit propagation and policy updates. InfoPPO restores the effectiveness of non-trivial discounting in long-horizon reasoning, retaining effective-horizon contraction while avoiding excessive attenuation of terminal supervision over long token sequences. Moreover, the information-time policy-improvement analysis naturally leads to a state-dependent update constraint, which we implement through adaptive clipping. By adapting the clipping threshold at each token position to the information density of its corresponding state, this mechanism enables more targeted policy updates while preserving proximal control. Theoretically, we extend performance-difference and policy-improvement analyses to the information-time MDP, deriving a policy-improvement lower bound when policy changes are regulated by information density. We further connect the general information-time analysis to practical LLM policy optimization by relating state-wise information density to local policy movement, while also providing theoretical grounding for the adaptive update mechanism. Experiments on Qwen3 models demonstrate consistent gains over competitive baselines across five challenging competition-style mathematical reasoning benchmarks. InfoPPO also maintains stable accuracy and response length across non-trivial discount settings under which token-time PPO deteriorates.

Mechanistic interpretability of vision transformers seeks to decompose model computation into human-readable units, but learned representations entangle many concepts in each neuron. Feature superposition is widely treated as the central obstacle to this decomposition, yet most mitigations (sparse autoencoders, dictionary learning) are post-hoc and leave the underlying network unchanged. We ask whether a spatial-locality training loss (TopoLoss) can act as a lightweight, training-time prior that improves interpretability of standard mech-interp tools. Training ViT on ImageNet-100 across multiple TopoLoss weights $α$, we measure causal sufficiency of topographic clusters via activation patching and feature geometry via sparse autoencoders fit to the same residual stream. At $α=1.0$, topographic clusters are 2.79$\times$ more causally sufficient than random unit sets of the same size, with the effect increasing monotonically in $α$. SAE L0 sparsity decreases by 11% and dead-feature fraction rises 19-fold, yet standard neuron-level monosemanticity scores are unchanged, indicating that topographic pressure acts at circuit level, concentrating causal mass into spatially local structures without disentangling individual neurons. This dissociation suggests current neuron-level monosemanticity metrics are insensitive to a class of real interpretability gains, and positions cheap architectural priors as a viable training-time complement to post-hoc tooling.

We study latent-space watermarking through a pretrained generator using a prescribed latent-to-output stochastic mapping, called the renderer. A watermark encoder selects the latent input using a message and secret key. For every message and semantic context, the released output must have exactly the desired conditional output distribution. For finite alphabets, we derive rate--key inner and outer bounds and characterize the coding and coordination requirements for realizing watermark communication through the prescribed latent interface. When the target output distribution of the generator uniquely determines the corresponding latent input distribution through the renderer, a strengthened converse yields the capacity region; the same region governs explicit preservation of the pretrained latent distribution. We extend the analysis to general jointly Gaussian models and identify a sufficient statistic of the latent that captures both the watermark-bearing information available at the generated output and the latent coordination required to preserve its target distribution. For the vector Gaussian model, we further characterize the optimal allocation of the secret-key resource across the resulting modes. Finally, we turn to an emerging robustness threat that is particularly natural in generative watermarking: an adversary can regenerate the released sample to obtain a fresh realization of the same underlying content while attenuating or destroying the embedded watermark. We incorporate this robustness axis into our framework and characterize the one-pass compound capacity of the scalar Gaussian model when the semantic context is known to the encoder but hidden from the detector, while the regeneration attack may depend on that context. Extending the analysis to multiple rounds of repeated canonical regeneration, we characterize the resulting watermark-capacity decay.

In-context learning (ICL) based on large language models (LLMs) has shown promising potential in alleviating performance bottlenecks caused by the limited availability of annotated data in Named Entity Recognition (NER). However, existing methods still face issues of retrieval misalignment and generation uncertainty, making their performance heavily dependent on the LLM's capabilities. As the parameter scale of LLMs decreases, their performance in few-shot settings deteriorates significantly. In this paper, we propose a novel unified retrieval-augmented framework, URA-NER, including three key components: Progressive Granularity Retrieval (PGR), Model-aware Representation Enhancement (MaRE), and Reason-aware Knowledge Verification. PGR is a two-stage retrieval mechanism that achieves stage alignment. It first retrieves demonstrations for span detection based on the query's global semantics, and then for type classification based on the specific entity context, providing fine-grained local information. Moreover, MaRE employs entity pre-recognition to guide the construction of representations, ensuring the query and demonstrations are aligned within the LLM's semantic space and attention pattern. In addition, to mitigate generation uncertainty, we propose RaKV, a closed-loop "generation-retrieval-verification" process. It explicates the LLM's reasoning paths, leverages them for the retrieval of external knowledge, and reorganizes the knowledge into verification evidence aligned with the original reasoning paths. We conduct extensive experiments on multiple low-resource NER datasets. Results demonstrate that URA-NER significantly enhances the performance of LLMs under low-resource settings, with particularly pronounced gains for smaller LLMs, achieving new state-of-the-art results on several benchmarks.

Self-attention is central to modern Transformer architectures, but its dense dot-product formulation makes it difficult to identify which internal directions are structurally important and which can be modified without disrupting the model. SVD-Inspired Attention (SVDA) addresses part of this problem by introducing a learned diagonal spectrum into the query-key score interaction, making latent attention directions explicitly inspectable through indicators such as spectral entropy, effective rank, sparsity, alignment, selectivity, and perturbation response. This paper examines the transition from diagnostic interpretation to operational intervention. A diagnosis--intervention--verification framework is proposed, and one intervention is evaluated: spectral energy retention in the attention-score pathway. Across FashionMNIST, CIFAR-10, CIFAR-100, and Food-101, the $ρ=0.90$ prescription removes 24.5--53.7\% of score directions, reduces parameters by 2.6--4.3\%, and reduces estimated MACs by 2.8--5.4\%. The paired mean accuracy change of the dimension-reduced model ranges from $-0.03$ to $+0.05$ percentage points over three seeds. These results support SVDA as an intrinsically interpretable attention mechanism whose learned spectrum exposes an operational coordinate system for deterministic and verifiable modification of attention-score formation.

Deception plays a central role in Intelligence operations, yet it remains difficult to analyse systematically without expert knowledge of reasoning patterns and cognitive manipulation. In computational argumentation, for instance, no scheme-level ground-truth corpora currently exist to support statistical validation. In this paper, we address this by introducing a set of ten argument schemes designed to model distinct forms of deception, each accompanied by structured premises and critical questions. In doing so, we introduce the first dedicated library of argumentation schemes specifically designed for deception analysis, providing a structured foundation for systematically modelling and analysing deception in narrative text. We then present \textit{DeceptionAnalyser}, a browser-based tool that implements these schemes through a two-stage methodology combining LLM-based premise extraction with critical-question-driven evaluation. Our aim is to provide a conceptual and methodological foundation for analysing deceptive reasoning in narrative text. This is precisely what we address in this paper by demonstrating how structured argumentation theory and AI-assisted analysis can support transparent, explainable assessments of potential deception. Because the schemes are designed to flag claims for scrutiny rather than to output a deception verdict, we do not benchmark classification accuracy; instead, we assess the \emph{reliability} of the methodology by measuring the consistency of the tool's premise and conclusion assessments across ten contemporary large language models and repeated runs. We find that scheme detection is highly stable for clear-cut deception and degrades gracefully, in interpretable ways, on more ambiguous intelligence-style narratives.

Sandboxed computer environments support multi-step reasoning with tools, executable programs, and persistent files, yet their extension from language models to vision-language models (VLMs) introduces a distinct state-management problem. Visual reasoning produces intermediate image-valued evidence---crops, masks, overlays, zoomed regions, and analytic renderings---that must remain addressable without accumulating unboundedly in multimodal context. We introduce VLM-in-Sandbox, a training-free framework for agentic multimodal reasoning in controlled computer environments. Its Visual Workspace registers generated artifacts in an image ledger, maintains a bounded active visual context, and lets the model explicitly promote selected evidence for subsequent inspection. This separates visual evidence generation, performed by sandbox tools, from visual evidence management. Across seven benchmarks and four base VLMs, VLM-in-Sandbox achieves the highest sample-weighted average accuracy among Vanilla VLM, Append-only Sandbox, and the proposed method. A compiler-matched $2\times2$ study on 1,260 examples further separates model-directed visibility from bounded retention: VLM-in-Sandbox reaches 66.27% accuracy with 18.6% fewer total tokens than the automatic, retain-all control. Over all 6,350 submitted GPT-4.1-mini examples, it produces 302 rescues and 142 regressions relative to Original Append-only. A local vLLM study with prefix caching confirms that the smaller request workload also reduces uncached tokens, time to first token, and end-to-end latency. These results identify explicit visual evidence state as a central abstraction for sandboxed VLM agents.

Image forensics is increasingly an open-world problem: manipulations range from fully synthetic images to localized edits, splicing and swapping, while most forensic detectors remain specialized to a single manipulation family. Agentic AI has recently emerged as a promising solution. In principle, such systems can assess the reliability of individual detectors, identify out-of-scope evidence, and arbitrate conflicting reports. However, it remains unclear which components actually drive performance and whether their benefits persist under distribution shift. To answer these questions, we study a training-free agentic framework built around specialist detectors, per-detector triage, and conflict-aware evidence arbitration. Using six configurations and three multimodal large language model backbones, we dissect the role of triage, prompting, and reasoning quality on both in-distribution and out-of-distribution data. Our results show that naive detector fusion suffers from severe false-positive rates on authentic images. Triage and prompting consistently improve performance by filtering unreliable evidence and exposing detector limitations. However, the dominant factor is represented by reasoning itself: A stronger judge substantially outperforms a weaker one, particularly under distribution shift. Most notably, manipulation recall is nearly saturated across all configurations, indicating that the main challenge of open-world image forensics is not detecting manipulations, but calibrating trust in specialized forensic tools and arbitrating conflicting evidence.

Predicting when maritime systems require maintenance can be critical, avoiding hazards and costly consequences. To address this problem, this paper proposes an explainable decision-making framework that integrates a neuro-fuzzy prediction model with a two-stage explainable component. The first stage of this component produces feature-attribution explanations, using gradient-based saliency maps, and the second stage extracts local rules using a fuzzy decision tree. The proposed framework is generic and can be integrated into any deep learning-based approach, rendering it explainable. To the best of our knowledge, this is the first fuzzy logic-based framework enabling both feature-level and local rule-based explanations of black box models. This approach aims to foster trustworthiness in decision making through user-understandable machine inferences. The performance of the proposed framework using a deep residual-based neural backbone is evaluated on various general-purpose public benchmark datasets, and its utility in maritime is demonstrated in the context of early fault detection in a naval propulsion system dataset. The results indicate that it can provide predictions outperforming relevant state-of-the-art approaches, with an average AUC-ROC (Area Under the Receiver Operating Characteristic Curve) value, reaching up to 99%, while offering the advantage of explainability.

Cross-domain Named Entity Recognition (CD-NER) aims to transfer the rich knowledge in the source domain to the target domain. Recent studies adopting decomposition or generation paradigms have achieved significant performance improvements, demonstrating high accuracy in entity span detection. However, during entity type classification, models severely suffer from entity type confusion, the erroneous tendency that models classify entities of one type in the text as another similar but incorrect type. To address this issue, we first propose a Multidimensional Confusion Quantification Model (MCQM) that quantifies a model's confusion extent between entity types from three dimensions: source-target hierarchy analysis, semantic similarity analysis, and explicit data evaluation. Moreover, we propose the Progressive Bidirectional Reasoning Chain (PBRC). PBRC leverages the source-target hierarchy and confusion analysis from the MCQM to prompt the LLM to generate two-stage reasoning information. The two-stage reasoning information is utilized to augment the knowledge of the model, significantly mitigating entity type confusion and improving the model's generalization performance. Experimental results demonstrate that our method achieves new state-of-the-art results on all domains of the CrossNER dataset.

Large language models (LLMs), when acting as agents, are expected to take observed data in context, infer the latent state space underlying the world, and leverage it for downstream prediction. However, prior work demonstrated that LLMs struggle to use representations learned in context on a graph tracking task, where the model needs to construct a representation of the graph governing data generation process and use it for subsequent predictions. In this paper, we show that extending this to few-shot settings, where each demonstration is generated from a different world with either the same or different graph topologies, enhances its prediction on 6 models from 4 model families. To understand this improvement, we linearly probe a low-dimensional world representation that encodes graph information in the hidden states. Notably, we find that few-shot demonstrations relocate the world representation and increase its predictive use. Specifically, for each model, these world representations shift in directions nearly orthogonal to their original subspace, and interventions on these representations selectively impair performance more than interventions on other subspaces. Consistent with this insight, we show that few-shot demonstrations with observations from different worlds improve performance on ARC-AGI-1&2, web agent tasks, and Othello. Our findings elucidate the role and internal mechanisms of few-shot demonstrations in in-context world modeling. More broadly, our work advances our understanding of how LLM agents learn from in-context observations and provides implications for their further improvement.

AI coding agents increasingly support software development beyond code completion, including planning, implementation, testing, and repository-level task execution. Their practical use, however, often remains only weakly connected to established software engineering practices. The aim of this work is to develop and evaluate a lightweight, spec-driven lifecycle for governed agentic software engineering. The lifecycle combines established software engineering practices with repository-local guidance through specifications, AGENTS.md, and phase-specific agent skill files. The approach was developed in the context of the FHNW course AI-assisted Software Development and applied by students to business-oriented software use cases. Its educational and practical applicability is explored through a student survey combining closed rating items with open-ended questions. The contribution of this work is a process-oriented framework that enables AI coding agents to operate with bounded autonomy within an explicit, reviewable, and test-oriented software development lifecycle.

Graph Retrieval-Augmented Generation (GraphRAG) has remarkably enhanced large language models on complex reasoning by leveraging structured entity topologies. However, existing frameworks heavily rely on standard autoregressive language models where the nature of inherent sequential generation severely hinders overall inference efficiency. Inspired by Diffusion Language Models (DLMs) that offer massive parallelism via continuous refine-in-parallel decoding, we aim to accelerate GraphRAG in the discrete space. However, it remains non-trivial for two challenges. First, partially denoised drafts are highly dynamic and uncertain, making dynamic graph grounding non-trivial. Second, raw denoising states are inherently noisy and unstable, making synchronous graph retrieval and multi-hop aggregation computationally prohibitive. To this end, we present LADDER, a novel framework that bridges diffusion language modeling with GraphRAG through graph-guided parallel decoding. Specifically, (i) we propose an event-driven self-clocking retrieval, inspired by our key insight that 88% of target entities emerge early in the partially denoised state, leading final commitment by an average of 5.7-9.6 steps. This mechanism dynamically triggers graph retrieval only when the set of graph-linkable entities expands, yielding an asynchronous self-clocking policy that bypasses learned gates or heuristic thresholds. (ii) An incomplete-query graph propagation module is designed to process the newly emerging entity queries using a specialized graph foundation model, continuously aggregating multi-hop evidence to sharpen parallel predictions and accelerate overall decoding convergence. Extensive experiments on three challenging multi-hop QA benchmarks show that LADDER raises average exact match from 39.6% to 45.2% while achieving a 4.1x latency reduction.

Context: Large language models (LLMs) can generate unit tests quickly, but high structural coverage does not establish that those tests execute reliably or detect faults. Existing evidence often treats coverage as the principal outcome and rarely compares prompt strategies and models through mutation testing at class level. Objective: This study examines how prompt strategy and model choice shape the executability, structural coverage, fault-detection effectiveness, and structural quality of LLM-generated Python test suites relative to human-written suites. Method: We evaluate multiple prompt strategies across a diverse set of current LLM configurations on the ClassEval benchmark. The evaluation combines execution outcomes, line and branch coverage, Cosmic Ray mutation scores, and structural quality indicators. Primary analyses treat successful execution as a prerequisite; paired comparisons use only classes shared by the relevant executable subsets. Results: Structural coverage is consistently near its ceiling and offers little discrimination among configurations. Executability varies substantially. The proposed prompt performs strongly for mutation score, but no prompt dominates across models. Model choice explains more variation than prompt choice, and their interaction shows that prompt effectiveness depends on the selected model. Human and LLM suites are evaluated on unequal executable subsets, so their relative mutation scores do not establish superiority. Conclusion: Reliable assessment of LLM-generated tests should treat executability as a gate and combine coverage with mutation testing and structural quality indicators. In practice, model selection should precede prompt tuning.

An Intraoperative Hypotension (IOH) event is a frequent complication during administration of general anaesthesia with serious downstream consequences, yet clinical management remains reactive and not predictive. Existing predictive models, however, ignore drug infusion history as a valuable signal for prediction despite its direct pharmacological relevance. Our model achieves an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.7360 and an Area Under the Precision-Recall Curve (AUPRC) of 0.1794, representing a 2.73-fold lift over the random guessing AUPRC baseline (0.0657), with the removal of propofol and remifentanil effect-site concentrations resulting in a 13.9% AUPRC drop compared to the full model. This is consistent with the hypothesis that pharmacokinetic trajectories encode impending haemodynamic changes before they manifest in the Mean Arterial Pressure (MAP). Additionally, this paper shows that training without lead-gap filtering degraded AUROC by 16.7%, empirically confirming that unfiltered models learn to detect ongoing hypotension rather than predict future events. Finally, a Mamba-based architecture achieves the aforementioned high prediction performance while maintaining a constant memory footprint across a range of sequence lengths, unlike the quadratic VRAM overhead typical of vanilla Transformers, making it the more practical choice for continuous intraoperative deployment.

Combining predictions from different models can improve performance at machine learning tasks, but the training of the individual models and the rule used to combine them are typically chosen separately, and by ad hoc means. Recent advances in distributional optimisation (i.e. where the optimisation occurs over the set of probability distributions) offer an opportunity for principled joint training, viewing the collection of models as a discrete distribution whose support points are to be optimised, but the potential of these methods is not well-understood. In this paper we (1) cast two standard combination strategies - ensembles and low-rank adapter averaging - as entropy-regularised distributional optimisation, observing that the resulting objective is convex in the ensemble case but not in the adapter-averaging case, so that existing convergence guarantees for mean field Langevin dynamics transfer only to the former; (2) assess existing and novel algorithms for this task, including a functional variant of variational gradient descent; and (3) report an empirical study spanning synthetic classification tasks and fine-tuning of large language models on a commonsense reasoning benchmark.

Electroencephalography (EEG) provides a non-invasive window into dynamic brain activity, yet modeling long-horizon EEG sequences remains challenging due to their high temporal complexity, substantial variability across subjects, and the lack of biologically meaningful sequence representations. Existing tokenization strategies, such as fixed-window and patch-based representations, discretize EEG signals according to artificial temporal boundaries, which may disrupt intrinsic brain-state dynamics. In this work, we propose Brain-Token Learning, a neuroscience-inspired framework that introduces Brain Tokenization for long-horizon EEG sequence modeling. Instead of partitioning EEG signals into predefined temporal segments, Brain Tokenization represents EEG as sequences of recurrent microstate-derived brain tokens, where each token corresponds to a quasi-stable large-scale brain state with variable temporal duration. Based on these biologically grounded tokens, we further develop a multi-scale token interaction module consisting of Latent State Aggregation and State Transition Modeling to jointly capture global brain-state context and local microstate transitions. We evaluate Brain-Token on five heterogeneous EEG datasets, including the newly collected long-horizon NeuroLong dataset and four affective or clinical EEG datasets (SEED, DEAP, MDD, and NSSI). Extensive experiments demonstrate that Brain-Token consistently outperforms conventional CNN/LSTM architectures, Transformer-based models, and domain adaptation methods across diverse EEG scenarios. Further analysis verifies the effectiveness of microstate-based tokenization and multi-scale interaction for learning robust and interpretable EEG representations. These results establish Brain-Token as a biologically grounded tokenization paradigm for long-horizon EEG sequence modeling.

We show that a quantized model that keeps its classification accuracy still changes $14$ to $46\%$ of its top-1 retrieval results, and that aggregate ranking metrics reveal only part of this damage. We tie this failure to the gap between the two highest scores and use that gap to decide when a quantized answer can be trusted and where additional precision should be spent. We show that the top-1 result is guaranteed to survive quantization only when this gap exceeds twice the largest rounding error. In classification, scores are the logits, and the loss function pushes the correct class away from other classes, encouraging this gap. In retrieval, scores are query-document scores, and nothing separates the top-1 item from the second. This gap can be measured without labels. Before deployment, it predicts which models will break under quantization, and at deployment time it tells, per input, whether the quantized answer still matches the full-precision answer. Most classification inputs have a gap wide enough to trust the quantized answer, but few retrieval queries do. That gap motivates a different fix in each task. In retrieval, spending extra bit-width on the layers whose quantization moves the gap most recovers up to three-quarters of an extra bit's benefit for half its cost. In classification, routing the few low-gap inputs to full precision recovers most of the lost accuracy at a fraction of the cost.

Text-to-speech models for Bantu tonal languages are challenged by a tonal system that is rooted in both the lexis (i.e., the inventory of words, stems, and affixes) and the grammar (i.e., morpho-syntax). To complicate matters, the standard writing systems of these languages often omit tone markings and syllable duration information, which must be disambiguated by the reader based on context. Motivated by linguistic descriptions of Bantu language tone systems, we propose an end-to-end text-to-speech model that augments the text encoding mechanism with a morpho-syntactic prior. We replace the standard phoneme encoder in the VITS architecture with a morpheme sequence encoder and a phoneme-to-morpheme attention network. We posit that, by using this explicit morphological modeling, we can capture the information required to produce the correct tone. Experiments conducted on the Kinyarwanda language, a tonal and morphologically complex Bantu language, reveal substantial TTS improvement from this morphological modeling. Specifically, the proposed method significantly improves the naturalness, intonation, and intelligibility of the produced synthetic voices.

Post-training often improves task performance but can degrade confidence calibration, leaving post-trained language models (PoLMs) more overconfident than their corresponding pretrained language models (PLMs). Because task-specific labeled calibration data can be costly or unavailable, the corresponding pretrained PLM provides a natural label-free reference for post-hoc calibration. Prior agreement-gated PLM-referenced calibration fits a scalar temperature using only examples on which the PoLM and its PLM reference agree, excluding disagreement examples because direct alignment can drive the fitted temperature excessively high and induce under-confidence. We revisit this binary treatment. A controlled reintroduction diagnostic reveals a non-monotonic aggregate effect: admitting a moderate fraction of disagreement examples can improve calibration, whereas the benefit diminishes as unit-weight inclusion approaches the full disagreement set. We introduce SupportCal, a label-free post-hoc method that retains agreement examples at unit weight and assigns disagreement examples continuous weights based on the own-base PLM's relative support and corroboration from pretrained references selected from a size-compatible candidate pool. We further characterize when the resulting weighted objective admits a finite optimal temperature. Across MedMCQA and MathQA, SupportCal yields lower ECE than the agreement-only baseline for nearly all evaluated target-model configurations; supplementary TweetEval Sentiment results show the same pattern on a fixed-label classification task.

Virtual screening aims to prioritize active compounds from large chemical libraries within a limited experimental budget. When applying Boltz-2 to virtual screening, a key challenge is how to use limited experimental data from the target assay to improve the prioritization of active compounds. We investigated whether fine-tuning the Boltz-2 affinity heads with a small number of binary activity labels could improve early enrichment of active compounds in hit discovery. We compared fine-tuning with 40-300 labels in a retrospective evaluation on eight MF-PCBA targets. With 300 activity measurements, fine-tuning increased the number of actives in the top 1% by a geometric mean of 1.77-fold across the eight targets and improved average precision (AP) by 2.14-fold relative to the control without fine-tuning. We also investigated whether rescoring a subset of candidates could retain the improvement in hit recovery by reranking only the top-ranked Boltz-2 candidates with the fine-tuned head. Restricting rescoring to approximately 10% of the evaluation set retained hit recovery comparable to full rescoring. These findings show that affinity-head fine-tuning with limited activity labels improves early enrichment with Boltz-2 and that this benefit can be retained when rescoring a restricted set of candidates.

What limits KV-cache compression at extreme bit-rates? We argue that it is not the choice of compression scheme, but how its budget is allocated across attention heads. Existing methods apply rank and bit-width uniformly, ignoring that each head has a different optimal mix of rank truncation and quantization. We show that co-optimizing rank and bit-width per head, using only standard low-rank projection and scalar quantization, dominates uniform allocation, with the largest gains at low bit-rates. Our method, KV-COBRA (Co-Optimized Bit-Rank Allocation), formalizes this as a resource-allocation problem: it balances rank-truncation loss against quantization loss within each head, then redistributes budget across heads to minimize total distortion. A fused Hadamard rotation equalizes per-channel variance, and reordering the SVD basis by attention-KL importance makes the solver query-aware. The same allocator extends to joint $K{+}V$ compression. On perplexity, zero-shot, and long-context benchmarks from $0.5$ to $4$ bits per dimension (bpd), KV-COBRA shows the smallest accuracy degradation among evaluated methods at low bpd, with no per-token overhead.

Instruction-tuned LLMs faced with underspecified queries often commit to a single interpretation rather than ask for clarification, producing confidently wrong answers. In our experiments, prompting alone is insufficient: models either ask for clarification on every query or ask vague questions that fail to recover the missing information. Addressing this failure requires learning two coupled skills: when to ask rather than answer and how to ask a question that recovers the disambiguating information. Existing recipes either address only one of these skills or require a separately trained critic. We propose CIGAsk, an RL recipe that teaches both skills through two complementary reward signals within a multi-turn GRPO loop. Counterfactual Information Gain (CIG) compares the gold-answer log-likelihood under a frozen reference model with and without the user response, providing per-turn credit that guides how to ask. The Asymmetric Ambiguity Bonus assigns a signed reward at the terminal token based on the gold ambiguity label, guiding when to ask. Across three clarification benchmarks spanning table, passage, and open-domain QA, CIGAsk-7B outperforms the strongest external baseline despite using a smaller backbone. It also transfers across datasets without per-dataset tuning while preserving single-turn QA performance on out-of-distribution benchmarks.

As Large Language Model (LLM) agents are applied in continuously interactive environments, driving the evolution of their own capabilities becomes a core problem for achieving long-term autonomy. Currently, environmental knowledge is typically treated as an external fixed input rather than as part of the agent's ongoing evolution. Reinforcement learning methods usually optimize policies through environmental interaction but tend to adapt only to fixed task distributions or single environments. This paper proposes TTSE (Two-Track Self-Evolution), a dual-track online self-evolution framework that separates evolving knowledge into FACT (environmental facts, whose reliability is continuously verified through interaction evidence) and TIP (task-conditioned implementation procedures). From a decision-theoretic perspective, we decompose the agent's excess risk into environment-representation regret and conditional-execution regret, characterize the conditions under which environment-conditioned policies strictly outperform condition-agnostic policies, and bound the downstream risk in terms of FACT identification error and cross-condition mismatch cost. In practice, TTSE's ablation experiments on GDPevo validate the advantage of dual-track evolution. On the classic agent task benchmarks ALFWorld and ScienceWorld, TTSE further demonstrates superior task adaptation. Moreover, TTSE is broadly compatible with existing skill self-evolution methods; combined with the Bayesian-Agent algorithm, a single-track ablation validates the dual-track advantage, substantially improving the aggregate score across the five major domains of SOPBench over three independent repetitions. Finally, on the real end-to-end task benchmark PinchBench, TTSE is integrated into a general agent framework via retrieval-based injection and stably outperforms the baseline across three independent runs.

Malware detection is a critical task in cybersecurity, and graph neural networks over control flow graphs have shown promising results for it. However, detectors are usually evaluated on a random split of a corpus collected over a single period, which cannot show how well a model generalizes to later samples. This study addresses that limitation with a strict temporal split: every model is trained on one period and scored once on a later one. Two corpora of control flow graphs, each node carrying 37 features, were extracted statically from 1,989 Windows portable executables: 459 graphs from 2024-2025 for training and 223 from 2026 for evaluation. Twelve variants and a flat-feature control were trained on the earlier corpus. The choice of message-passing operator changes robustness to the shift significantly, and every pairwise gap that survives correction separates an aggregating architecture from one built around a learned attentional readout. The ranking also reverses: the flat control, which sees node features but no topology, is the best in-distribution model and among the worst across the boundary, so a conventional benchmark would have rejected message passing. Neither recalibration nor ensembling substitutes for the operator choice. Attributions do not shift, but explanation validity is architecture-specific, and the most accurate operator on the later corpus is the hardest to explain. An architecture derived from the finding matches the best searched operator without search. The shift affects both malware and benign classes alike, so these are results about robustness to distribution shift, not malware evolution.

Change point detection (CPD) identifies abrupt and significant changes in sequential data, with applications in human activity recognition, financial markets, cybersecurity, manufacturing, and autonomous systems. Traditional CPD methods often face computational challenges in high-dimensional settings and typically provide limited explanations for detected changes, which can restrict their practical usability. This paper introduces a CPD framework that improves scalability and interpretability by leveraging the Sliced Wasserstein (SW) distance. Our contributions are fourfold: (1) we transform multivariate sequential data into one-dimensional scores using the SW distance, making the resulting representation compatible with existing CPD methods; (2) we analyze the distributional behavior of random slices of the SW distance and show that, under suitable assumptions, they can be approximated by a Gamma distribution, providing a principled basis for threshold calibration; (3) we propose a self-adapting online CPD algorithm that combines this SW-based score with an adaptive quantile-based threshold; (4) we introduce a model-specific framework for generating contrastive explanations for annotated change points. Empirically, our method reduces false positives by at least $48\%$ on average compared with popular online and offline CPD baselines, while maintaining competitive or superior detection performance. Code is available at https://github.com/jsve96/SWCPD_Code. At the same time, it produces interpretable change-point annotations, making it practical for deployment in high-stakes applications.

GUI agents predict click coordinates as digit-token sequences, but standard text-LLM confidence estimation methods rank correct clicks from wrong ones only weakly. GUI-specific alternatives use K samples or new supervision, but still leave room for improvement. We trace part of this to place-value asymmetry: bounding-box correctness often makes higher-place digits more important than lower-place digits, so uniform aggregation weakens the signal that determines correctness. The fix is to weight each digit's Shannon entropy by its place value. We call this Place-Aware Coordinate Entropy (PACE). Across fixed-scale agents on ScreenSpot-Pro and ScreenSpot-v2, PACE wins both AUROC and selective accuracy on all primary comparisons in a single forward pass, matching or outperforming K-sample baselines at a fraction of the cost. PACE provides a per-click confidence estimate that turns coordinate-token internals into a practical confidence signal for GUI agent deployment.

Text-to-speech (TTS) corpora are costly to record, yet many utterances add little new phonetic information. Core-set selection reduces this cost by choosing a small training subset under a fixed audio-duration budget. We represent a corpus as a phonotactic graph that links each utterance to its most phonemically similar ones, and we first test whether this graph has structure. In Bangla and English corpora, its clustering is 199 and 56 times that of a size-matched random graph, and its modularity is more than twice that of a degree-preserving random graph. We then propose Community Representative, a selector that samples across graph communities and spreads its choices within each one, starting from utterances rich in rare phonemes. At every budget and in both languages, it covers more rare phoneme bigrams than random and entropy-based selection, and this lead holds on held-out utterances. TTS models trained on its 20% core-sets have a significantly lower character error rate (CER) than models trained on equal-duration random or entropy-based subsets in both languages. When all models train for the same number of epochs, the Bangla core-set model also outperforms full-corpus training (3.93% vs. 4.47% CER) with 4.5x less training time.

Deploying language-conditioned manipulation without a dedicated GPU requires efficient inference and action chunks that cover the delay between policy queries. We present vla.simd, a CPU inference engine that combines shared SIMD micro-kernels, reusable computation, and target-specific optimization. We relate query latency and execution horizon to action availability under lagged and time-aligned execution, distinguishing action supply from feedback frequency. Across six policies and four CPUs, vla.simd achieves approximately $1.4\times$ median speedup over compiled PyTorch references while preserving fp32 numerical fidelity. We also introduce IMPACT, an ACT-based policy with cached text representations and language-modulated visual features. IMPACT is the only language-conditioned policy in our evaluated set that supplies at least 30 actions/s on the Raspberry Pi 5: after a 90 s thermal soak, it supplies 33.5 actions/s in fp32 and 81.2 with int8. Separate GPU evaluations yield $76.4\%$ mean success across four LIBERO suites without robot pretraining; instruction-shuffling tests demonstrate selection among familiar goals. Trials with IMPACT on an SO-101 arm and SmolVLA on a UR10e with a Robotiq gripper demonstrate CPU deployment on two robot embodiments.

Unsupervised anomaly detection (UAD) aims to localize abnormal regions in medical scans without pixel-level annotations. A typical strategy seeks to reconstruct a pseudo-healthy image that preserves subject-specific anatomy. Recently, diffusion models have been proposed to perform UAD. However, these methods rely on heuristic noise schedules or synthetic corruptions to balance subject-specificity and anomaly removal. In this work, we propose an alternative formulation of UAD as a Bayesian inverse problem under a diffusion prior. First, we introduce a latent spatial anomaly mask that models pixel-wise consistency between a test image and its latent corresponding pseudo-healthy image. Then, we propose an approximation of the unknown generation process that links healthy anatomy, anomalies, and the observed image, enabling a well-defined likelihood within the Bayesian framework. Building on recent advances in diffusion-based inverse problem methods, we jointly infer the pseudo-healthy image and the anomaly mask via annealed posterior sampling. We evaluate our approach on FDG PET (ADNI) and FLAIR MRI (BraTS 2021), demonstrating improved anomaly localization performance compared to other diffusion-based approaches and validating the contribution of our introduced model. Our code is available at https://github.com/HuguesRoy/UAD_DAPS.

Tool-use agent traces identify messages and API calls, but procedural analyses also need explicit units of action and inspectable links to their evidence. We present Canonical Procedural Actions (CPAs), an annotation protocol that records a procedural function, its first agent-event anchor, the agent events that realize it, and separate contextual evidence. Multiple actions may share a message anchor without an inferred within-message order. A retail case study produces a versioned 24-entry codebook through open induction, recorded consolidation, and successive application audits. Two isolated LLM contexts annotate 32 trajectories disjoint from development at the trajectory level, producing 499 and 491 occurrences with anchor-label overlap A=0.982. Requiring identical context-event references reduces overlap to 0.798. These are structural repeatability measures, not semantic accuracy: 16 of 26 task IDs also occur in development, and historical tool payloads were truncated to 110 characters. Retrospective controls show that collapsing all labels raises overlap to 0.986, while simple endpoint rules reproduce the tool-anchored portion with 0.997 overlap. Assistant-message actions have 0.971 overlap, with a per-label minimum of 0.816. Applying the frozen codebook to 244 further trajectories yields 4,058 records, including eight diagnostic outcomes. The contribution is an explicit, auditable annotation instrument and a case study of its construction and measurement limits; human-reference validity and downstream utility remain to be established.

We study the problem of \emph{adversarially robust} PAC learning. In this framework, the learner observes independent samples from an unknown distribution over $\mathcal{X} \times \{0,1\}$, as in classical PAC learning. However, given a perturbation map $\mathcal{U} : \mathcal{X} \to 2^{\mathcal{X}}$ known to the learner, the goal is to output, with high probability, a predictor that correctly classifies \emph{every} perturbation $z \in \mathcal{U}(x)$ of most future examples $(x,y)$ drawn from the same underlying distribution. We determine the \emph{optimal} $\mathcal{U}$-independent sample complexity of this problem in both the realizable and agnostic settings. More specifically, for every concept class $\mathcal{H}$ of $\operatorname{VC}$ dimension $d$, we prove upper bounds of $\mathcal{O} \big( d/ε+ \log(1/δ)/ε\big)$ in the realizable setting and $\mathcal{O} \big( d/ε^2 + \log(1/δ)/ε^2 \big)$ in the agnostic setting, together with an optimal first-order refinement of the latter. These bounds match the corresponding lower bounds for classical PAC learning. Consequently, and perhaps surprisingly, adversarial robustness incurs \emph{no additional} distribution-free statistical cost, uniformly over all perturbation maps. Our bounds improve exponentially on those of [Montasser, Hanneke, and Srebro; COLT '19]. On the technical side, we present short and elementary proofs based on a new algorithmic principle that we call \emph{binomial-bagging}. We believe that binomial-bagging and its analysis may be of independent interest.

The effectiveness of agent memory ultimately depends on whether the underlying LLM gives each memory in context an appropriate degree of influence over its response. Yet this capability has remained largely overlooked. To assess this capability, we introduce MemCalib, a benchmark grounded in realistic memory-system scenarios for evaluating memory use and advancing optimization algorithms. Results on the MemCalib test set reveal that frontier open- and closed-source models struggle to use memory appropriately. They frequently over-use or under-use memory rather than matching each proposition's actual use to its target level, leading to biased, low-quality responses. Experiments with common post-training algorithms, including group relative policy optimization and on-policy self-distillation, further reveal a clear directional skew: trained models improve in one direction while deteriorating in the other. We therefore propose MemCalib-RL, an ordered bidirectional counterfactual credit-assignment algorithm that separates over- and under-use signals and localizes their credit to response tokens through exact atom ablation. Results across model families and scales (Qwen3-8B, Ministral-3-8B-Instruct, and Qwen3.5-35B-A3B) show that MemCalib-RL achieves the best overall performance while better balancing over-use and under-use, with gains generalizing beyond MemCalib in external benchmark evaluation. Further experiments support its design choices and robustness and provide insight into its training dynamics.

The shipping industry has a significant impact on the global economy, emphasizing the need for operational availability and safety through the use of effective maintenance techniques. During the last decades, predictive maintenance (PdM) has emerged as a promising solution compared to the existing conventional maintenance systems. This is because it offers several advantageous functions, such as damage predictions for vessel components, reduced downtime, improved and extended life of machinery, as well as higher safety during voyages. However, existing methodologies developed for performing PdM do not provide explanations of their results to users, so that they can understand the failures that may occur. To address this limitation, this paper proposes a novel framework based on a fuzzy decision tree and a deep residual neural network, aiming to perform explainable PdM on naval vessels. The proposed framework is able to generate fuzzy local rules based on the dataset used, and can provide explanations of its outcomes, using cause-and-effect relationships, in a way that are understandable to users, thereby gaining their trust. Experiments using a publicly available dataset demonstrate the effectiveness of the proposed framework, as it achieves an accuracy of 99.24%.

Neural networks, both convolution or transformer based, are essential for modern computer vision systems. However, they are vulnerable to small perturbations, almost imperceptible to humans, which significantly alter the model's prediction. These adversarial attacks are often considered to be a significant threat to the implementation of neural networks in safety-critical applications. Most attacks utilize the white-box threat model and therefore require full access to the target model, making them unrealistic to use in practice. We propose a novel approach under the more realistic black-box threat model that utilizes concepts from reinforcement learning to optimize perturbations with a non-differentiable target model. Reinforcement learning algorithms have already been optimized to be query efficient, making them an ideal starting point when designing black-box adversarial attacks. We show the success of our reinforcement learning inspired black-box adversarial attack (RIBA) in generating adversarial perturbations using only a small number of queries to the target model, by comparing it to state of the art attacks on different models on the Cifar10 and ImageNet data sets. RIBA takes $25.4\%$ fewer median queries to generate attacked images against a ResNet-18 on Cifar10 and $22.5\%$ fewer median queries to fool a Vit-B/16 model on ImageNet. Additionally, we demonstrate that RIBA can match the performance of white-box attacks on an adversarially trained model.

Large language models (LLMs) have shown remarkable progress in natural language understanding, yet their effectiveness in specialized fields like astronomy and astrodynamics remains limited due to challenges in multi-step reasoning, symbolic manipulation, and domain-specific terminology. To address this, we present Taramandal-GPT (Constellation-GPT), a domain-adapted framework built on the Qwen3-8b backbone, enhanced with a Retrieval-Augmented Generation (RAG) pipeline and a fallback mechanism for improved contextual precision. We evaluate it on the Astrodynamics Problems Benchmark (APBench), a dataset of 299 questions covering foundational to advanced levels of space science. Using a dual evaluation method - numeric margin-based scoring and semantic similarity assessment - Taramandal-GPT achieves competitive performance against state-of-the-art open- and closed-source models, with notable strength in thinking-intensive tasks. These results highlight the value of specialized LLMs for domains demanding accuracy and interpretability, positioning Taramandal-GPT as a step toward reliable Artificial Intelligence (AI) assistants for astrophysics, spacecraft engineering, and space exploration.

Reinforcement learning (RL) improves reasoning in vision-language models (VLMs) but can induce chain-of-thought (CoT) obfuscation: an operational, non-intentional outcome where task reward or accuracy rises while traces become less grounded and monitorable. Prior work largely documents this decay behaviorally, leaving its representation-level correlates and actionable controls unclear. We find that template- and ground-associated activations become less separable during RL; matched interventions support the contribution of selected features to monitorability degradation. Guided by this evidence, we propose Targeted Anti-obfuscation with Mechanistic Enforcement (TAME), which uses Sparse Autoencoders (SAEs) to combine behavioral feedback with targeted suppression of template-associated activations during RL. Its asymmetric constraint penalizes template activations only above their pre-RL baseline, anchoring the localized features while behavioral feedback promotes grounded refinements. Across VIRL-39k, SPA-VL, and two model families, TAME improves CoT monitorability by up to 30.9 and 16.7 percentage points over Group Relative Policy Optimization (GRPO), respectively. Blinded human evaluation finds higher human monitorability on both datasets, and two held-out monitor families reproduce the monitorability gains. Task accuracy changes are small and mixed, and general-capability benchmarks show task-specific trade-offs. These results provide a path from behavioral monitoring to representation-level oversight for more auditable RL-trained multimodal systems.

Uncovering latent variables and their causal relations from observed data is a fundamental yet challenging problem. Existing methods often rely on restrictive assumptions, such as linear relations or invertible mixing functions. To better address this problem under general nonlinear mixing procedures, we propose a condition called the cross-Hessian Rank Constraint (HRC), which serves as a primitive rank-based tool for nonlinear latent causal discovery. In particular, we show that a rank-based property arises from the cross-Hessian of the observed-data log-density in the nonlinear case, revealing information about the latent variables, and reduces to the Tetrad constraints in the linear Gaussian case. More specifically, when two groups of observed variables are d-separated by a set of lower-dimensional latent variables, the rank of this cross-Hessian is equal to the dimension of the latent variables, under a mild affine derivative assumption on the conditional log-density derivatives. This assumption can be naturally satisfied when the noise level is low or the relevant nonlinearity is moderate. As a downstream application, we instantiate HRC in the pure one-factor measurement setting for locating latent variables and recovering their causal structure up to Markov equivalence. Experimental results on synthetic and real-world datasets support the theoretical claims.

We reproduce and stress-test the work of Yu et al. (2023), who characterize how language models (LMs) arbitrate between memorized knowledge and contradictory in-context statements. We replicate their world-capitals experiments on 31 models spanning Pythia, GPT-2, Qwen3, and Ministral families, including base and post-trained variants, and extend evaluations to five additional knowledge relation types from the ParaConflict dataset. We empirically confirm most of their original findings: larger models and higher-frequency entities tend to favor memorized answers, with substantial family-level variance. However, several conclusions do not generalize cleanly: entity-frequency effects disappear on Qwen3-14B and 32B; post-training shifts the memory-context trade-off inconsistently across families; question phrasing alone can change a model's reliance on memorized knowledge by up to 80 percentage points; and semantically unrelated prose can mimic coherent supporting context. Our results clarify where Yu et al.'s claims hold and to what extent they generalize to other prompts.

Electroencephalography (EEG) provides non-invasive monitoring of brain activity and is widely used in emotion recognition, motor imagery and sleep staging. Although within-subject decoding has achieved considerable progress, cross-subject generalization remains a central challenge in practical applications. EEG decoders are typically trained with Adam/AdamW under a fixed second-moment decay coefficient, even though cross-subject learning involves low signal-to-noise ratios, subject variability, and gradient nonstationarity. A fixed coefficient implicitly assumes that gradient statistics are homogeneous across layers and time, which can limit model's adaptability to cross-subject EEG signals and degrade generalization. To address these issues, we propose AFOR, a tensor-wise adaptive optimizer that converts the fixed second-moment decay coefficient into a dynamic coefficient estimated online from local gradient state. AFOR combines a Residual-Alignment Signal Scorer (RASS) and an Adaptive Forgetting Controller (AFC). RASS summarizes local gradient residuals and directional agreement into a signal-quality score, and AFC maps this score through self-referential normalization to a bounded per-step decay coefficient, with cumulative-product initialization correction maintaining consistency under time-varying decay. Under a strict cross-subject protocol on three EEG benchmarks that cover three representative fields, AFOR achieves the best average performance among the compared optimizers, improving the mean test accuracy over Adam by 3.00%, 2.07%, and 4.38%, respectively.

Test oracles for quantum transpilers typically judge correctness by comparing compiled output against a reference: a statevector, a sampled distribution, or a unitary compared modulo global phase. A companion empirical study measures how often that choice fails. Roughly 28% of merged Qiskit transpiler bug-fixes (95% Wilson CI 19-40%) repair a fault that corrupts layout metadata, global phase, or run-to-run reproducibility while output stays correct: invisible to a black-box output-equivalence oracle by construction. This paper closes that gap with a fault-class-matched, layout-aware, width-tiered oracle family: a layout/permutation contract checker and a contract-level metamorphic relation (MR-1) for the metadata channel, a global-phase tracker for the phase channel, and a determinism runner for reproducibility. Verified from source on nine real, merged Qiskit transpiler regressions (three per channel), the output-equivalence oracle is blind throughout and the matched mechanism fires on every case. A 675-configuration sweep of the contract/metadata invariant finds no false positive. Synthetic mutant families confirm reliability at scale: 1.00 sensitivity and specificity across 36 mutants apiece for the contract/metadata and global-phase channels, and 1.00 sensitivity (95% CI 0.44-1.00) for reproducibility on the three circuits where the mutation is constructible. The contract checker costs two to six orders of magnitude less than a plain output check, the global-phase tracker is comparably cheap within its exact tier, and only the metamorphic relation carries a bounded cost. Ported natively to pytket/tket, the global-phase mechanism transfers cleanly, an identical 1.00/1.00 result with phases recovered to double-precision accuracy, evidence against a Qiskit-specific artifact.

Long-term time series forecasting has made significant progress by leveraging multi-scale information to capture hierarchical temporal patterns and model long-range dependencies. However, temporal downsampling in existing multi-scale methods inevitably smooths detailed temporal fluctuations, and this information loss is further aggravated by their emphasis on dominant trends across scales, resulting in insufficiently expressive representations. To address this, we propose a Multi-Scale Wavelet Mixing (MWMixer) model, which incorporates a Bidirectional Frequency-Bands Mixing strategy to recover lost temporal details across scales, enabling complementary cross-scale information interactions. Then, a Dynamic Scale-Adaptive Fusion module learns time-varying weights for each scale to fuse multi-scale forecasts into the final prediction, enhancing the flexibility of multi-scale aggregation. In addition, a cross-scale consistency loss aligns each coarse-scale prediction with the interval-averaged fine-scale outputs, while a multi-scale supervision loss enforces prediction accuracy at each scale, promoting consistent learning across scales. Extensive experiments on seven real-world datasets demonstrate that MWMixer achieves competitive performance in long-term forecasting.

Quantum computing's environmental footprint remains poorly understood relative to classical infrastructure, and as quantum computing moves toward cloud delivery, Quantum Cloud Service (QCS) providers lack actionable guidance beyond platform-level carbon-accounting frameworks. Objective: This study extends the carbon-aware quantum computing (CQC) framework from a platform-level to a service-level model that translates empirical life cycle assessment (LCA) findings of a superconducting quantum computer into guidance for QCS providers. Method: We modeled the CQC framework via service-level embodied-carbon allocation, load-independent and load-proportional operational decomposition, and a workload-resolved application offset on the basis of results acquired through a cradle-to-grave LCA of a superconducting quantum platform. Results: The five-year footprint is 583 t CO2e (GKP) and 10,570 t (surface-code), dominated by embodied carbon (77.3-85.2%), with operational-embodied parity not reached until 17.0-28.7 years versus 2.7 years for classical comparators. This reorders provider levers: utilisation yields the largest gain (19.7x), followed by service life extension (59.9%) and electricity supply (6.5x), while operational efficiency and renewable procurement offer limited leverage. Conclusion: Superconducting quantum computers are structurally embodied-carbon-dominated, inverting classical sustainability intuition and motivating direct power measurement and cross-architecture validation as quantum infrastructure scales.

Traditionally, the reliability of news publishers is assessed by expert organisations that evaluate editorial practices, transparency and factual standards at source. When this process is translated into a computational approach, the problem is often formulated at the level of individual articles, with models being trained on a set of pre-labelled articles and their performance being evaluated in a test phase. In this work, we investigate news source reliability inference as a source-level prediction problem. We propose a two-stage framework in which transformer-based language models first estimate the reliability of individual articles and subsequently aggregate article-level predictions to infer the reliability of previously unseen publishers. To approximate realistic deployment conditions, we enforce a strict publisher-disjoint evaluation protocol, ensuring that no publisher appears in both training and test sets. Experiments on 19,476 political news articles from 439 English-language publishers labeled with NewsGuard reliability ratings show that aggregation substantially improves robustness and performance, increasing accuracy from approximately 0.60 at the article level to 0.69 at the publisher level. Finally, we analyze how prediction errors vary across political orientations, revealing statistically significant associations between political leaning and misclassification patterns. Overall, our findings show that publisher reliability can be inferred from aggregated textual signals alone, supporting scalable and content-based approaches to automated news source assessment.

The Platonic Representation Hypothesis (PRH) claims that independently trained models converge on a shared statistical model of reality, yet recent work finds only weak pointwise similarity between models. In this paper, we show that what models share is not the location of samples in representation space, but the directions (displacement vectors) between them. Under a single orthogonal alignment--rotation and reflection only--these displacement vectors are substantially preserved across 44 independently trained vision and language encoders spanning modalities and asymmetric capability pairs, consistent with the PRH evidence. The samples' absolute positions are not, consistent with recent counter-evidence. Both arise from a single decomposition: representations split into a shared semantic component that is linearly aligned across models, and a private capability component that is not. We trace this geometry to concept-level structure: within a model, parent concepts are orthogonal to their child variation vectors; across models, concept displacements are parallel. Our theory falsifiably predicts (and experiments confirm) that fine-tuning preserves pointwise similarity but collapses displacement, and that relational distillation does the opposite. A major implication is that, because semantics align linearly but capabilities do not, capabilities can be imported from one model to another using a single cached forward pass through the source. We call this Shadow Casting. As a proof of concept, our SHADOWCLIP instantiation outperforms strong fine-tuned baselines at orders of magnitude less compute. A cache can be released alongside open model weights, letting one model's capabilities be downloaded and imported into any number of other models without fine-tuning.

Modern AI model training imposes unprecedented computational demands, making it a key contributor to datacenter energy consumption. Yet a significant fraction of the energy consumed during training does not translate to useful computation due to bottlenecks throughout the training pipeline. We present PAFT, a phase-aware, dynamically adaptable GPU frequency tuning system that reduces energy consumption of training workloads with minimal performance overhead. The key insight behind PAFT is that bottlenecks represent an energy optimization opportunity, rather than purely a performance problem: when GPUs are bound to stall, PAFT opportunistically reduces their clock frequencies to match the pace of bottlenecked devices, saving energy without impacting execution time. PAFT achieves this by continuously monitoring pipeline behavior and applying fine-grained frequency adjustments, adapting to workload and system changes. Experiments conducted on twelve widely used models show that PAFT consistently outperforms all baselines, achieving energy savings of up to 46% with an average overhead of 4%.

Sparse attention reduces the quadratic cost of global self-attention while retaining strong empirical performance, but how its restricted interactions shape the evolution of token representations remains theoretically underexplored. Modeling tokens as particles on the unit sphere, we introduce opinion leader dynamics, a framework that identifies two mechanisms through which token groups converge internally while maintaining distinct limiting directions. In the explicit model, fixed representatives induce a potential that attracts tokens toward distinct local maxima. In the implicit model, disconnected interaction groups evolve toward separate consensus directions. We formulate both models as reverse Wasserstein gradient flows and establish exponential convergence under suitable conditions. We further connect these theoretical predictions to token evolution in frontier sparse-attention LLMs that motivate our framework. Across four benchmarks, Kimi-K3, MiniMax-M3, and DeepSeek-V4-Flash consistently exhibit clearer cluster separation and higher clustering scores than the dense-attention model GLM-4.7-Flash in projected token representations. These observations support the relevance of the predicted multiple-group structure to trained frontier LLMs, while finite-particle simulations illustrate the theoretical convergence behavior. Together, our results connect restricted token interactions to distinct group-level attractors, providing a dynamical account of how sparse attention can support alignment within groups while preserving separation between them.

Evaluation benchmarks for Indian language automatic speech recognition (ASR) suffer from two systematic biases: optimistic scores from clean, controlled audio conditions, and pessimistic scores from overly rigid transcription standards that penalize valid linguistic variations. We introduce Vimarsha, a 100-hour benchmark spanning all 22 scheduled Indian languages, designed to address both distortions. Vimarsha combines demographically diverse on-field recordings with carefully mined in-the-wild audio selected for acoustic difficulty, alongside a lattice of variations framework that encodes multiple valid transcriptions per utterance. Evaluations of 10 state-of-the-art ASR models reveal substantial shifts in model rankings under realistic conditions, geographic and demographic performance disparities, and systematic failure modes across speaking rates and acoustic environments.

This paper describes the SKstars submission to SHROOM-Visions 2026, a shared task on fine-grained hallucination detection in large vision-language model outputs. The task requires systems to identify hallucinated character spans, assign hallucination categories, and provide confidence estimates for their predictions. Our approach combines zero-shot predictions from Qwen2.5-VL-72B-Instruct with those of a LoRA-adapted Qwen2.5-VL-7B-Instruct model. The outputs of the two models are integrated through a lightweight ensemble procedure, followed by span refinement and confidence adjustment. We evaluate the main system components on a small internal development subset and report the performance of the submitted system on the official English test set. SKstars achieved a Cor+Lbl score of 0.2902, ranking 15th among 29 teams, and obtained Cor and IoU scores of 0.3642 and 0.3151, respectively, ranking 18th on both metrics. The results show that combining a large zero-shot model with a smaller adapted model provides a practical framework for multilingual and fine-grained hallucination localization, while also highlighting the difficulty of transferring development-set improvements to hidden test data. Code and predictions: https://github.com/aliathar1401/SK-Stars-shroom-visions-2026

Speculative decoding losslessly accelerates large language model inference by having a lightweight draft model predict future tokens for verification by the target model. Recent block diffusion drafters further reduce drafting latency by predicting multiple tokens in parallel. However, existing block drafters project target hidden states at every input position into a separate drafter-side KV cache, incurring per-request memory and KV-write overhead that grow with concurrency; directly reusing target KVs in place removes this cache but fails to sustain draft quality throughout the block. We propose a hybrid target-context injection method that complements direct target KV reuse with target hidden states only at the last input position, requiring no separate drafter-side KV cache. Building on this design, we propose H-Spec, a hybrid Mamba-attention parallel drafter that consumes the two target-context sources through complementary modules. Mamba modules are initialized with projected last-token target hidden states, while attention modules reuse target KVs in place. Despite its recurrent formulation, Mamba's parallel scan allows H-Spec to preserve block-parallel drafting. Across three target models and diverse tasks, H-Spec improves over the best baseline by 5.0--13.3% in mean accepted length and 5.3--12.6% in batch-size-1 inter-token latency speedup. Under concurrent serving, H-Spec consistently achieves higher throughput while maintaining lower KV cache utilization than baselines across evaluated concurrency levels.

Looped Language Models (LoopLMs) perform "latent reasoning" by recursively refining internal latent representations with shared weights, offering a more effective alternative to explicit verbal reasoning. Despite their effectiveness, we find that LoopLMs remain prone to loop instability: unstable refinement across iterations can produce localized uncertain "hard" tokens associated with reasoning errors. To address this, we propose LoopCD, loop-wise contrastive decoding that enhances the reasoning performance of LoopLMs by intervening on these tokens at inference time. Specifically, we exploit the internal dynamics of LoopLMs and contrast the logits from earlier iterations with logits from the last refined iteration to form the final sampling distribution. We find that this strategy is highly efficient, introducing only negligible inference overhead and requiring no additional training, while effectively improving reasoning performance by naturally refining reasoning-critical hard tokens. Extensive experiments show that our method improves the performance of recent representative LoopLMs across various reasoning tasks.

Evaluation scores used around LLM systems -- including reward models, rerankers, and LLM judges -- can track surface form instead of the quality they claim to measure. When presented with a terse correct solution and a commented buggy solution for the same MBPP problem, a public preference reward model selects the correct one no better than a coin flip (0.507). Subtracting the predictable surface component from such scores is increasingly common, but removal alone does not yield a more valid measurement: the removed component may carry construct-relevant signal, and residualization cannot tell which is which. Under designed interventions -- unit-test labels with comment-only edits -- residualization attenuates the reward model's format effects by about 0.12 on both correct and buggy code, while the correct-versus-buggy margins move by less than 0.01. In observational NLI and QA settings, we freeze a held-out replication before scoring and re-evaluate it using labels from disjoint annotators; this supports only a narrower conclusion: better agreement with the construct labels on a pre-declared slice where a surface-only predictor errs, not a repaired score. Full-population agreement falls in every observational setting with a reported positive slice gain, and within-question ranking falls in every such QA setting. When construct and surface features are entangled, residualization can decorrelate a score while degrading construct alignment, and, in a controlled model, configurations just as damaging to construct alignment pass every pre-adjustment check, so no committed gate is a guarantee. We assemble these distinctions into a reporting protocol whose outcomes, refusal included, state what an adjusted score may be claimed to show: an audit-time diagnostic reported beside the construct-alignment cost it incurs, never a replacement for the raw score.

Large language models have achieved remarkable progress on Text-to-SQL through reasoning-enhanced fine-tuning, yet existing approaches predominantly rely on massive instruction corpora under the assumption that scale drives performance. We challenge this paradigm by investigating a fundamental question: what is the minimal data requirement for effective Text-to-SQL instruction tuning? We propose LIMIT(Less Is More for Instruction Tuning in Text-to-SQL), a data-centric framework that demonstrates strong database reasoning can emerge from an extremely compact training set when examples are strategically selected. LIMIT operates through four stages: difficulty-aware filtering that identifies samples within the model's learning frontier, chain-of-thought synthesis with consistency-based selection, multi-dimensional quality scoring via LLM-as-judge, and genetic algorithm optimization that jointly maximizes schema coverage and sample quality. On the BIRD and Spider benchmark, LIMIT selects only 796 and 863 samples while achieving 100% table coverage, enabling Qwen3-8B to reach 69.1% and 88.9% execution accuracy.This result surpasses methods trained on 20 times more data and establishes a new state-of-the-art among open-source approaches. Our findings suggest that careful data curation, rather than scale, is the key to efficient Text-to-SQL learning.

Legal information in Bangladesh is inaccessible to most citizens. Statutory text is English-only, trained lawyers are concentrated in urban centres, and cloud-dependent AI fails where mobile connectivity is unreliable, a setting in which hallucinated legal text causes direct harm. The system addresses statutory interpretation only; queries that require judicial precedent or case-law reasoning fall outside its scope. We target the statutory access gap by compressing a 9-billion-parameter Gemma-2 teacher into a 2-billion-parameter student through two-phase progressive knowledge distillation. Phase 1 performs supervised fine-tuning on 9,429 quality-gated legal question-answer pairs (65% acceptance from 14,514 generated queries); Phase 2 minimises sparse Kullback-Leibler divergence against the teacher's top-50 per-token logits at temperature tau = 4.0, implemented via QLoRA (4-bit NF4, rank-32 LoRA adapters). Prior legal language models target general legal English; this system specialises in Bangladeshi statutory law. Every response is grounded through hybrid retrieval combining dense semantic search (60%) and BM25 (40%) across 36,029 statutory passages from the Bangladesh Constitution and national legislation. On a 50-query English benchmark, the distilled model reaches ROUGE-L 0.4715 and BERTScore F1 0.5679, a 103% ROUGE-L and 143% BERTScore gain over the retrieval-augmented undistilled baseline (ROUGE-L 0.2323, BERTScore 0.2340). The adapter quantises to 1.6 GB (GGUF Q4_K_M) and runs at 4-8 tokens per second on a Pixel 6 with no network access. Cross-lingual evaluation on 50 Bangla queries yields ROUGE-L 0.4083 and BERTScore 0.8133, showing effective retrieval from Bangla input against an English-only corpus. In a single-evaluator pilot, a practising lawyer rated 50 responses at a weighted mean of 4.16/5 (90% rated 4 or 5), supporting utility beyond text-overlap metrics.

Long-horizon language agents receive sparse terminal feedback, while intermediate rubrics provide structured but potentially misspecified assessments of progress. In resettable training environments, counterfactual continuation rollouts can measure local credit, but exhaustive replay is costly. We propose Credo, a framework that couples evolving semantic rubrics with selective, execution-based credit correction. A frozen judge maps visible transitions to rubric features, and a credit head predicts the change in expected terminal reward associated with the realized transition. Independently sampled two-sided replays correct prediction residuals using their recorded inclusion probabilities. We derive conditional unbiasedness and a variance decomposition that connects two design choices: which rubric features to retain, and where to allocate a fixed expected replay budget. The resulting criterion weights prediction errors by policy-score sensitivity and missing replay coverage; its allocation rule additionally accounts for continuation cost. We also describe a practical mixture with terminal leave-one-out advantages and distinguish its clipped, token-normalized PPO implementation from the ideal policy-gradient estimator. This preliminary report provides the method, proofs, an exact finite-model audit, and a controlled evaluation protocol. It makes no claim of empirical superiority on language-agent benchmarks.

Synchrotron data reduction, detector calibration followed by azimuthal integration of terabyte-scale diffraction series, is a multi-step, expert-bound bottleneck that increasingly limits the science rate of user facilities. LLM agents promise to collapse it, but driving a real pipeline with a stochastic model creates a failure mode chat benchmarks cannot see: an agent can report a calibration that was never computed. Correctness here is a property of what executed, not of the transcript. We present APEXA, a deployed multi-agent framework (61 tools over heterogeneous compute, run as a single reasoning loop) automating calibration and integration from natural language at a major light source. We make three contributions. First, execution-integrity enforcement: a deterministic tool-layer guard that refuses to surface any result not backed by an executed tool call, with a parser tolerant of cross-model tool-call format drift: in deployment, a frontier model fabricated a complete calibration-comparison report for commands that never ran, which the guard converts to an explicit non-result; the same code gates an optional motor-control surface at 0/200 adversarial violations against a simulated IOC, versus 15/200 for an equivalent safety prompt. Second, we release APEXA-Bench, an evaluation harness of 58 facility tasks (50 base plus an 8-task cross-detector slice) organized by a four-class physical-consequence taxonomy, the first benchmark axis we know of separating a wasted compute cycle from a damaged instrument; its cross-detector grading against NIST-traceable lattice constants surfaced two latent pipeline bugs. Large-scale agent scoring is left to a full-length study. Third, we validate APEXA on real beamline data: from one natural-language prompt it recovers detector geometry and integrates a full attenuation/exposure sweep. We release the framework, harness and traces.

As LLM agents increasingly interact with external tools through standardized protocols such as MCP, tool-interface design becomes a critical yet underexplored factor. How funψtionality is decomposed into tools affects whether an agent can select the right tool and construct valid arguments. This choice is especially consequential at the edge, where resource constraints limit which models can run locally and scaling up is often not an option. We present MCP-GRANITE, an open-source extensible benchmark framework that treats tool-interface granularity as a controlled variable for MCP-based agents, evaluated under edge and IoT scenarios. It comprises 81 multi-step scenarios across 9 domains, instantiated at 4 granularity levels from fine-grained primitive tools to a single tool. We evaluate 9 locally deployed models (268M-20.9B parameters) across 8,748 trials using task completion, tool selection F1, argument accuracy, latency, and resource-usage metrics. Results show that a 4-tool interface offers the best trade-off, improving task completion by 16.4% over fine-grained primitives and 33.6% over a single monolithic tool, while nearly doubling argument accuracy. Model size is only weakly correlated with task completion and strongly with latency, while its association with argument accuracy is less robust, and a 3.2B model at the optimal granularity outperforms a 20.9B model at a mismatched one. These findings identify tool-interface granularity as a key design parameter for MCP-based agents.

Most existing time series forecasting methods rely solely on numerical observations, overlooking rich contextual information from auxiliary texts. Recent multimodal approaches attempt to incorporate textual signals, but they often treat text as static features or use large language models as forecasting backbones, limiting their ability to capture temporal dynamics and increasing computational cost. To address these challenges, we propose TAC-Time, a unified framework that transforms textual information into additional temporal channels. By modeling text features jointly with numerical sequences in a shared temporal backbone, TAC-Time preserves temporal continuity and periodic structures while remaining efficient and scalable. This formulation also enables systematic interpretability analyses. We show strong cross-modal dependencies through attention and frequency-domain analyses, and identify predictive textual signals whose correlation-aware alignment yields partial forecasting improvements. Extensive experiments on real-world multimodal benchmarks demonstrate that TAC-Time outperforms prior methods.

Visual search on large e-commerce catalogs must serve both "similarity" queries that ask for items resembling an uploaded image and "modifier" queries that comprise an image and text describing a desired modification (e.g. a color change or style swap). The latter is the setting known as composed image retrieval (CIR). Existing CIR methods, however, treat relevance as binary and train on triplets with a single positive target - a poor fit for real catalogs where many candidates partially satisfy a user query and ranking across that partial-match spectrum drives the customer experience. We propose a methodology for training CIR retrievers on graded relevance, consisting of: (i) a VLM to curate training data, generating both queries (object detection + modifier synthesis) and 4-level relevance labels without manual annotation, (ii) an iterative relevance-feedback loop that expands the training set by mining hard negatives from the in-training retriever, and (iii) a hierarchy-aware angular objective to train the retriever directly on the graded labels rather than collapsing them to a binary split. We call this methodology GradCIR and instantiate it on a PaliGemma2 bi-encoder trained on 3.5M graded pairs curated from raw Walmart catalog data. A controlled graded-vs-binary ablation isolates the supervision granularity and shows lift of 4.9%-5.9% in NDCG@10. The same recipe applied to other multimodal encoders lifts early-fusion backbones by up to 8.5% NDCG@10. On the public FashionIQ benchmark, GradCIR (applied to PaliGemma2) reaches 0.6703 average recall when fine-tuned, slightly ahead of the strongest peer-reviewed supervised baseline we compare against, and matching or exceeding all published CLIP-L-class zero-shot CIR methods. The system is deployed in production at Walmart, where it's serving live visual-search user traffic.

Four-dimensional (4D) Radar has emerged as a key sensor for environmental perception, providing range, azimuth, elevation, and Doppler measurements while remaining robust to illumination changes and adverse weather conditions. However, conventional Radar preprocessing methods, such as constant false alarm rate (CFAR) detection, select measurements primarily based on signal-level criteria and may therefore discard information valuable for downstream perception during point cloud generation. In addition, existing 4D Radar perception pipelines typically optimize Radar data processing and downstream perception independently, preventing task objectives from directly guiding the preprocessing stage. To address these limitations, we propose a Scene- and Task-Aware Radar (STAR) Preprocessor together with an end-to-end training framework. The STAR Preprocessor incorporates scene context and downstream task objectives to generate task-relevant Radar points, enabling the Radar representation to be optimized directly for perception. On the K-Radar benchmark, the proposed method achieves 74.3 AP, outperforming the previous state of the art by 5.6 AP points. Furthermore, applying the task-relevant points generated by STAR to various existing 3D detectors improves detection performance in most evaluation settings and yields an overall positive average gain over point clouds produced by conventional preprocessing.

Speculative decoding accelerates large language model (LLM) inference by using a lightweight draft model to generate multiple candidate tokens that are verified by the target model in a single forward pass. Its speedup is largely determined by the acceptance length, yet existing draft-model training methods mainly optimize cross-entropy or Kullback-Leibler (KL) divergence as proxies. These objectives encourage distribution matching but do not directly optimize acceptance length, and the acceptance mechanism also differs between greedy and sampling-based decoding. In this work, we propose acceptance-length-aware training losses that directly optimize the expected number of accepted tokens within a speculative window. For greedy verification, we derive an expected accepted length (EAL) loss that explicitly maximizes expected acceptance length. For sampling-based decoding, we introduce a window total variation (WTV) loss that optimizes the overlap between temperature-scaled draft and target distributions while accounting for sequential acceptance dependencies. Both objectives can be further combined with a group-relative reinforcement learning stage (GRPO) using simulated acceptance length as the reward. Experiments across different target and draft models, tasks, and decoding settings show that our losses consistently improve acceptance length over KL-based training. WTV provides particularly strong gains under sampling-based decoding, while EAL better matches greedy verification. These results show that directly optimizing the acceptance objective, with losses tailored to the decoding mode, is more effective than conventional distribution-matching objectives.

Language model agents are increasingly used to simulate social interaction, and the resulting transcripts read as though the agents understand one another. We ask whether that appearance rests on a model of the partner's mind or on the surface record of what the partner said. We build a social simulation in which both questions have exact answers: 40 multi-issue negotiations whose hidden preference weights and whose full Pareto frontier are known by construction. Two model families negotiate across 160 dyads, every transcript is frozen before any measurement, and 2880 counterfactual probes then hold the evidence byte identical while moving one factor at a time: the reader's own stake, the partner's tone, an identity label, and the order of recursion. The agents are socially fluent and economically poor. They reach agreement in 96.2% of dyads with 0 protocol failures, yet only 0.7% of deals land on the Pareto frontier, they leave 20.5% of the available joint value unclaimed, and they miss the one issue on which their interests are perfectly aligned in 76.6% of deals; on the frontier and on that aligned issue, a package drawn at random from the set both sides would accept does as well. The probes locate the failure. Swapping only the reader's own payoff sheet, while the partner's words and offers stay identical, moves the inferred top priority by 15.0 percentage points, which is egocentric projection rather than inference, while a tone rewrite moves it by 5.3 percentage points and an identity label by 0.0. Most tellingly, an agent predicts what its partner believes about it 72.5% of the time while that partner's belief is itself correct only 51.2% of the time: the agents track the conversation far better than they track the mind behind it.

Group-relative reinforcement learning compares rollouts of the same prompt, but independent environment noise can obscure these comparisons. We study paired rollouts, which share an event-keyed noise schedule within each group while preserving each rollout's marginal distribution. Pairing removes the between-schedule component of reward-contrast variance, but need not reduce gradient variance. For one-sided grader noise, we derive an exact condition for reduction and give a counterexample in which reward contrasts improve while gradient variance increases. A controlled study trains a 2B tool-use agent under tool faults and grader flips, with three seeds per design. The protocol was registered with a disclosed, previously completed pilot. Under tool faults, pairing improves final noisy-test success by +5.1 percentage points on average, with all three seed differences positive, but misses the registered learning-curve criterion. The criterion is also missed under grader flips: the validation-AUC difference is +0.003 (95% interval [-0.029, +0.033]). A gradient probe on eight distinct checkpoints from two fault-trained trajectories finds lower mean-centered covariance traces under both noise types: 21 to 30% for grader flips and 40 to 63% for tool faults. These finite-sample measurements support the variance mechanism without establishing a general learning-speed benefit. The results distinguish improving reward comparisons, reducing estimator variance, and improving learning.

On-policy distillation (OPD) pays twice for each fresh batch: the student generates trajectories and a stronger teacher scores them. Existing methods improve which trajectories are scored and how the teacher signal is constructed, but usually consume it with one actor update. We introduce CLOOPD, a closed-loop framework separating teacher-signal acquisition from student-side realization. CLOOPD selects an adaptive $α$ waypoint inside a KL envelope, freezes the scored batch and its advantages, re-forwards the student after each actor pass, measures realization, and allocates actor work under a separate token budget. The framework includes deterministic two- and three-pass policies, token-priced CLOOPD-TPMR, and a budget-matched control. Across six 300-step runs on an 8-H20 node, every CLOOPD policy improves the one-pass TOP-D anchor at comparable teacher-token scale: macro accuracy rises from 15.41 to 17.78 with CLOOPD-Fixed2 and 19.36 with CLOOPD-Fixed3. At step 100, CLOOPD-Fixed3 reaches 15.35, nearly matching TOP-D at step 300 while using 67.2% fewer teacher-scored tokens and 28.0% fewer GPU-hours. Earlier 8-A100 ablations show adaptive $α$ eliminates observed trust-envelope violations; a third pass adds headroom. These results position CLOOPD as a framework for budgeting how fully students learn from teacher-scored tokens.

Generative models have recently demonstrated considerable promise in speech super-resolution (SSR). Nevertheless, the majority of existing work has concentrated on standard or versatile SSR configurations, leaving the extreme setting with severely limited spectral inputs largely unexplored. In this regime, current approaches exhibit marked performance degradation, underscoring the need for dedicated solutions. To bridge this gap, we introduce P2Flow, a phoneme-aware progressive flow matching (FM) framework designed for extreme SSR with three main strategies. First, our model leverages phonetic information to reconstruct missing spectral components. Furthermore, it employs a progressive architectural design that hierarchically restores distinct frequency regions. Finally, we incorporate post-training of the vocoder to enhance overall waveform fidelity. Extensive experiments are conducted on the TIMIT and VCTK datasets under both 1 kHz to 16 kHz and 2 kHz to 16 kHz settings, demonstrating that P2Flow yields state-of-the-art results across multiple evaluation metrics.

Traditional data systems face profound limitations in the AI era, relying on human-crafted pipelines, lacking semantic understanding of heterogeneous data, and operating through rigid, reactive processing. To address these challenges, we propose a new paradigm called the Data Agent, designed to manage, process, and analyze data with minimal human intervention. Data agents autonomously execute a wide range of data-related tasks, transforming traditional data systems by shifting from manual design to autonomous orchestration, from literal manipulation to semantic interpretation, and from reactive to proactive processing. Our Data Agent system includes six components: semantic data organization, semantic operators, agentic pipeline orchestration and optimization, feedback-driven refinement, memory management, and proactive adaptation. Building on this foundation, we also develop two specialized agents: the data analytics agent and the data science agent. Experiments on real benchmarks demonstrate significant performance gains of our data agent over state-of-the-art methods. We identify open challenges to guide future research in building fully autonomous data systems.

LLM agents can change their own future behavior, raising a basic control question of which self-generated changes should be allowed to persist. We formulate this as admission control for self-modification. The agent may propose changes to its operating instructions, while an external runtime gate controls persistence. We implement this principle as a model-agnostic self-healing harness that runs a Detect, Notice, Heal, Validate loop around an otherwise unmodified agent. The agent authors candidate behavioral rules in an external workspace, where they receive provisional execution authority during evaluation and acquire persistent cross-episode authority only after measured improvement on the triggering failure without regression beyond a fixed margin on protected cases. Replay provides matched evidence when available, forward trials provide a weaker fallback, and a corpus-level guard re-tests the accumulated active rule set. Across 16 matched Baseline and Harness runs spanning AppWorld, Terminal-Bench, and $τ^2$-Bench, the gate rejected 383 replay-decided proposals. Of these, 211 (55%) improved their triggering failure while degrading a case that previously worked. This shows that locally beneficial self-modifications can introduce collateral regressions often enough to materially affect gate decisions, providing direct empirical motivation for external admission control. Task-completion score is higher under the Harness in all 16 pairs, with two paired bootstrap intervals excluding zero, while repeated-trial reliability is higher in 12 pairs, tied in 4, and lower in none. Because adaptation modifies the policy-inducing context while leaving model weights fixed, admitted changes remain inspectable, reversible, and compatible with closed-weight models.

Judge-specific sensitivity is useful for aggregating pairwise LLM evaluations, but its interpretation depends on which systematic presentation effects the ranking model includes. We introduce OSCAR, an order-aware framework for scoring and calibrating AI rankings, and study position as one such effect. In released judgments from 18 evaluators, the all-response A-minus-B score difference ranges from $-63.11$ to $98.31$ percentage points. Matching question text, response texts, candidate identities, and judge within the released table gives an overall difference of $24.22$ points (95% interval $[22.90,25.54]$), conditional on the released text mapping. A controlled calculation isolates the potential consequence: with true sensitivity fixed at one, omitting a position intercept of four reduces the population-optimal slope to $0.0771$. We extend sensitivity-based ranking with judge-specific position, length, and family terms, characterize local omission-induced displacement and an identification failure, and propagate prompt-cluster uncertainty to adjusted comparisons. Across four released datasets, position provides the largest stand-alone predictive improvement. Refitting bootstrap comparisons show more selective gains from the full model over position-only adjustment. In dependent binary simulations, adjusting both the mean and covariance yields 94.4--95.2% coverage; correcting either alone is insufficient. At $N=10{,}000$, OSCAR reduces mean neutral-target RMSE from $0.1158$ under the sensitivity-only model to $0.0237$.

Atomic activity understanding aims to recognize and localize structured traffic behaviors that jointly encode motion patterns and their grounding in road topology. Unlike conventional action recognition, atomic activities are multi-agent, multi-label, and topology-aware: multiple activities co-occur while many agents remain inactive. We introduce Action-Slot, a structured action-centric representation learning framework. Slot attention is widely used for object-centric decomposition, but its permutation-invariant design and object-level inductive bias are misaligned with atomic activity semantics. We reformulate slot learning as structured activity decomposition through three designs: (1) category-aligned action slots that anchor slots to predefined activity categories, (2) parallel spatio-temporal slot updating for holistic video-level reasoning, and (3) background and negative-slot regularization that enforces competition between foreground activities and irrelevant regions. Together these establish an activity-centric inductive bias that disentangles concurrent and asynchronous activities directly from raw video. Beyond recognition, the learned representations encode transferable spatio-temporal grounding signals. We further propose an attention-difference-based pseudo mask selection framework that suppresses false positives by measuring attention changes before and after candidate region removal, enabling weakly supervised localization without dense annotations. To support systematic evaluation, we introduce TACO, a balanced synthetic dataset with full atomic activity coverage and pixel-level annotations. Experiments on OATS, TACO, and annotated nuScenes show superior recognition, strong sim-to-real transfer, and state-of-the-art weakly supervised localization.

Black-box machine learning models increasingly deliver strong predictions, but extracting useful information from them, such as a set of important features, remains challenging. Existing model-agnostic methods primarily estimate feature importance or conduct inference on it rather than directly selecting features, whereas many feature selection methods are model-specific or rely on the model-X assumption. We introduce LOCO-guided Adaptive Minipatch Sampling (LAMPS), a model-agnostic ensemble framework that uses any black-box regression algorithm as its base learner to select features important for predicting the response. The base learner need only produce predictions and need not perform feature selection itself. LAMPS operates within a minipatch ensemble framework that subsamples both observations and features, allowing leave-one-covariate-out (LOCO) feature importance scores to be easily computed. It adaptively concentrates minipatch sampling on features with high LOCO scores while maintaining exploration. The resulting sampling probabilities rapidly separate signal from noise features after a few iterations, enabling selection through simple thresholding. We establish that LAMPS achieves exact feature selection in high-dimensional settings, provided that the base predictive models are sufficiently well trained on average. Extensive experiments on synthetic and real data show that LAMPS outperforms state-of-the-art feature selection methods, with particularly strong performance in the presence of correlated features.

Active perception and manipulation are crucial for robots to interact with complex scenes. Existing benchmarks struggle to evaluate how robots effectively acquire and maintain information in memory in an active manner. To this end, we introduce ActiveArena-Sim, an active-perception simulator with controllable viewpoints and large-scale workspaces as the foundation. Built on this, we propose ActiveArena-Bench, which comprises 35 tasks across 5 fine-grained categories, covering visual exploration and interactive information acquisition. Each task is difficult to solve from passive observations alone, requiring multi-round evidence acquisition and memory-based reasoning. The benchmark provides rich memory annotations, standardized training data, and ID/OOD protocols featuring disjoint scenes, unseen distractor configurations, and novel backgrounds. Moreover, we present ActiveArena-VLA, a modular suite of 13 vision-language-action configurations for controlled studies of memory writing, memory capacity, proprioceptive state, subtask supervision, and high-level planning in active perception. Benchmark results reveal a substantial ID-OOD gap: uniform memory sampling, increased memory capacity under reliable write policies, proprioceptive inputs, and subtask supervision improve OOD generalization, while planner-guided memory management and decision-making achieve performance close to the best-performing configuration using only sparse memory. ActiveArena thus provides a unified testbed to develop and diagnose models for active perception and manipulation.

Retrieval-augmented generation (RAG) is hard to monitor in production: exhaustive relevance labels do not exist for non-stationary multi-million-passage corpora that re-index in real time. As a result, retrieval quality is generally understudied and often deprioritised in favour of generation-oriented metrics. In this work, we propose auditing retrieval coverage by probing for evidence of missing documents rather than enumerating every relevant one. Our method Re:CAP (REtrieval Coverage Audit by iterative Probing) is a reference-free audit loop applied to a deployed RAG pipeline's initial answer and retrieved context: it identifies the topics already covered, generates probing questions for plausibly missing topics, retrieves candidate documents, and applies an LLM-as-judge to retain only those that introduce previously-unretrieved information. On four public benchmarks, Re:CAP recovers 9-29% of gold labels that flat BM25 top-500 cannot reach, rising to 48% on TREC-COVID. On MuSiQue Re:CAP beats flat hybrid top-500 by +12.9 pp on recall at less than half the document budget. An ensemble BM25, dense, and hybrid baseline (top-500 each) still leaves out 21.2% of gold docs on TREC-COVID that Re:CAP recovers; human annotators judge that 78.9% of those structurally distinct documents add new information to the baseline answer (Fleiss $κ$ = 0.79, n = 123), and 73.9% on live production traffic (n = 180). End-to-end recall is reproducible to within $\pm$1% across three independent runs, making Re:CAP a stable instrument for periodic retrieval audits.

Identifying linear time-invariant (LTI) dynamical systems is challenging when trajectories are short, noisy, or high-dimensional. Traditional system identification typically treats each system independently and cannot exploit shared structure across related systems. We propose PBML-LTI, a PAC-Bayesian meta-learning framework for few-shot LTI system identification that learns a transferable prior over task-specific dynamics while preserving task heterogeneity. Each task corresponds to an unknown LTI system, and the meta-learner uses training trajectories to learn a data-dependent prior over transition matrices. For a new system with limited data, PBML-LTI performs Bayesian adaptation under this prior to obtain a task-specific posterior, providing accurate estimates and principled uncertainty quantification. A key challenge is temporal dependence, since LTI trajectories violate the i.i.d. assumptions underlying most PAC-Bayes meta-learning analyses. We address this with a martingale PAC-Bayes analysis for dependent trajectory losses and derive a support-query predictive-risk bound that motivates a fit-KL meta-training objective. The bound clarifies the roles of empirical fit, posterior complexity, and prior quality in few-shot adaptation under sequential dependence. We further derive corollaries for transition-matrix recovery and multi-step trajectory prediction, connecting uncertainty-aware meta-identification with finite-sample guarantees for dependent dynamical data.

Tool-calling LLM agents are increasingly deployed in enterprise applications. However, effective evaluation and optimization require high-quality, diverse task datasets that are often difficult to obtain due to privacy and other constraints. Existing synthetic task generation methods often produce generic tasks that ignore an agent's underlying state or database and fail to reflect real-world usage diversity. We propose EdgeGen, a synthetic task generation framework that extracts compliance rules from an agent's specification and uses them to generate database-grounded edge-case tasks designed to violate these rules. When combined with existing synthetic data generation techniques, EdgeGen enables agent improvement through finetuning and harness optimization. The resulting pipeline forms a fully automated closed-loop system that requires no human annotation. Finetuning on data generated by EdgeGen yields a consistent mean progress improvement of 2 percent to 42 percent on tau2bench airline domain, while other baseline methods show degradation for some models. On the other hand, for harness optimization, our method shows a mean progress improvement of 10 percent and 30 percent over the human-curated and base harnesses, respectively, for the Gemma-4-e4b model.

Causal Bayesian Optimization (CBO) combines causal inference with Bayesian optimization to enable sample-efficient intervention selection in systems with causal structure. This survey provides a systematic review of CBO through a unified BO-loop perspective, showing how causal assumptions shape intervention search spaces, surrogate models, acquisition functions, and decision policies. We organize existing methods by graph and system-knowledge assumptions, environment, intervention representation, surrogate architecture, and decision rule, and connect CBO to causal bandits, Bayesian experimental design, safe optimization, policy search, and causal abstraction. We also introduce a reproducibility-oriented benchmark spanning hard- and soft-intervention settings, with standardized GAP and a new trajectory-aware Path-Aware GAP (PA-GAP), evaluating seven CBO methods and a non-causal BO baseline across thirteen datasets, three budgets, and two metrics. Results show that no method dominates uniformly: rankings depend on dataset, budget, metric, and how causal information is used, while strong non-causal baselines remain competitive in several settings. Controlled graph-misspecification and omitted-variable stress tests further show that rankings can change substantially when learner-side causal information is perturbed. We conclude by identifying key open challenges, including robustness to causal-assumption violations, scalable unknown-graph optimization, mixed intervention types, realistic cost models, stronger theoretical guarantees, and integration with modern representation learning and causal abstractions.

Test-time adaptation (TTA) addresses distribution shift using only unlabeled test data. Existing methods typically adapt pretrained models by updating their parameters, limiting both what is adapted and where adaptation can occur within the network. We instead keep the pretrained network frozen and steer its intermediate representations. We introduce SPeaR (Steering Primitive for Realigning Representations), which inserts lightweight learnable modules at stage boundaries and optimizes them directly from the test stream, requiring neither source data nor supervised warm-up. Each primitive is optimized using a gated objective that reduces uncertainty only when adaptation is beneficial, along with a diversity regularizer to prevent collapse, and a multi-depth anchor to stabilize adaptation. We show that steering early representations is the most effective strategy, and that the same primitive transfers across convolutional and Transformer architectures. Across CIFAR-10-C, CIFAR-100-C, and ImageNet-C, SPeaR consistently matches or outperforms methods that adapt orders of magnitude more parameters, remains robust across a wide range of batch sizes, and preserves source-domain performance during continual adaptation.

Questions scraped from the web are used across academia and industry as a proxy for what people want to know. Across QA training data, retrieval benchmarks, and content strategy, questions on a page are assumed to reflect human intent. We test this assumption at scale by extracting 13.4B question occurrences across 110 FineWeb snapshots (2013-2025), and report three findings. First, you can tell who is asking: provenance (the host/page of questions) leaves a signal in question form, and a logistic model can separate genuine user questions from templated/manufactured ones at AUC 0.725 via length and surrounding context rather than question type, though only 0.554 against commerce FAQ writing. Second, question frequency does not measure demand: the most-frequent questions are boilerplate/templated (over 70% of the top thousand), so occurrence counts measure how often a string was published and not how often it was asked. Third, over twelve years the genuine share of occurrences fell by 79% (42-56% after controlling for crawl composition), with question length and context decreasing. We present the first diachronic, occurrence-level measurement of web question provenance, and find the crawlable web's questions have shifted from being asked by humans toward manufactured for machines to read.

In many real-world planning tasks, agents must tackle uncertainty about the environment's state and variability in the outcomes of any chosen policy. We address both forms of uncertainty as a first step toward safer algorithms in partially observable settings. Specifically, we extend Distributional Reinforcement Learning (DistRL)-which models the entire return distribution for fully observable domains-to Partially Observable Markov Decision Processes (POMDPs), allowing an agent to learn the distribution of returns for each conditional plan. Concretely, we introduce new distributional Bellman operators for partial observability and prove their convergence under the supremum p-Wasserstein metric. We also propose a finite representation of these return distributions via psi-vectors, generalizing the classical alpha-vectors in POMDP solvers. Building on this, we develop Distributional Point-Based Value Iteration (DPBVI), which integrates psi-vectors into a standard point-based backup procedure-bridging DistRL and POMDP planning. By tracking return distributions, DPBVI lays the foundation for future risk-sensitive control in domains where rare, high-impact events must be carefully managed. We provide source code to foster further research in robust decision-making under partial observability.

Automated biomedical evidence synthesis depends on retrieving published studies, but the biomedical literature is systematically skewed toward positive findings. Deeper retrieval can therefore make a system \emph{more} likely to falsely infer benefit when the true effect is null. We formalise this phenomenon as \emph{evidence drift} and prove that, under a standard publication-bias model, the false-positive probability on null-effect queries follows a strictly increasing large-sample envelope in retrieval depth, approaching one. Empirically, on a held-out test set of 140 Cochrane-derived queries, drift rises monotonically from 7.9\% to 15.7\% as the retrieval budget grows from 3 to 20 steps, and concentrates in the null-effect class. We present DACG-agent, a drift-aware causal-graph agent that incrementally builds a causal knowledge graph from PubMed abstracts and applies a two-layer stopping policy with complementary roles: a KL-divergence monitor that detects posterior convergence (the accuracy layer), and a Bradley--Terry process reward model (PRM) whose online decline detection halts retrieval once evidence quality peaks (the efficiency layer). Against full-budget retrieval, DACG-agent reduces evidence drift from 15.7\% to 6.4\% and improves null-effect accuracy by 21 percentage points (40.0\%$\to$61.4\%) while using 67\% fewer retrieval steps; overall accuracy rises from 61.4\% to 69.3\% (95\% CI 61--77). A simulation confirms the drift result transfers from the analysed vote-counting aggregator to the deployed noisy-OR one.

Visual analytics (VA) enables sensemaking through interactive visualization, but effective analysis often requires experts to translate high-level intents into long sequences of interface operations and iteratively interpret visual feedback. We study whether modern vision-language models (VLMs) can take on this role as autonomous VA operators that observe the interface, plan multi-step exploration, execute interactions, and adapt based on intermediate visual feedback. To support systematic development and evaluation, we first introduce WidgetVA, a widget-centric agentic VA framework that standardizes interactive components as structured widgets with unified action (e.g., filter and zoom) and perception-query (e.g., selection summaries) APIs. This standardization supports two modes of system construction: wrapping an existing VA system to make it agent-operable without rebuilding it, and composing a new system from widgets as modular building blocks. To help agents coordinate across widgets rather than plan each interaction from scratch, each widget further packages reusable analytical workflows, giving agents more than a bare set of callable functions to plan over. Building on this framework, we present WidgetVABench, a benchmark of single- and multi-widget VA tasks that require agents to perform multi-step interactions to uncover evidence and produce verifiable results. Each task also provides fine-grained reference annotations so that WidgetVABench can score Answer, Reference Trace Similarity, and State separately rather than collapsing agent performance into one success score. Experiments across multiple VLMs show that our framework provides an effective scaffold for agentic VA, while the diagnostic measures expose persistent limitations for future work. The WidgetVA framework and WidgetVABench have been released in https://github.com/Hiverwin/widgetva.

Real-world document processing systems rely on rigid, predefined schemas, yet critical target fields often lack direct visual counterparts on the page. Extracting these implicit values requires multi-hop derivation, such as aggregating sub-categories or reasoning over visual marks. While existing methods handle explicit text spans or simple implicit queries, they fail at multi-hop visual reasoning even after standard fine-tuning: models retrieve incorrect visual evidence, or retrieve it correctly and then skip the intermediate steps of the derivation. To address this, we introduce DocMIDE, a fine-tuning framework that trains compact vision-language models to retrieve visual evidence explicitly before deriving an answer. DocMIDE constrains generation to a plan-retrieve-derive structure and optimizes it with Group Relative Policy Optimization under a four-component, rule-based reward that scores output format, the retrieved evidence block, every intermediate derivation step, and the final value against a verified reference trace. On a 4,151-pair implicit extraction benchmark, DocMIDE raises accuracy from 70.8% to 95.9% on Qwen3.5-4B from only a small set of annotated examples, and transfers to a second backbone architecture. Supervised demonstrations alone do not close this gap at any budget we tested; rewarding the intermediate steps is what does.

Long-horizon autonomous intelligent systems rely on heterogeneous components such as large language models, databases, external APIs, and rule engines, while their external states continuously change during execution. Re-executing the entire workflow after every change introduces substantial redundant computation. This paper proposes an incremental consistency execution method based on task fact contracts, field-level dependency masks, and state perturbation result invariant domains. After an initial verified execution, the system constructs conservative invariant domains for critical inputs and uses them to determine whether downstream results can be safely renewed without re-invoking expensive components. When re-execution is required, only the smallest affected output fields are recomputed, and an equivalence barrier prevents unnecessary downstream propagation. A submission-time version consistency gate further ensures the safety of side-effecting actions. Experiments on industrial fault diagnosis, enterprise analytics, and LLM-based multi-tool assistants show that the proposed method significantly reduces expensive component calls and end-to-end latency while maintaining high consistency and low incorrect-reuse rates.

Transformer models built on the attention mechanism have become a central building block in modern deep learning, yet softmax attention remains a major bottleneck for long-context workloads. While FlashAttention makes the forward and first backward passes I/O-efficient, it does not support backward-over-backward (BoB), which enables exact differentiation through the backward pass for applications such as second-order optimization, test-time training, gradient-based memory, and meta-learning. Existing BoB implementations either materialize large intermediate tensors or exhaust GPU memory at long sequence lengths. We present FlashBoB, an exact, I/O-efficient algorithm for BoB in softmax attention that keeps computation within on-chip tiles and avoids all $N \times N$ intermediate tensors, where $N$ is the sequence length. The key insight is a hierarchical affine structure in the softmax double backward: two row-wise scalars determine all outputs through affine transformations. This yields a two-pass schedule with bounded on-chip static random-access memory (SRAM) usage and minimal off-chip high-bandwidth memory (HBM) traffic. FlashBoB achieves $Θ(N^2 d^2/M)$ HBM traffic ($d$ is the head dimension and $M$ is the memory size) and, within the standard FlashAttention-style score-recomputation model, matches the inherited large-cache lower bound for exact forward attention. Empirically, it scales exact attention BoB to $N=262\text{K}$ on a single A100 80GB GPU, where prior PyTorch exact baselines fail by $N=16\text{K}$, and is up to $6.3\times$ faster than FlashBack. These results make exact second-order attention practical at long-context sequence lengths where prior implementations cannot run efficiently.

Open-weight large language models (LLMs) can be copied, modified, and redeployed behind black-box APIs, making post-release ownership verification difficult. Existing black-box fingerprints often rely on secret query-key pairs that reproduce predefined responses, and can therefore be easily disrupted by fine-tuning, pruning, quantization, model merging, and serving-time prompt changes. We propose SimPrint, a recoverable semantic fingerprinting framework for black-box LLM ownership verification. Rather than relying on isolated exact matches, SimPrint encodes a private owner signature into a coded semantic fingerprint domain, distributing ownership evidence across natural binary question-answering probes. It implants only base-deviating probes through a low-interference batch update that preserves the original model behavior, and later recovers the signature by parsing suspect-model responses into reliable bits or erasures with an error-correcting recovery mechanism. Because verification only uses input-output queries, SimPrint remains applicable when model weights or activations are inaccessible. Experiments on three open-weight LLMs show that SimPrint reliably recovers the owner signature in both clean and modified settings, remains robust under fine-tuning, pruning, quantization, model merging, and serving-time perturbations, and maintains comparable downstream utility.

Generative AI enables scalable production of educational videos, but current systems largely focus on producing visually coherent content rather than supporting learning. As a result, generated videos often lack explicit pedagogical structure, reliable quality control, and mechanisms for assessing learner understanding or addressing misconceptions. In this work, we introduce PIVOT (Pedagogy-guided Instructional VideO Tutoring), a generative video tutoring framework for STEM learning via learning-centered instructional support.1 Inspired by conventional teaching workflows, our framework integrates pedagogy into the full generation pipeline: it first uses instructional principles to guide storyboard generation, then produces verified multimodal videos through code-centric generation and a pedagogical verification harness, and finally connects videos with assessment and misconception-aware remediation. Experiments and expert evaluations across four STEM domains show that our framework produces educational videos with pedagogically aligned content, clear and engaging presentation, coherent instructional flow, and perceived effectiveness for learning. These findings suggest a human-centered perspective on educational content generation: generative systems should be evaluated and designed not only by what they produce, but also by how they support teaching practices, learner understanding, and corrective feedback.

Best-of-$N$ is a widely used inference strategy for complex reasoning, whose effectiveness depends on whether sampled candidates can cover diverse and high-quality reasoning paths. However, post-trained reasoning models often suffer from \emph{exploration collapse}, where independent rollouts repeatedly follow similar reasoning paths and limit the gains from increasing the rollout budget. Existing methods alleviate this issue by promoting broader exploration, but do not explicitly guide exploration toward continuations that make meaningful progress, resulting in limited exploration efficiency. To address this, we propose \emph{\underline{S}tate-conditioned \underline{P}rogress-guided \underline{S}teering} (SPS), a training-free latent steering framework. Specifically, SPS constructs a state-conditioned Direction Bank containing multiple progress-guided steering vectors for different prefix-state regions. During online inference, SPS retrieves a suitable steering vector based on the current prefix state and applies it at high-uncertainty transitions to guide the next reasoning step toward meaningful progress. Extensive experiments across multiple model scales and benchmarks demonstrate that SPS consistently outperforms strong baselines. Further analyses validate the effectiveness of its key designs and offer valuable insights for future research. The code is available at https://github.com/rattlesnakey/SPS.

New Guinea is the world's richest island flora (~2,856 orchid species), yet most species are represented by only a handful of photographs, far fewer than direct species-level classification requires. Methods for fine-grained identification in such species-rich, data-poor floras are needed, and it remains unclear which backbone architecture and pretraining strategy best support them. We built a two-stage system that first predicts the genus of a query photograph, then retrieves visually similar reference images of candidate species using FAISS. We compared four pretrained backbones -- two Vision Transformers (ViTs; DINOv2, BioCLIP 2) and two CNNs (ConvNeXt V2-L, EfficientNetV2-L) -- fine-tuned under an identical protocol on a fixed, species-stratified partition of 16,701 photographs spanning 120 genera and 1,350 species, assessing accuracy, calibration, error structure, species retrieval, and open-set detection of novel genera. DINOv2 attained the best genus performance (macro top-1 66.9%, 95% CI 63.7-70.6; global top-1 88.9%); both ViTs outranked both CNNs, and general-purpose self-supervised pretraining (DINOv2) outperformed domain-matched biological pretraining (BioCLIP 2) by 7.1 points of macro top-1. Errors concentrated on two abundant genera acting as error attractors. DINOv2 embeddings achieved species Recall@5 of 86.6% and genus Recall@5 of 98.7%; temperature scaling reduced every backbone's Expected Calibration Error to about 0.03; and a distance-based open-set gate flagged unseen genera (mean AUROC 0.958). A self-supervised Vision-Transformer backbone combined with embedding retrieval is an effective, deployable strategy for fine-grained identification in species-rich, data-poor floras. The system is released as an open web application (the New Guinea Orchid Identifier), offering a practical template for other hyperdiverse, under-documented taxa.

In this paper, we propose MECT, a speaker verification model that integrates the Mixture-of-Experts (MoE) mechanism into a CNN-Transformer backbone with optimized block structure and stacking scheme. Specifically, we investigated four MoE variants that span utterance-level and frame-level granularity with dense and sparse routing strategies. The MoE mechanism proves to be effective over the baseline without MoE with only a small increase in parameters. We further scale MECT to a series of model sizes, all maintaining compact parameters and low computational complexity. In particular, MECT-B2 achieves state-of-the-art performance on VoxCeleb1 and delivers strong results on CN-Celeb, demonstrating its effectiveness across diverse datasets. In addition, we establish a streaming inference paradigm through causal retraining, which maintains strong performance at a chunk size of 100ms.

Medical image interpretation is central to diagnosis and care, yet adapting general-purpose multimodal large language models (MLLMs) often requires resource-intensive domain-specific fine-tuning. Here we introduce representation-guided in-context learning (RG-ICL), a training-free inference framework that retrieves query-aligned demonstrations using frozen encoders, without task-specific parameter updates. Across eight datasets spanning histopathology, radiology and retinal fundoscopy, RG-ICL improved classification (mean gain 20 percentage points) and visual question answering (VQA) (mean gain 13 percentage points) over no-context and conventional ICL, approaching or exceeding training-based comparators. Which cases were retrieved mattered more than how many: 6 query-aligned cases outperformed up to 32 randomly selected ones, whereas fixed or random cases often reduced accuracy below baseline. For VQA, aligning reference cases with both image content and question intent produced further gains. These findings indicate that for medical image interpretation, curating which reference cases an MLLM sees is a practical alternative to retraining it.

Crash datasets that carry an investigator narrative hold information the coded fields omit. Coding those narratives at scale has been blocked by three obstacles. Frontier large language models are costly at that scale, their generated text cannot be verified, and no rule says how much output a human must check. This paper formulates narrative coding as gated, typed decisions answered by Jev, a System One model that returns probabilities over analyst-defined options and generates no text. A screen covered 499,500 Texas narratives and 195,857 were coded with a 27-question schema. Cost is governed by schema size rather than narrative length. The probabilities are audited against coded fields and against 2,416 blinded human judgments drawn under a stated sampling design. Two frontier large language models are benchmarked on the same records. Against human labels the typed model attains an F1 of 0.908. One frontier model gains 0.059 and the other is indistinguishable from it. Calibration varies by model rather than by paradigm, so each model must be audited. Recalibration on the same labels reduces calibration error by a factor of 3.3. Agreement with coded fields understates fidelity to the narrative by a median of 0.26 in kappa. A resolution-floor bound covers any model that reports probabilities on a discrete grid. A review budget over flagged records gives the records a human must read per variable and per year. Adding the calibrated variables to the coded fields raises the injury and fatal crashes attributed to nine factors by 10,747 per year.

Edge deployment motivates forecasting models with compact parameter storage and low-bit representations. Deep equilibrium models (DEQs) obtain implicit depth by repeatedly applying a shared layer, reducing the parameter cost of explicit layer stacking. Their usual Anderson solver, however, searches for update coefficients in the continuous real domain. We propose Q-DEQ, which formulates local updates in DEQ forward solving as discrete optimization problems. Candidate directions are constructed from the current state and iteration history, and a local quadratic residual model is used to evaluate their combinations. Binary encoding of the direction coefficients yields a quadratic unconstrained binary optimization (QUBO) problem that can be solved by simulated annealing (SA) or a coherent Ising machine (CIM). After fixed-point solving, a re-forward pass applies W8A8 fake quantization to the shared layer's weights and activations. We evaluate Q-DEQ with an iTransformer backbone on five multivariate time series forecasting datasets. Relative MSE differences from the explicit multi-layer baseline range from $-1.16\%$ to $+2.90\%$, with lower MSE on two datasets. DEQ parameter sharing reduces parameter counts by factors of $1.80\times$--$3.82\times$; combined with W8A8, static weight storage is reduced by factors of $4.3\times$--$12.8\times$. Local QUBO problems solved using CPU-based SA and the Kaiwu CIM physical backend produce closely matching downstream forecasts. These results establish local discrete solving as a viable component of DEQ time series forecasting and provide a route for executing fixed-point updates through different combinatorial optimization backends.

Dense pass surfaces give, for every pitch cell, whether a pass played there would arrive, whether the carrier would choose it, and what the possession would then be worth. The networks that draw them read the state as a raster of per-cell counts, losing where inside a cell each player stands. LiDAR detectors, bird's-eye-view perception and graph weather models move entity features onto a grid, binning each entity to a cell or learning the transfer. We evaluate the interpolated form: each player's features are scattered bilinearly onto the grid at the player's measured coordinates, so the surface loss trains the per-player encoder end to end. Those systems adopt an interface; this paper measures one. On 53,628 passes from the 2022 World Cup, painting improves selection likelihood over the same core fed rasters alone by about a quarter of a nat: in every match of an eight-fold cross-validation, with every arm tuned over five seeds, and after retraining on seven Bundesliga and 2. Bundesliga matches from another provider. Thirteen pre-specified studies locate the gain: painting the nine raw player features with no encoder carries three quarters of it, and the learned encoder and message passing add a smaller, resolved increment. Painting also helps the original SoccerMap and a canonical U-Net, whereas offset channels, a finer raster, an attention painter and a raster-free decoder do not. Frozen across the provider boundary the likelihood advantage is lost; injected tracking error compresses it. These results concern observed-endpoint prediction, not calibrated evaluation of hypothetical passes.

This paper presents an LLM-based system that translates natural-language access control policies (NLACPs) into executable Rego code for Open Policy Agent (OPA). It provides a modular, end-to-end pipeline for policy detection, component extraction, schema validation, linting, compilation, and automated test generation and execution. The system is designed to bridge the gap between human-readable access requirements and machine-enforceable policy-as-code (PaC), with a focus on deployment reliability and security correctness. We evaluate the system on 372 ACRE-complete access control statements with non-null subject, action, and resource annotations against a direct single-prompt LLM baseline to isolate the contribution of structured decomposition and schema-aware validation. The system achieves a 50.3% end-to-end policy correctness rate, compared with 15.3% for the baseline, representing a 3.3x improvement. A policy is counted as correct only if it satisfies compilation, linting, and both positive and negative tests, making this a strict measure of deployable correctness. On security-critical patterns, the system generates correct deny semantics for 87.5% of deny policies (baseline: 37.5%), ownership conditions for 100% of ownership-qualified policies (baseline: 40%), and status-qualified conditions for 100% of status-qualified policies (baseline: 55.6%). These results indicate that structured decomposition and schema-aware validation play a critical role in improving the reliability of LLM-generated authorization policies.

Generating coherent meta-reviews from multiple peer reviews is challenging when reviewer evidence conflicts and varies in reliability. Existing approaches typically formulate meta-review generation as a multi-document summarization task and aggregate reviewer feedback uniformly, making it difficult to determine which opinions should be prioritized under disagreement. In this paper, we study meta-review generation through reliability-aware evidence aggregation. Our framework first extracts aspect-level opinions from peer reviews and identifies conflicting evidence within each aspect. It then estimates opinion-level support and review-level quality to measure evidence reliability. Based on these signals, the framework assigns reliability-aware weights to reviewer feedback, enabling the generator to prioritize better-supported arguments while preserving diverse perspectives. Experiments demonstrate that our method consistently improves meta-review generation over strong baselines on both automatic and human evaluations, with clear gains in conflict recognition and resolution under high-conflict review scenarios. The code and implementation details are publicly available at https://github.com/Wangxz729/reliability-aware-meta-review.

In this paper, we investigate the limitations of fixed, hand-crafted connectors in hierarchical semantic representations for open-vocabulary object detection. Existing methods establish semantic relationships between base categories and unseen novel categories by placing a fixed connector between adjacent super-/sub-categories. However, such fixed connectors may not optimally capture the relationships within a semantic hierarchy. To address this limitation, we propose interconnected hierarchical semantic representations (InterHier), which utilize a prepended learnable context to globally guide the interpretation of prompts containing hierarchical relationships. InterHier operates in two main stages. First, it constructs a hierarchy-aware prompt by integrating super-/sub-categories and prepending a learnable context. Second, it optimizes this learnable context to align visual region embeddings and textual embeddings. InterHier consistently improves performance over methods that rely on fixed connectors and can be seamlessly integrated into existing open-vocabulary object detection models. Experiments on open-vocabulary object detection benchmarks demonstrate that InterHier achieves competitive performance against state-of-the-art methods.

We present a method for synthesizing reactive character behaviors for continuous games as compact, human-readable programs. Game AI practice still relies heavily on manually authored behavior trees, state machines, and scripts, while academic reinforcement learning typically produces opaque neural controllers that are expensive to train and difficult to edit. Our approach bridges this gap by searching directly over a domain-specific language for continuous-space game policies. The language is designed around reactive geometric decisions and includes higher-order constructs such as direction maximization. These constructs help discretize a continuous behavior space into enumerable program structures. To make program search practical, we introduce a large set of synthesis antipatterns that remove redundant program forms while preserving behavioral coverage. We further combine bottom-up symbolic enumeration with top-down guidance from a coding agent. Our resulting method, agentic sketching, has the agent propose high-level policy structure and call an enumerator to complete local program slots. We evaluate the method on a benchmark of 14 continuous games, ranging from classic control tasks to multi-agent football. We find that pure enumeration is often more efficient than using a coding agent alone, while the combined method substantially outperforms both. Our results suggest that programmatic policy search can be a practical authoring tool for game AI: designers specify reward functions, and the system discovers editable behaviors that are effective, portable, and often surprising.

Neural operators are data-driven models that learn mappings from inputs that parameterize partial differential equations, such as spatially varying coefficients, initial conditions, forcing terms, boundary conditions, or geometries, to solution fields or quantities of interest. Once trained, they can serve as surrogates for classical numerical solvers in many-query settings that require repeated evaluations for varying inputs. We address the question of when, and then why, neural operator surrogates outperform classical numerical solvers, in terms of cost for a given accuracy. We focus on the post-training, many-query limit, in which data-acquisition and training costs are treated as fixed and fully amortized. Even in this deliberately favorable regime for neural operators, there are regimes in which classical solvers outperform the surrogate models. We compare the cost-accuracy performance of neural operator surrogates and classical numerical solvers through a reproducible benchmark study comparing neural operators with problem-matched classical solvers on representative problems in computational science and engineering, focusing on prediction error, per-query floating-point cost, and wall-clock runtime. Neural operators are most competitive at low-to-moderate accuracy requirements. Their floating-point cost advantage depends strongly on the problem structure, arising when they avoid temporal or nonlinear iterations or predict a reduced quantity of interest rather than a full solution field. Additional wall-clock speedups result from dense tensor operations that are well suited to modern hardware. As the target accuracy is tightened, achieving the required accuracy with neural operators becomes increasingly challenging, and classical solvers outperform surrogates in this regime; thus classical solvers will remain important for verification and high-accuracy computation.

Commercial large language models are increasingly deployed across African fintech infrastructure for fraud detection and customer communication, yet no Nigerian or African continental regulatory instrument specifies what pre-deployment evaluation such systems must undergo before procurement. This paper reviews African fintech AI governance across global, continental, and Nigerian instruments, and shows that safety is affirmed as a principle while pre-deployment evaluation is operationally unspecified. Generic safety benchmarks cannot surface the failure modes most relevant to this domain, since none contain Nigerian institutional content or test for false positive misclassification of legitimate financial communications. These claims are demonstrated using SafeAlert, a purpose-built evaluation kit applied to six commercial models across three system prompt conditions. Results show that models resisting generic harmful content requests still produce complete fraud scripts under specific framing, and that several models misclassify most legitimate Nigerian bank communications as suspicious or fraudulent, a failure invisible to standard safety evaluation. The paper concludes with a regulatory framework proposing pre-deployment evaluation requirements for the CBN, NITDA, SEC, and the AU, arguing that the identified gap reflects an absence of regulatory specification, not a shortage of technical or financial resources.

Validation of generative social simulators often stops at face validity: emergent network structure is compared descriptively, without quantified parameter uncertainty or an adequacy check. We present an adequacy-aware calibration protocol that couples amortized posterior estimation with a synthetic identifiability assessment, a matched-sample-size adequacy check (prior-predictive reachability plus per-statistic posterior-predictive localization), a diagnosis-guided repair, and a statistic-held-out audit. We demonstrate it on a real second-hand luxury resale market with four channel-by-residency cells, each a bipartite buyer-brand network, using a forward model built from persona profiles elicited once, offline, by a language model. The behavioural parameters are recoverable in all four cells, though calibration is approximate and overconfident for one parameter. The observed summary falls outside the simulator's reachability reference in every cell, with the mean purchased tier as the pervasive discrepancy. The repair meets the value-block criterion in two of four cells but does not restore adequacy, and the held-out audit surfaces a buyer-breadth-dispersion miss no earlier diagnostic detected. A profile-source ablation finds the language-model profiles beat a flat rule baseline in all four cells, yet within-category brand relabelling causes no consistent degradation, so the profiles are a partially validated input whose value rests on structure, not brand identity. Making no causal claim, we conclude that an independent-aggregation account, without agent interaction or a buyer-breadth mechanism, cannot jointly reproduce the market's purchased-tier level, head-brand concentration, community structure and buyer-breadth heterogeneity.

Text-controlled time series generation aims to synthesize sequences that follow natural-language descriptions while remaining faithful to real data distributions. Existing paradigms often couple semantic understanding and sequence modeling in a single continuous latent space, lacking explicit local semantic anchors and separation between global continuous attributes and local discrete shapes. As a result, key local structures may be smoothed, missed, or misplaced. We propose Shape Lexicon (ShapeLex), which decouples text-to-sequence generation into discrete symbolization of local shapes and continuous modeling of global attributes. ShapeLex first induces a reusable vocabulary of discrete shape units, such as rises, spikes, and sharp drops, from training data, forming an interpretable symbolic space. An autoregressive generator then selects shapes according to the textual description, adjusts attributes such as position and duration, and composes them in temporal order into a shape skeleton. Finally, a mixture-density scale head models and samples the overall level and volatility to restore realistic global scale. Experiments on twelve public datasets, real user-written text, and downstream forecasting tasks show that ShapeLex generates series that better match real data distributions than existing methods. In addition, paired supervision is automatically synthesized from the learned vocabulary, avoiding annotation costs that grow with dataset size and improving scalability.

Large language model agents increasingly answer financial questions by searching regulatory filings. Such questions are often deceptively under-specified: Meta Platforms' "operating income" is $46.75B consolidated but $62.87B for the Family of Apps segment, and each reading is exactly verifiable against the filing. A capable agent should recognize the ambiguity and ask, rather than commit to a plausible but unintended reading. Existing financial benchmarks cannot measure this, because one gold answer per question cannot separate agents that resolve the ambiguity from those that guess the common reading, a blind spot we call the single-gold illusion. We release FinInteract, a bilingual (English/Chinese) benchmark of 173 instances that pairs each question with a default and an intended interpretation across a five-category ambiguity taxonomy, and grades whether an agent elicits the right clarification and then integrates it. Re-grading identical outputs against the default rather than the intended reading inflates GPT-4o's accuracy by 3.1 times, confirming the illusion. Beyond it, we find that models answer above 90% once the interpretation is supplied but at most 28.9% when they must elicit it themselves, that targeting is uneven across a taxonomy well powered for entity scope and metric definition and exploratory elsewhere, and that conditioning on the ambiguity category improves resolution at both inference and training time.

How large language models (LLMs) integrate patient risk with clinical cost tradeoffs remains poorly understood. We investigated how four open-weight LLMs (Qwen-2.5-7B/32B and Llama-3.1-8B/70B) internally represent cost tradeoffs, how these representations relate to clinical predictions, and whether decisions shift as predicted by the specified cost direction and magnitude. Using a public diabetes dataset, we varied 11 false-negative (FN) to false-positive (FP) cost ratios across three phrasings and examined representations and behavioral outputs. Patient risk was linearly recoverable on par with conventional classifiers (AUC $\approx 0.83$), and cost direction was recoverable in every model. However, representational shifts in cost direction tracked output changes only in the two larger models, and responses to cost magnitude were predominantly direction-agnostic. Only 2 of 12 model-phrasings showed both opposing responses to increasing FN versus FP costs and cost-correct ordering. Representationally, a direction fitted on one cost side did not invert when transferred to the other, as expected under mirror-symmetric encoding. These findings suggest that LLMs encode risk and cost information but do not reliably integrate them into cost-correct decisions. Clinical evaluations should therefore include tradeoff tests, phrasing sensitivity, and default operating points alongside predictive performance.

Multi-robot collaboration could enable more efficient and scalable solutions to complex robotic tasks, but collaboration under partial observability remains challenging. Natural-language communication offers a promising approach to coordinating robots under partial observability. However, in decentralized manipulation, jointly learning explicit inter-robot communication and skill-level action selection from multimodal demonstrations remains underexplored for small vision-language models (VLMs) intended for on-device deployment. To address this gap, we introduce RoboTalk, a synthetic data-generation pipeline and dataset of 7,950 multimodal trajectories spanning 53 mobile-manipulation kitchen tasks for training small VLMs to communicate and coordinate. The dataset includes a leader-follower planning protocol, tool calls (perception, manipulation, navigation, and communication), rationale traces, and diversified natural-language communication. Fine-tuning open-source models on our dataset can reach 77% success on novel held-out tasks, a significant improvement over the untuned open source models, which had a success rate of around ~2%.

Linear models with an $L_1$-norm penalty remain state-of-the-art for high-dimensional ($d > 1,000,000$) tasks, offering a straightforward method for solving real-world industry problems. Despite their widespread use in industry and utility, many $L_1$ solvers are not effective for general use, are prohibitively slow, and are ineffective in parallelization. This makes them difficult to train in an MLOps pipeline on large industry-scale corpora. In this work, we test several proposed ``state-of-the-art'' solutions from the literature and find that older methods are currently far superior for general use. We also identify several recommendations for academics to perform research that avoids erroneously overconfident results, which can prevent the transition to production use. Equally surprising, we find that a new and simple baseline, using LBFGS on a sub-gradient, is highly effective with minor tweaks, despite being dismissed in the literature for theoretical non-convergence. In practice, we find it is an easier-to-support and easier-to-scale method for production use.

Tracking neurite morphology over time helps characterize structural changes during neuronal development and deterioration, but long-term time-lapse imaging is resource-intensive and difficult to scale. Forecasting future morphology could reduce this burden. Existing neurite digital-twin models such as gated spatiotemporal attention (gSTA) produce a single deterministic forecast without representing variability among plausible futures. We introduce Morphology-Gated Residual Diffusion (MGRD), a compact stochastic surrogate that jointly forecasts twenty future neurite-morphology frames from ten observed frames while conditioning on morphology features derived from the latest observation. On controlled phase-field trajectories, MGRD reduces trajectory-wise mean MAE by 9.7% relative to a matched control while updating 4.46 times fewer parameters. On human iPSC-derived neuron microscopy, MGRD improves all four reported metrics over gSTA, including a 39.6% reduction in trajectory-wise mean MAE and a 45.3% increase in skeleton F1. Without mouse-domain retraining or fine-tuning, MGRD also improves MAE and skeleton F1 on mouse cortical-neurosphere microscopy across 10-40-min sampling intervals and forecast horizons beyond 13 hours. Repeated sampling provides a case-level variance score for ranking forecast difficulty. Retaining approximately 60% of the lowest-variance cases reduces mean MAE by 17.6% on iPSC microscopy and 16.8% on simulation data. MGRD uses 1.01% of gSTA's parameters, requires less than one tenth of its training-update time, and generates a 50-step DDIM trajectory 7.9% faster when morphology features are cached. These results establish MGRD as a compact stochastic surrogate for neurite-morphology forecasting and case prioritization across simulation and microscopy datasets.

Building general-purpose agents for industrial deployment requires integrating multiple capabilities, each typically acquired at a distinct stage of training. Yet there is currently no well-established recipe for Agent Continual Learning (ACL), with little understanding of the trade-offs among existing integration paradigms. To address this gap, we introduce ACLArena, a framework for comprehensively studying, analyzing, and evaluating ACL. We first build a sequential training pipeline and conduct an in-depth analysis that explains the mechanisms of forgetting and generalization from two complementary perspectives, the model level and the token level. Guided by these analyses, we systematically compare multi-teacher on-policy distillation, self-distilled fine-tuning, and model merging to assess their ability to recover previously learned capabilities while preserving newly acquired ones. Through extensive experiments, we develop a detailed understanding of how capabilities transfer across stages. Finally, we propose a new ACL recipe that combines offline replay over high-quality trajectories with a routed network of multiple LoRA experts each specialized via RL, substantially improving the agent's ability to learn across multiple domains. Comprehensive experiments on four reasoning and agentic tasks, evaluated under both in-domain and out-of-domain settings, demonstrate the value of our analysis and the effectiveness of our approach.

Renewable forecasts are evaluated by statistical skill (e.g., CRPS), but grid operators pay for realized dispatch cost. We diagnose what drives dispatch value in a single-period newsvendor-style economic dispatch using real public data from two European transmission systems (CWE, DE-4TSO). Spatial coherence across forecast sites falls below the pre-specified 1% practical-significance threshold: a controlled ablation holding per-zone marginal forecasts bit-identical and varying only cross-zone dependence (10 configurations, 3 seeds, paired-bootstrap confidence intervals) shows a coherence gain of at most 0.64% of dispatch cost, indistinguishable from zero in 3 of 10 configurations, reached only under an unrealistic 8-fold forecast-error stress test. Decision-focused training, an established paradigm in this venue, delivers a robust 2.82-5.19% gain. A parametric Gaussian-copula approximation matches the empirical copula at realistic error magnitudes but performs worse than no dependence under extreme stress. A single-seed sweep shows that a 12% energy-score gain changes cost by less than 0.1%. Results characterize this single-period dispatch class; a lightweight four-period extension supports the same conclusion. For this dispatch class, spatially-correlated scenario generation provides limited operational value on its own; grid operators and forecast vendors should instead evaluate dependence models by downstream decision value and prioritize decision-focused training.

Agentic memory is becoming essential for long-horizon AI agents, yet many existing systems rely on autoregressive LLMs to control how memories are organized, retrieved, and used, placing expensive generation on the critical path of memory operations. We introduce \textbf{\method}, a new agentic memory architecture inspired by System-One/System-Two cognition. System One captures fast, lightweight decision-making, whereas System Two performs slower, deliberative reasoning. Jev-Mem brings this division of labor to agentic memory through a dedicated System-One control plane, a structured multi-relational memory plane, and a System-Two reasoning plane. The System-One controller governs memory typing and relational organization during construction, and dynamically performs query routing, retrieval-budget allocation, graph traversal, candidate scoring, and adaptive stopping during retrieval. System Two is invoked only for complex reasoning and answer synthesis. This design improves both memory effectiveness and system efficiency: on LoCoMo Jev-Mem achieves an overall LLM-as-a-Judge score of 0.777, an 11.0\% relative improvement over the strongest baseline, while reducing memory construction time to 158\,s, a 6.6$\times$ speedup over the fastest competing memory system, and lowering average query latency to 0.93\,s, a 36.7\% reduction.

AI agents now report vulnerabilities faster than maintainers can review them. Reports often depend on security properties specific to the application, and require considerable human labor to process. To mitigate this, we introduce a framework for evaluating vulnerability reports via probes, executable checks of security properties. A reported exploit is evaluated by replaying it against the application and running the probes: a triggered probe indicates both that the exploit succeeded and which security property it violated. As a probe encodes a security property rather than a known vulnerability, it can detect vulnerabilities that were not known when the probe was written. We instantiate the framework as MobileCybench, a benchmark for vulnerability discovery by AI agents in 13 Android applications, with 495 probes written and reviewed by the authors. We evaluate 5 coding agents (OpenCode with GPT-5.5, GPT-5.6-Sol, and GLM-5.2; Claude Code with Opus 4.8 and Opus 5) under 4 settings: as a malicious app on the victim's device or as a remote attacker with a low-privilege account, each with either only an obfuscated APK or access to the application's source code. Given only the obfuscated APK, the top agent, OpenCode with GPT-5.6-Sol, triggers probes in 53.8% of applications in the malicious-app setting and 16.7% in the remote-attacker setting. With source code, the trigger rate across all agents and both attack settings increases from 28.8% to 32.8%. Building and running the benchmark surfaced 23 previously unreported vulnerabilities, the majority of which have been confirmed by maintainers.

Attention encoder-decoder (AED) Speech Foundation Models achieve strong ASR performance but can generate acoustically unsupported text when inputs contain no speech, weak acoustic evidence, or unreliable transcription. We propose AURA: Activation-editing with Uncertainty-Routed Adaptation, an ultra-efficient representation-editing method that freezes the pretrained model and applies sparse scale-and-shift edits to decoder cross-attention heads. AURA dynamically routes edits using cross-attention uncertainty features that capture over-concentration, diffuse attention, and abrupt frame shifts. We evaluate AURA on four datasets spanning non-speech hallucination and speech grounding stressors, including imperfect-label child speech, imperfect-label adult speech, and disfluent speech. On non-speech audio, AURA reduces hallucination rate from 89.18% to 1.94% without prior hallucination-head identification. On imperfect-label corpora, AURA approaches LoRA WER while using roughly 500x fewer trainable parameters. Sensitivity analysis and qualitative cross-attention examples are consistent with AURA's uncertainty-routed editing behavior, supporting dynamic activation editing as a practical path for grounding AED speech models.

Foundation models are endowing autonomous systems with greater intelligence, enabling a more comprehensive understanding of the environment through visual perception. A representative example is Human Mesh Recovery (HMR), which provides useful estimates of a target's 3D pose and shape that can benefit tactical missions. However, the size and power demands of such models make them difficult to run on edge platforms and limit their real-time performance, undermining the requirements of tactical edge deployment - especially for active perception, where a mobile robot must plan its next-best view on-board and cannot offload computation under contested communications. We present LEAP-NBV, a lightweight active-perception framework that runs foundation-model-driven Next-Best-View (NBV) planning on-board an edge device. To this end, we distill a family of large HMR teachers, each into a compact 32M student, with an offline mesh objective, then quantize the vision encoder to FP16 and characterize its on-device accuracy and latency. Within an occlusion-aware active perception loop, we evaluate all configurations on the same held-out benchmark and deploy the end-to-end pipeline on an NVIDIA Jetson Xavier NX, reporting measured on-device latency and energy. Distillation recovers 6-7 mm of Procrustes-aligned mean per-vertex position error (PA-MPVPE) over the undistilled student on the test set. Selecting the edge-optimal compression model brings the HMR engine to ~12 ms at a small accuracy cost and runs the full closed loop at 3.6 FPS and 2.6 J per frame, achieving a 2.0x speedup and 3.0x lower energy than the uncompressed model while nearly matching downstream task quality.

Two transformative classes of AI systems are reshaping how organizations operate: \textit{digital AI}, which reasons over enterprise knowledge to power chatbots and agent workflows; and \textit{physical AI}, which learns to control robots and autonomous systems from video, gameplay, and sensor telemetry. Both face the same foundational bottleneck: raw knowledge at scale, spanning heterogeneous modalities, locked in private corpora that existing AI infrastructure cannot access reliably or efficiently. We propose \textit{Universal Knowledge Perception (UniK)} as a common platform for both classes, covering the full knowledge lifecycle (ingestion, enrichment, indexing, retrieval, and continuous evaluation) across modalities from rich text and video to molecular data and sensor telemetry. We present UniK, built on Polymath Retrieval (multi-index fusion over automatically enriched indices) with no task-specific fine-tuning. Across five digital AI domains (medical literature, open-domain QA, chemistry, legal video proceedings, and government open data) UniK combined with an open-source 70-billion-parameter model consistently matches or outperforms frontier proprietary LLMs that are orders of magnitude larger: 76\% RAG accuracy on government data versus 47\% for GPT-5; 77.9\% on medical QA without fine-tuning; topping all open-source chemistry pipelines. We show that the same infrastructure directly addresses the data curation, indexing, and retrieval challenges facing physical AI world model training, where the knowledge problem is harder but structurally identical.

We develop exponential family synthetic controls (EFSC), a distributional version of synthetic controls for a panel of datasets. Each cell of the panel corresponds to a dataset drawn from an exponential family whose natural parameters factorize probabilistically across units and times. We estimate the latent factors using black-box variational inference. This replaces the usual weighted-average view of synthetic controls with a flexible probabilistic model that operates on full distributions. We propose causal estimands based on divergences between pre- and post-intervention distributions induced by the posterior of the natural parameters, together with distributional placebo tests to support causal inference and assess the significance of the estimated effects. We validate the proposed framework on synthetic and real data. Across a variety of exponential-family distributions, EFSC accurately recovers causal effects induced by exponential tilts, together with the corresponding divergences between treated and counterfactual distributions. The framework also captures effects induced by structural perturbations of the latent factors and by heavy-tailed noise contamination. Finally, we apply EFSC to study the expansion of Medicaid under the Affordable Care Act (ACA) and its impact on the distribution of health insurance coverage across U.S. states. Code is available at https://github.com/blei-lab/efsc.

Diffusion models are increasingly adapted from generation to conditional prediction, where a conditioning signal is combined with an evolving noisy representation of the target. In fully supervised segmentation, however, the conditioning image can already support direct target prediction, so endpoint performance alone establishes neither reliance on the added diffusion state nor a deterministic advantage over image-only prediction. For state reliance, we disrupt target-derived state content or correct image-state pairing during retraining of twelve published methods across three datasets, with ten matched seeds per setting. All 40 original-method comparisons whose evaluated-mask routes remained downstream of noised-quantity reconstruction exhibited state reliance, whereas all 30 comparisons with a segmentation-supervised bypass preserved reference performance. Rerouting five originally bypass-capable methods by forcing segmentation supervision through noise-to-mask reconstruction converted all 30 corresponding comparisons from preserved performance to state reliance. For deterministic utility, matched image-only counterparts achieved similar or better performance in 28 of 35 settings overall, including 16 of 20 whose native methods relied on both audited state properties. These results identify supervision path as a determinant of state reliance in the audited methods. Separately, matched image-only counterparts show that diffusion-specific computation often provides no deterministic endpoint advantage, including in methods that rely on the audited state properties. More generally, when conditioning already supports strong target prediction, diffusion-specific claims require additional evidence that the added state is used and that diffusion-specific computation improves the claimed capability beyond a matched condition-only counterpart.

LLMs can generate fluent descriptions from tables, but their outputs may remain logically unsupported by the structured data. We introduce STAT-TO-TEXT, a controlled task in which LLMs generate quantified natural language inferences from statistical tables using quantified constructions such as all, some, no, and most. To evaluate these inferences, we use an LLM generated Python checker code which when executed verifies the corresponding truth conditions against the table. We compare four open-weight LLMs across model families and scales, evaluating faithfulness, logical accuracy, table coverage, and diversity. Our results show that model scale and family matter, with the largest model (GPT-OSS-120B) consistently producing the most faithful inferences without sacrificing greater table coverage and quantifier diversity, as opposed to smaller models. These findings are supported by human annotation, which shows that the automated checker closely aligns with human judgments.

Screening a phone call for fraud needs a trustworthy probability after every caller turn, in milliseconds. Jev-style typed decisions promise exactly that: declared options go in, one calibrated probability per option comes out of a single forward pass, with no generated text. We test an open implementation of this readout, JevLite, on scam-call screening: Qwen3-4B is LoRA-tuned so that the temperature-scaled softmax over two answer-label logits is P(scam). On 41 held-out CallScreenBench scenarios (577 per-turn decisions) a three-seed ensemble reaches AUROC .974 with calibration error .052, non-inferior to an LLM judge (MiniMax-M3) at a pre-registered .02 margin, with no false alarms on legitimate calls, decisions 1.14 turns earlier under the same hang-up rule, and 64.5 ms per decision on one consumer GPU, 4.9x lower than the same backbone fine-tuned to generate its answer. The gain is in the readout and calibration, not accuracy: a fine-tuned ModernBERT encoder is not significantly worse, the recipe was selected with test-set exposure, and all callers are synthetic. We claim no architectural novelty; the contribution is the application and an evaluation reporting calibration, false alarms and decision timing alongside AUROC.

Scientific agents are mostly evaluated on whether they complete tasks or recover known results; we instead study variation across repeated open-ended campaigns. Sixteen separately initialized sessions of one model-harness configuration received a frozen database of 12,499 metal-organic frameworks, a methane-storage objective, a pinned protocol and a one-week budget. Strategies diverged into four approaches spanning 100--5,000 screened structures, and eight built 2,253 hypothetical structures. Yet the agents recovered the same materials frontier near 200 cm^3/cm^3, and an independent calculation of the database's porous region found its nine best structures all among their reports. Enforced checks on half the agents raised fresh-run reproduction from one of eight to eight of eight but could not detectably improve conclusion validity, because fifteen of sixteen agents selected the same audit-excluded entry, an incomplete structure whose missing anions created artificial pore volume. Replicated agents thus reveal both robust conclusions and common-mode errors from shared inputs.

Previous work on Egyptian Arabic in NLP has focused largely on the prestigious Cairene Egyptian Arabic (CEA) dialect, resulting in a lack of representation for the less prestigious Sa'idi Egyptian Arabic (SEA) dialect both in LLM and resource development. Does this lack of representation influence an LLM's view of the acceptability of SEA (upstream), and does an upstream bias against SEA lead to worse performance (downstream)? We investigate the upstream effect of SEA dialectal features on LLM preferences in a Targeted Syntactic Evaluation (TSE) task which reveals a significant bias against SEA across multiple LLMs. We then analyze the effect of these same features on downstream model performance on MMLU benchmarks and show that models experience a degradation in performance when presented with SEA. This work highlights the need for further exploration on how sub-dialectal variation impacts language technologies.

Programmers write formal specifications, and LLMs implement them, proving that each implementation matches its spec. Taken to its extreme, this makes specification languages the new programming languages. We argue that an unambiguity constraint is key: in addition to proving that its implementation satisfies the specification, the LLM must also prove that any other implementation satisfying it must produce the same outputs on the same inputs, i.e. that the relation formed by the constraints is deterministic. This leaves the LLM no leeway on program semantics: as with a conventional compiler, the generated code never needs to be read and can be regenerated from the spec at any time. Under this constraint and with a powerful LLM, the difference between a specification language and a programming language becomes essentially meaningless, and the LLM essentially becomes a part of the compiler toolchain. The arrangement doubles as a strong form of AI control: an untrusted model writes the code, yet its work is tightly checked by a verifier. To demonstrate that our LLM-in-the-compiler paradigm is feasible when supported by our unambiguity constraint, we present Djinnlang, a high-level specification language built for this future. A Djinnlang program consists only of specifications --- the programmer never writes executable code. In place of a traditional compiler, a symbolic translator lowers each spec to Dafny stubs and proof obligations, and a driver harness orchestrates an LLM that fills in implementations and proofs, all checked by the Dafny verifier. We evaluate our language and implementation on multiple examples and we show that it is self-hosting: an LLM can implement the Djinnlang translator from its specification and the reimplementation can verify itself.

AI agents that carry a multi-step computer task through on their own became ordinary tools in the past year, and the same autonomy is available to anyone whose task is harmful. We ask what that means for a relying party -- an insurer, a lender, an auditor -- whose evidence is a filed PDF. AgentForge-Bench measures how reliably an off-the-shelf coding agent, driving one of seven open-weight models with a shell and the stock Python PDF stack, alters one dollar amount, date or address in a real filed financial document from a single sentence of intent, graded by rules rather than by a model. Across 1,750 cells, 1,419 (81.1%) satisfy the verifier, and 808 (46.2%) also survive every stricter filter: visible, localized, typeface-matched, original value gone document-wide. A deterministic script with no model in it solves 98 of the 125 documents; the agents solve 124, and none the script solves alone. Agents misreport 41% of their wrong edits as done, no model refused, and the cheapest verified forgery costs 2.4 cents. The raw rate overstates the threat by about a factor of two; the strict rate is still large.

Expressive speech synthesis has advanced through prosody modeling, yet generating structured poetic speech, such as haiku, remains challenging. Prior work on prosody transfer improves expressiveness, and fine-tuned poetry TTS (text-to-speech) systems capture verse intonation. However, these models do not model haiku's 5-7-5 syllable structure or line-ending pauses. We present a cascaded system, HaikuS2S, combining ASR (automatic speech recognition), LLM (large language model)-generated haiku, and TTS fine-tuning on both prose and custom haiku datasets. Our evaluation focuses on emotion similarity, speech quality, and prosody alignment. In our experiments, we see that our prosody and tonal alignment improve significantly with our fine-tuned systems, particularly the one trained on both general poetry and haiku. We also see that we maintain similar emotion similarity scores across all systems.

Cardiovascular magnetic resonance (CMR) provides comprehensive cardiac assessment but remains underutilized because of the complexity of acquisition, post-processing, and interpretation. Existing artificial intelligence (AI) methods address isolated tasks, limiting clinical integration. We present ORION-CMR (On-scanner Reporting with Integrated fOunda-tioN Model), the first clinically evaluated scanner-native end-to-end CMR foundation model. Pretrained on 12,896,733 CMR images from 9,258 studies, ORION-CMR performs sequence classification, ventricular function assessment, late gadolinium enhancement (LGE) detection, binary and multiclass disease classification, and local large language model-based report generation in approximately 90 seconds. The framework. was evaluated on public benchmarks and clinically validated in a multi-vendor cohort of 68 subjects with normal examinations, congenital heart disease, dilated cardiomyopathy, and myocardial infarction. ORION-CMR outperformed supervised baselines and the previously published CMR foundation model (CMR-FM), achieving state-of-the-art performance for LGE classification and scar segmentation. Clinical evaluation achieved an AUC of 0.96 for normal-versus abnormal classification and 0.88 for multiclass disease classification, while generated reports demonstrated 81.4% agreement with expert interpretation. These results demonstrate the feasibility of real-time scanner-native AI-assisted CMR analysis and automated report generation.

In-band full-duplex (IBFD) multiple-input multiple-output (MIMO) systems enable simultaneous transmission and reception on the same frequency band, improving spectral efficiency for next-generation wireless networks. However, IBFD-MIMO systems are susceptible to self-interference (SI), which may overpower signals of interest (SOI). In this scenario, blind source separation (BSS) algorithms can be adopted to remove SI and perform joint sensing and communication (JSAC), but BSS algorithms mostly assume an idealized linear and quasi-stationary signal model, which does not hold under realistic radio frequency (RF) impairments, such as I/Q imbalance, carrier frequency offset (CFO), phase noise, and power amplifier nonlinearity. This paper proposes a two-stage echo state network (ESN)-based scheme that is superior to BSS under these realistic conditions. A frozen ESN is trained offline to characterize the static SI path, while an adaptive ESN, updated online via recursive least squares, tracks the time-varying SOI path using sparse pilot symbols. We evaluate the proposed scheme's SOI recovery performance and acquisition speed with different block sizes, comparing it against other recurrent neural networks (RNN), such as long short-term memory (LSTM) and gated recurrent unit (GRU). Simulation results show that the proposed approach outperforms BSS, LSTM, and GRU in both efficiency and SOI recovery, demonstrating the viability of ESNs for real-time, nonlinear self-interference cancellation in realistic IBFD MIMO systems.

Effective communication between users and AI agents is essential for human-AI collaboration. The XY problem is a well-known communication pitfall where a person asks about their attempted solution rather than their actual problem. We extend prior sycophancy evaluation to the XY problem in agentic settings, evaluating whether agents can resist plausible but misleading suggestions from users and communicate their reasoning. We introduce XYEval, a meta-evaluation framework that can transform an existing benchmark into an XY problem evaluation. We evaluate five models across six diverse benchmark suites. Agents suffer large XY drops under XY mutation across benchmarks, with relative drops reaching up to 46.7%. With $τ^2$-bench, we further show that agent performance drops more when encountering a pedantic user who requires detailed explanations before approving a better solution. Our findings suggest that current agents lack the ability to effectively reason and communicate when facing misleading suggestions. A simple system instruction baseline that encourages awareness of XY problems only offers partial mitigation. Extensive trace analyses provide behavioral insights into how and why these XY drops occur across execution trajectories. Our results show that mitigating the XY problem remains challenging, requiring agents to both recognize user misdirection and clearly communicate the underlying problem.

Robust linear fits can resist response contamination yet remain too dense or unstable for useful global explanations. We propose penalized distillation, which fits a smoothly clipped absolute deviation (SCAD) estimator to a robust initial estimator's empirical fitted surface along a safeguarded coordinate-descent path and evaluates candidate states separately for fidelity, parsimony, perturbation stability, and held-out prediction. The new results attach to the states the algorithm actually computes. Conditional on a fixed uncontaminated design, deterministic bounds transfer response-replacement boundedness from the initial fit to every retained path state. Turning to fixed dimension, we characterize the oracle-support branch by its empirical-Gram projection and influence function, give conditions for covariance-weighted least-squares approximation equivalence, and establish a path-conditional generalized information criterion. By contrast, at large dimension-to-sample ratios the full-coordinate robust fit collapses without warning, and screening restores the construction. Under a sure-screening framework, the robustness bound and the support and selection guarantees transfer to the screened fit. Simulations separate robustness transfer from support recovery, efficiency, and computation across the dimension-to-sample ratio, with p up to 240, and the signal density, which isolates what the sparse stage adds once the screen over-selects. In a duplicate-grouped superconductivity study, the distilled estimator remains predictively stable under prespecified training-response shifts but retains 66.8--68.8 of 81 slopes. Stronger sparsification reduces the model to 12.6--14.0 slopes only at visible fidelity and prediction cost. Distillation therefore preserves predictive stability on these data without substantiating a compact coordinate-level explanation.

Post-training turns a general next-token predictor into a chat model with a persistent assistant persona. If that persona is a character the model plays only on its own turns, its preferences should govern what the assistant says, not what the model predicts other speakers will say. We test this boundary and find that it does not hold: a safety-relevant preference of the assistant---for harmless over harmful tasks---shapes the model's predictions even on the user's turn, where the assistant is not the one speaking. We find that this preference is small or near-zero in pretrained base models, that it emerges through post-training, replicated across open-weight model families, grows with scale, and can be moved by narrow finetuning that never touches user turns. We claim that this is evidence that post-training does not merely install a shallow assistant persona, but instead generalises beyond just the local assistant turn, into the model's representation of the user.

We present a two-part neural framework for forced three-dimensional incompressible Navier--Stokes flow. Part~I develops the computational forcing system. A physics-informed neural model generates structured external-force trajectories, candidates are optimized through differentiable PDE rollouts or PPO-Clip, and selected forcings are frozen and checked by independent fixed-force replay. Part~II provides the mathematical certification layer. It separates neural candidate discovery from continuum analysis, derives integrated reciprocal-vorticity criteria that imply Riccati-type growth and finite-time loss of smooth continuation, develops a validated computational-to-continuum transfer strategy, and establishes a conditional positive-probability closure for a nondegenerate neural output law. The proof is complete at the continuum level.

Density-Ratio Rescoring (DRR) augments a classifier trained at the original class prior with a survey-raking dual score. Raking reweights the majority sample to match minority feature moments within a tolerance. DRR marginally standardizes the dual and base scores and combines them with a fixed weight of one half, using the fitted dual directly for prediction without resampling or refitting the base classifier. Under exact population matching and a correctly specified log-linear tilt model, the dual equals the log density ratio up to an additive constant. A class-separation analysis characterizes the signal strength and correlation conditions under which fusion improves separation under common within-class covariance. On 24 tabular benchmarks, evaluated over 30 trials and five base learners, DRR at the D=128 random-feature setting improves average precision over the standardized base on every dataset, with a mean gain of 0.034. It exceeds the shared-dual raking-and-relabeling resampler on 22 of 24 datasets, with a mean gain of $0.092$, and on all eight one-versus-rest tasks of a shared gene-expression cohort. These results demonstrate the effectiveness of using raking duals as reusable scores for improving rare-class ranking while retaining classifiers trained at the original prior.

We study whether LLMs can produce executable workflow artifacts that remain consistent across graph structure, tool implementation, schema bindings, and runtime wiring. In this setting, correctness depends on cross-layer consistency: a workflow may be structurally plausible, yet still fail because tool implementations, schema bindings, or runtime execution do not align. Existing benchmarks largely evaluate these capabilities in isolation or rely on trajectory-level proxies, leaving open whether generated workflow artifacts execute end-to-end. We introduce \textbf{MCPGen}, an executable benchmark for Model Context Protocol (MCP) workflow development. MCPGen contains 100 self-contained MCP projects across 16 application domains and evaluates three diagnostic tasks: workflow reconstruction, tool creation, and backward-compatible workflow extension. We evaluate 11 representative LLMs in a single-turn foundation-model setting, assessing generated artifacts through static analysis, unit and integration tests, and process-isolated end-to-end execution. Models reach 88.5\% on workflow reconstruction, but no model exceeds 57\% end-to-end execution success. Per-tool unit-test pass rates reach 63.8\%, while project-level integration success does not exceed 45\%, suggesting that integration remains a major bottleneck even when isolated tool tests pass.

Pretrained EEG foundation models are increasingly proposed as general-purpose encoders for brain-computer interfaces, yet recent benchmarks disagree about when their representations transfer to downstream tasks. We audit LaBraM and CBraMod on motor imagery under a validation-locked protocol in which preprocessing, architecture, optimization, freeze depth, checkpoint, temperature, and method selection are determined using training-session data only. On four-class BCI Competition IV-2a, every supervised comparator evaluated here outperforms every foundation-model configuration, including validation-selected fine-tuning. We then examine a key confound: foundation models and task-specific decoders are normally evaluated using different input pipelines. Retraining three supervised architectures on the broadband arrays consumed by the foundation models produces matched-input accuracy differences of opposite sign across architectures: broadband input improves ATCNet by 0.078 accuracy while reducing EEG Conformer accuracy by 0.088. None of the three individual matched-input terms is significant after multiple-comparison correction at n = 9, so we treat the sign variation descriptively rather than as a formal architecture-by-pipeline interaction. These observed sign differences suggest that a single comparator may not provide an architecture-invariant decomposition of a pretrained-versus-supervised performance gap. The four-class deficit also does not reproduce uniformly across motor-imagery datasets: on two-class BNCI2014-004 we cannot detect the same separation between fine-tuned CBraMod and the supervised comparators. Finally, validation-fitted temperature scaling returns foundation-model calibration error to the supervised range despite substantially lower four-class accuracy.

Chess involves complex reasoning in a deterministic environment, which makes it a useful setting for studying the mechanisms of computation inside transformers. The Maia-3 chess transformer takes Elo, a measure of competitive chess skill, as an input to the pre-trained network, so we can vary the skill the network is conditioned on with no change to its weights. Here we investigate how turning this skill dial affects self-attention. Ablating every attention head at every Elo from 700 to 2500, we find 1) increasing skill pushes the causal center of mass of the computation deeper, monotonically, for every chess piece and move type we measured; 2) the depth migration is much greater for specific tactics, especially knight forks, than for other move types; 3) the migration consists of deeper heads getting recruited for more specialized computations while one shared shallow head keeps a roughly constant contribution. These results may shed light on how conditioning inputs redistribute computation in larger transformers.

Large language models (LLMs) drift out of date the moment their pretraining ends, yet retraining from scratch is prohibitively expensive. Continued pretraining (CPT) is the natural remedy, but it is typically evaluated through a continual learning lens that assumes disjoint data streams. This is a poor fit for time-incremental updates on web-scale crawls, where successive snapshots share substantial URL overlap by design. We study time-incremental CPT in this realistic regime: continued pretraining on FineWeb-Edu dumps drawn strictly from after each model's knowledge cutoff, evaluated across six open-weight models spanning three families (OLMo2, Llama-3.1/3.2, Gemma-3-1B) and four parameter scales (1B-3B-7B-8B). We organize our findings around four practical questions. (i) Is knowledge acquired? Yes, but heterogeneously, and without catastrophic forgetting: five of six models also improve on pre-cutoff factual recall, and the gains track pretraining saturation (driven primarily by token budget per parameter). (ii) What does it cost? Almost nothing: the macro-average across a thirteen-task suite stays within 0.01 of the base for every model. (iii) What is the recipe? Data quality dominates quantity (a curated 6B-token slice matches a broader 40B one); the optima for knowledge acquisition and general capability are separated by roughly an order of magnitude in learning rate; and LoRA at sufficient rank matches full CPT. (iv) Does it survive deployment? CPT gains transfer through SFT, while DPO's effect is family-dependent. Together, these results paint a more optimistic picture of time-incremental CPT than the prior continual learning literature suggests.

This paper presents an end-to-end framework for reconstructing overhead power utility network topology and extracting span-level physical metadata from large-scale aerial LiDAR. The pipeline begins with semantic segmentation of the input point cloud using an improved KPConv-based model, in which data sampling and loss functions are adapted to emphasize pole and conductor (wire) classes. Network topology inference then proceeds in two stages: (i) pole instances are obtained by clustering pole-class points and validating candidates using geometric criteria, including height and verticality estimated via PCA, and (ii) candidate pole pairs are evaluated using a heuristic method and a lightweight ResNet-based classifier on 2D top-view projections of pole and wire point distributions to determine whether a physical conductor span exists. By explicitly classifying candidate spans, the approach mitigates common failure modes of heuristic connectivity rules in dense or cluttered scenes and under partial wire observation. For each validated wire, attributes regarding utility infrastructure geometry are computed, including endpoint conductor heights, ground elevation, sag-related lowest-point features, conductor arrangement, and wire width. Evaluation on multiple real-world aerial LiDAR datasets demonstrates decimeter-level endpoint height accuracy and approximately 9% relative improvement in recall for topology reconstruction compared to heuristic nearest-neighbor baselines, with larger gains in complex layouts.

Simulation-based evaluation provides a scalable and repeatable alternative to real-world evaluation of vision-language-action (VLA) policies. However, reconstruction errors can cause simulated policy performance to diverge from real-world performance, motivating the need to assess reconstructed environments for downstream VLA policy evaluation. We present ReVeal, a real-to-sim assessment framework combining workspace reconstruction, reconstruction-level assessment, and matched closed-loop policy evaluation. Novel-View Mesh Fidelity (NVMF) and Annotated Planar Geometry Fidelity (APGF) assess observation and planar geometric fidelity, respectively. We also develop PGSR-D, a reconstruction pipeline incorporating monocular depth supervision to improve geometry where multi-view visual cues are limited. Across 8 assessment scenes, NVMF and APGF consistently distinguish the fidelity of 2DGS, PGSR, and PGSR-D. Matched evaluations of GR00T, SmolVLA, and pi0.5 across 8 humanoid manipulation tasks show consistent ordering between reconstruction fidelity and real-sim performance agreement across pipelines. Further analysis of the evaluation workspaces shows that higher fidelity is associated with stronger real-sim agreement.

Generative models of neural activity could help characterize tissue dynamics, compare experimental conditions, and simulate population activity for applications ranging from disease and drug-response studies to closed-loop experimentation. Existing approaches, however, typically assume a fixed set of sorted neurons, whereas high-density microelectrode arrays produce extremely sparse, array-wide binary spike volumes in which the observed subset of electrodes varies across assays. We introduce a discrete generative model that represents this activity using a shared vocabulary of spatiotemporal motifs. A residual vector-quantized autoencoder learns the motif vocabulary, while a factorized masked transformer predicts where activity occurs and which motif appears at each active location. We evaluate the model on 31 assays spanning human brain organoids and acute \emph{ex vivo} human hippocampal tissue. The learned motifs are broadly reused: assay identity explains only $9%$ of the entropy in motif use, and motif overlap across tissue types is comparable to overlap within them. When representation quality is evaluated independently of the generative prior, our approach achieves $5.2\times$ the voxel-level reconstruction average precision of a matched flat tokenizer. For masked completion and free generation, the full model achieves $1.4$--$2.6\times$ the site-level average precision of the matched generative baseline and outperforms it across all four families of generation metrics. These results establish a compact, reusable representation for array-wide spiking activity without learned assay-specific parameters, providing a scalable foundation for generative modeling across diverse neural preparations.

This paper introduces a novel algorithm for predicting conditional joint distributions of vector-valued targets in stochastic systems whose randomness is intrinsic rather than arising from observation errors or additive noise. Multivariate quantile regression also involves modeling conditional joint distributions but represents a less challenging task. It predicts the probability that vector-valued targets fall within predefined regions, identifies regions corresponding to predefined probability levels, or performs both tasks simultaneously. The proposed identification technique employs ensembles of Kolmogorov--Arnold networks (KANs) as flexible function approximators. Although the suggested technique is not theoretically restricted to KANs, KANs are particularly well suited to the proposed construction and are therefore used throughout this study. In addition to the training procedure, this work introduces a new discrepancy measure for joint distributions and a goodness-of-fit (GoF) test based on it. This GoF test was initially developed to validate and calibrate the proposed identification technique and is used here in an ad hoc manner. Although the test could be tabulated for broader use, such a tabulation is not pursued in this work. The test is also applicable more generally.

Tree speculative decoding verifies multiple candidate continuations in one target forward pass. For attention-only transformers, the verifier mainly needs an ancestry mask. Recurrent-hybrid language models break this assumption: a candidate row must also carry the recurrent state that native sequential decode would have produced along its root-to-node path. Otherwise, a verifier can use a correct attention mask while still conditioning on an impossible recurrent history. We present GDN Tree-Scan, a served verifier for Gated-DeltaNet hybrid language models integrated into vLLM. The system combines FlashAttention-2 tree-bias attention, branch-local GDN scan/replay, device-side multidraft commitment, and accepted-chain-only state publication. On the public Qwen3.6-27B-FP8 checkpoint, in a clean batch-one (B=1) SWE/Codex decode gate at temperature 0.6, a six-node root-branch tree increases committed tokens/event by 17.2% at near-native verify-forward time and reaches 23.88 token-weighted decode tokens/s versus 18.80 for native five-step MTP (E5), a 27.0% token-weighted decode-throughput gain. The per-request-equal latency view is +4.0%, and end-to-end task wall time remains prefill-heavy. Empirical equivalence evidence is scoped to recurrent-oracle probability-rescore (p-rescore) closure within the observed native flip floor, not a full distribution-distance proof.

Agentic AI extends LLM security beyond generated content to persistent state, autonomous actions, tool use, and interactions with humans and other agents. Existing threat classifications often emphasize individual dimensions, obscuring connections among entry points, affected components, and security consequences. The known threat landscape also differs from the coverage demonstrated by empirical research. Through a structured review of 66 studies published from 2022 to 2026, we introduce T={S, B, P, A}, a cross-dimensional representation linking affected functional or system surfaces {S}, interaction or trust boundaries {B}, violated security properties {P}, and empirically examined architectures {A}. We analyze 22 artifact-backed red-teaming studies and 11 representative security benchmarks to characterize empirical coverage and evaluation maturity. Within the selected studies, evidence concentrates on prompt/reasoning, memory, and tool-mediated attacks, predominantly in single-agent settings. Persistent, Human--Agent, complex multi-agent, systemic, and long-horizon threats receive less coverage. These findings describe the selected corpus rather than establish gaps across all empirical research. Heterogeneous metrics, limited adaptive defense evaluation, architectural imbalance, and incomplete execution-state capture further constrain comparison and reproducibility. We derive 13 open research questions to guide more systematic, architecture-aware, and reproducible security evaluation of agentic AI.

Mechanistic interpretability defines features as the fundamental units of a neural network and circuits as the weighted subgraphs that carry out its computation. Because individual neurons are polysemantic, Cross-Layer Transcoders (CLTs) were introduced as a way to approximate a model's circuits by generating an attribution graph. The nodes of that graph, however, are unlabeled features: reading a graph means pruning it and then working out by hand what each surviving node means. To make CLTs easier to use for circuit discovery, we introduce Circuit-Diff, which intervenes on the model itself with a low-rank factual edit and takes the features whose role in the attribution graph changes under that edit as related to the edited knowledge. On the edits we examine, the flagged nodes are not only detectors of the object token: read off the CLT's released feature dashboards, they include features for the history, geography and associations surrounding the old and new objects. We formalize the method, measure how reliable a frozen CLT remains after a factual edit, test the selected nodes causally by patching them on up to 24 CounterFact edits, give a case study, and release an open-source implementation built on the circuit-tracer package, together with two further tools (multi-prompt aggregation and rule-based supernode labeling).

Automated kernel vulnerability reproduction is essential for bug triage, patch validation, and regression testing, but still lacks an effective and efficient solution. The core challenge is twofold: a reproducer must first recover the trigger scaffold needed to reach the vulnerable state and determine the precise concrete values that actually trigger the bug. Existing directed fuzzing approaches are ineffective at recovering the necessary trigger scaffold, while LLM- only generation is brittle because it struggles with concrete-value discovery and runtime nondeterminism. We design SyzHarness, a framework that combines LLM reasoning with coverage-guided fuzzing for patch-based Linux kernel vulnerability reproduction. Given a patch, SyzHarness uses an LLM agent grounded by code navigation tools to synthesize a parameterized fuzzing harness that fixes the prerequisite setup logic while exposing only uncertain, bug- critical input parameters to be mutated by Syzkaller. SyzHarness then translates this harness into a Syzkaller- compatible interface and iteratively refines it using hierarchical reachability feedback. We evaluate SyzHarness on multiple datasets of triggerable real-world Linux kernel vulnerabilities. On 100 KernelCTF cases, SyzHarness achieves a 78% bug reproduction success rate. On the SyzDirect benchmark, SyzHarness achieves a 73% bug reproduction success rate, substantially outperforming prior directed greybox fuzzing. On 50 recent, known-triggerable syzbot bugs fixed after March 2026, SyzHarness reproduces 40/50 (80%) using only the fix commits as input.

Software delegates more of its branches to models every year: which queue a ticket enters, whether a command is safe to run, whether a claim clears without a person. What the program needs back is not prose. It is one of n declared options and a number it can threshold. Today that costs a round trip to a frontier model -- hundreds of milliseconds, a per-token bill, and a parser -- for a question that is usually a conjunction of three clauses. this-that-model-1.0 is a 2B-parameter typed decision model. Its answer is read directly from the hidden state at a designated position and restricted to the option set the caller declared, so no text is generated, nothing can be malformed, and every question in a request is answered in the same forward pass. It decides in 30.9 ms on one laptop GPU and generates zero output tokens doing it, where a frontier API call costs 8758 ms and the hosted systems that answer these questions well spend between 21 and 212 generated tokens per question thinking first, billed for every one. It sustains 32 decisions per second on one consumer GPU and never lets the state leave the machine. On a third party's recorded cohort of 68 decision questions, on their inputs and their wording, it scores 0.941 with a Brier score of 0.042, against 0.765 and 0.133 for the hosted service Jev on the same items. One pass of our 42-family internal suite takes 32 seconds and 0.000217 USD of electricity; the most accurate hosted model we measured needs 155.2 minutes and 10.636 USD. We also report where it loses. On multi-step arithmetic, which a single forward pass cannot carry intermediate results through, it scores 0.560 against their 0.98 to 1.00, and a targeted second training round improved the five task families it was written for and transferred to none of the other 13. The model is open-sourced in https://huggingface.co/flock-io/this-that-model-1.0

Robot foundation models learn manipulation from large demonstration corpora, but surgery is missing from those corpora: across the 780-hour Open-H surgical collection, one dataset carries synchronized force and none covers an aesthetic procedure. Liposuction is the hard case, because the instrument works under the skin and the surgeon operates by feel and by judgment. Humynex Robotics builds curated expert datasets for this kind of procedure. HumynexSurg-1 is the first release: a master liposuction surgeon performing on porcine abdominal tissue while narrating every decision, recorded with synchronized suction pressure, six-axis hand force/torque, top-down RGB-D video, side video and a lavalier microphone -- 14 episodes, 42,738 frames, 35.6 minutes, 356 utterances of which 95% compile into a liposuction-specific label schema. The capture follows a patent-pending sensing plan organized around the quantities a policy needs, so a channel captured today by a model can be upgraded to a sensor tomorrow without changing the data format. This release captures the instrument motion as a tool-hand track in the side video and provides the force channel as state; the funded capture adds a measured 6-DoF handle pose, a validated force channel, ultrasound imaging of the fat layer, and palpation sensing. As a proof of concept, NVIDIA Isaac GR00T N1.7 fine-tunes on the dataset with no custom code in under an hour per run and learns the recorded sessions; scaling probes on the same episodes show where further gains come from: every new session lowers the error on an unseen session. The dataset, its label schema, its quality-assurance reports and its evaluation protocol are the product; the next capture, many short sessions across fat regions with the sensors named here, is what the probes point to.

We present a collaborative streaming anomaly detection system for high-speed data streams that explicitly integrates human analysts into the decision loop. The system combines heterogeneous detectors and aggregates their outputs through a normalization-based weighted consensus, complemented by artifact-aware rules to stabilize anomaly scoring under deployment. To improve interpretability, it derives surrogate models that approximate the ensemble consensus and expose human-readable sensor conditions associated with anomalous behavior. Analysts can actively intervene by reviewing anomaly episodes, adjusting consensus behavior, and refining surrogate rules used for anomaly prediction, producing a human-adjusted ensemble. We evaluate the approach on an industrial stream with 260\,000 events and 3 anomalous episodes, showing robust detection and actionable human-AI interaction.

Temporal Information Retrieval (TIR) has been increasingly critical given the rise of Retrieval-Augmented Generation (RAG). Since temporally mismatched evidence can be highly misleading, TIR aims to retrieve documents that are both semantically and temporally relevant to a query. Two TIR paradigms have emerged - temporal retrievers and temporal re-rankers - differing in how temporal relevance is modeled. While these paradigms provide complementary strengths, our analysis reveals that each alone falls short of robust TIR: temporal retrievers provide flexible query understanding via learned representations, but often fail to explicitly account for temporal constraints; temporal re-rankers can enforce such constraints more explicitly, but often rely on predefined re-ranking rules. To address this, we propose Q-TIE, a re-ranking framework based on learned Temporal Intent Extraction (TIE). By introducing a TIE model that maps each query's temporal constraint into a unified interval representation (i.e., $\langle t_{start}, t_{end} \rangle$), Q-TIE generalizes beyond predefined rules via model-based learning while explicitly modeling temporal constraints as a separate signal - jointly achieving what each paradigm typically trades off. Experiments demonstrate that Q-TIE consistently outperforms existing TIR methods with stronger generalizability across temporal query types, and provides a lightweight yet effective add-on for temporally-aware RAG pipelines. Code: https://github.com/ssoy0701/Q-TIE.

Modern music streaming platforms face a persistent tradeoff: exploiting familiar content versus driving the exploration of novel items. While users frequently desire discovery, they hesitate to select unknown artists over proven favorites. Providing transparent, natural language rationales that explain why an unexplored item is recommended lowers this barrier. However, while Large Language Models (LLMs) excel at this nuanced explainability, their real-time deployment is severely bottlenecked by prohibitive inference costs and computational overhead. In this paper, we present an industry case study of a decoupled recommendation architecture that successfully scales exploration without compromising latency. Our system isolates LLM inference asynchronously offline, pre-computing personalized candidate pools of undiscovered artists alongside tailored rationales. Large-scale online A/B experiments validate our design. We demonstrate that combining LLM-backed recommendations with these explanatory rationales significantly reduces the trust barrier for new content, yielding statistically significant improvements in both user exploration and overall engagement on the discovery surfaces.

Clinical multimodal models must often predict before all chest X-ray (CXR) and electronic health record (EHR) inputs are available. Existing approaches align observed representations, model missingness, or reconstruct across modalities, but do not jointly exploit within-patient and clinically similar inter-patient evidence. We propose GLR-MM, a Graph-Based Global-Local Reconstruction framework for early ICU mortality prediction. It maps five CXR-EHR modalities to a shared space, reconstructs missing embeddings through complementary local cross-modal and global graph-attention branches, adaptively fuses their estimates, and optimizes class-balanced prediction, reconstruction, and contrastive objectives. On 9,620 MIMIC-derived ICU stays, we evaluate 10%, 30%, and 50% random modality missingness with shared deterministic masks. MUSE performs better under mild and moderate missingness, whereas GLR-MM achieves higher AUROC and AUPRC at 50% by 0.0088 and 0.0249, respectively. These results indicate that graph-guided reconstruction is most useful when inputs are severely incomplete.

The rapid growth of resident space objects is increasing the complexity of space situational awareness sensor tasking, challenging classical optimization methods as they allocate finite, heterogeneous, and distributed sensing resources across ever-larger catalogues. Existing deep reinforcement learning approaches show promise in reduced settings, but fixed-dimensional state and action representations limit their ability to scale to large, dynamic catalogues and distributed sensing networks. We introduce VISTA (Variable-Entity Intelligent Sensor Tasking Architecture), a scalable deep reinforcement learning architecture for persistent uncertainty-driven catalogue maintenance across variable object populations and sensor configurations. VISTA combines physics- and mission-informed top-K retrieval with entity-centric attention, recurrent memory, and pointer-based action decoding, thereby keeping each agent's observation and action spaces independent of catalogue size. We evaluate VISTA across different scenarios, from fixed-size single-sensor benchmarks to large-scale space-based tasking and heterogeneous cooperative sensing. With 30 orbiting targets, VISTA recovers the catalogue 31.2% faster than the fixed-dimensional recurrent baseline. In the large-scale regime, VISTA reduces five-hour uncertainty by 97.5% relative to the strongest classical reference and by 99.3% relative to the recurrent learner. Zero-shot tests up to 20,000 objects reveal near-linear relations between sensing capacity, catalogue size, and recovery horizon. Learned policies also exhibit sensor modality adaptation and generalization to population and initial-uncertainty shifts. Together, these results demonstrate that VISTA provides a scalable framework for adaptive space situational awareness sensor tasking across large, distributed networks of heterogeneous ground- and space-based sensors.

Multimodal Large Language Models (MLLMs) commonly reuse visual encoders pretrained with CLIP, although the features of these ViTs are ultimately consumed by autoregressive LLMs. We refer to this mismatch as the semantic-interface gap and introduce MedMLIP, a framework that pretrains the visual encoder through report generation with a frozen LLM, while employing Local Relational Distillation (LRD) to preserve relationships among visual patches to avoid visual collapse. We pretrain MedMLIP on IU-Xray and Open-PMC-300K and evaluate the resulting encoders on VQA-RAD and SLAKE. Only the ViT is transferred, while the guiding LLM and projector are replaced, allowing us to assess cross-LLM transferability. Our cross-LLM transfer experiments demonstrate the value of pretraining visual encoders for their autoregressive LLM interface while trying to preserve more fine-grained visual information. Code and the pretrained model are available at https://github.com/SkyCol/MedMLIP

Disaster-risk-reduction archives describe hazard events in prose that databases such as EM-DAT (Delforge et al., 2025) cannot ingest directly. We present an LLM pipeline that generates candidate georeferenced event records using a controlled hazard vocabulary and fixed schema, retaining evidence for review. Applied to 10,000 documents from PreventionWeb, the knowledge hub managed by UNDRR, it produced 3,572 records from 1,913 documents across 24 hazard types and resolved 81% of location mentions to OpenStreetMap geometries. On 171 human-positive document windows from a stratified 217-document reference set, GPT-5 achieved 86.0% pooled attribute $F_1$, versus 44.2% for the spaCy-gazetteer baseline. Evaluation pools hazard families, location strings, and event years within documents, without assessing their assignment to individual events. GPT-5.4 ranked highest among ten LLMs (86.6% $F_1$). Verbatim evidence occurrence was 72.0% for GPT-5 and 47.2% for GPT-5.4, measuring textual traceability without establishing attribute support. We report production failure modes and automated label and location-rule compliance checks. Prompts, schema, and outputs will be released for adaptation to national reporting archives.

Decentralised federated learning removes the aggregation server but makes collaboration dependent on transient peer availability. In mobile and intermittently connected systems, evaluating a promising peer consumes contact time and may cause the exchange opportunity itself to vanish, so that the evidence a learner gathers about a peer is perishable: it decays because links expire and because peer models drift while old measurements age. This paper develops a self-contained theory of optimal stopping for the resulting peer-selection problem. We formalise a receiver's within-contact decision as a finite-horizon Markov optimal-stopping problem with costly information acquisition and a future-arrival outside option, and prove that it admits an optimal policy characterised by a reservation value (Snell-envelope structure). Around this formulation we prove: (i) stage-uniform, drift-aware concentration and a maximin certification rule that is correct with high probability together with a finite-sample identification bound; (ii) a mobility-aware value of-information stopping rule and comparative statics showing that higher link hazard lowers the value of continued probing and enlarges the stopping region; (iii) a closed-form value of waiting under marked-Poisson contact arrivals, together with a search-theoretic reservation value whose comparative statics we characterise; and (iv) a myopic-optimality theorem establishing that, in sufficiently volatile (monotone) mobility regimes, the one-step confidence-safe rule is a sound surrogate for the optimal policy and never stops prematurely. We instantiate the theory as PROSE (Perishable-evidence Reservation-value Optimal Stopping for Exchange), a lightweight, fully local policy, and delineate the static contact and drift-free limits in which classical sequential decision problems are recovered. The development is entirely analytical.

Scheduling clients for model training is critical in federated learning due to both data and system heterogeneity. Most previous works focus on the quality of the scheduled clients to achieve faster convergence, shorter wall-clock convergence time, or better average model performance. They rarely consider the diversity of clients, which is important to counter heterogeneity and improve performance for the worst-off clients. In this work, we advocate the use of determinantal point processes (DPPs) to model and enhance the diversity in client scheduling. We first design the kernel matrices of DPPs using gradient information and quality scores, which inherently enables a flexible quality-diversity trade-off. Applying fast MAP inference over DPPs, we propose Adaptive Determinantal Client Scheduling (ADCS) in FL. We further quantify the gradient approximation error of ADCS and develop convergence analysis for general biased client selection in FL with non-convex loss functions. We conduct comparative numerical experiments showing that ADCS outperforms state-of-the-art client scheduling algorithms, including both quality-based and diversity-based ones.

Greenhouse gas emissions estimates can be derived using inverse methods by combining atmospheric concentration observations with chemical transport models. The latter traditionally use physics-driven simulators such as Lagrangian Particle Dispersion Models (LPDMs), which are expensive to run and do not scale well to modern satellites' high resolution data. Previously we developed a performant atmospheric transport emulator that approximates LPDM outputs ("footprints") over South America ~1,000X faster than the UK Met Office's LPDM. Expanding towards global emulation is not straightforward, as atmospheric transport is regionally heterogeneous. This paper evaluates spatial transferability capabilities of models across four world regions: South America, East Asia, South Asia, North Africa using both region-specific and multi-region models, and leave-one-region-out experiments. Regional differences are characterised in the context of input variable and output footprint distributions. This work builds intuition in cross-region generalisation and transfer learning, aiding regional performance towards efficient global emissions estimates.

We study the complexity of finding $(δ,ε)$-Goldstein stationary points of nonsmooth nonconvex Lipschitz functions. By now, it is known that randomized first-order algorithms can solve this task with a dimension-free oracle complexity [Zhang et al., 2020], whereas deterministic algorithms cannot, as their complexity must scale at least linearly with the dimension $d$ [Jordan et al., 2023, Tian and So, 2024]. This leaves open whether deterministic algorithms can nevertheless solve the problem with oracle complexity polynomial in $d$. We answer this question negatively by proving a lower bound of order $(1/ε)^{Ω(d)}$ for deterministic algorithm, closing the exponential gap between the previously known lower and upper bounds and resolving an open problem posed by Jordan et al. [2023]. We further discuss several extensions and implications of this result to weaker stationarity notions, finding a descent direction and deterministic smoothing. Overall, our results establish an exponential computational advantage in nonsmooth nonconvex optimization offered by randomization.

A fundamentally challenging question in K-12 education is about the effects of taking more advanced or challenging classes. It is particularly complex because students (and/or their parents) choose whether to enroll in these classes, making causal analysis challenging. In this paper, we begin to tackle this question by taking advantage of a novel dataset from a public school system in the US. This dataset records students' course enrollment decisions, prior academic histories, demographics, and subsequent outcomes around the time of a district-wide change that introduced optional open-enrollment advanced middle-school courses in subject areas. This is a rich observational dataset, but enrollment in advanced classes is driven by student characteristics and choices rather than random assignment. This creates a core identification challenge: the same factors that influence enrollment in advanced courses are also predictive of academic outcomes. As a result, simple comparisons between enrolled and non-enrolled students are confounded, and naive estimates may reflect underlying differences in student ability, motivation, or support rather than the impact of coursework itself. Our analysis shows that enrolling in advanced English courses has a net positive but modest effect on student achievement outcomes. However, these benefits are unevenly distributed: some students with relatively large predicted gains ("middle achievers" in prior years) are less likely to enroll than others. Some other groups (e.g. Black students and those with lower socio-economic status) also demonstrate significantly lower propensity to enroll. This gap between predicted benefit and observed enrollment illustrates how careful data analysis can extract actionable insights from large observational datasets, including identifying students who appear well-positioned to benefit but do not select into advanced options.

Mitral regurgitation is the most common heart valve disorder worldwide, affecting over 2% of the global population, rising to at least 10% in adults over 75, and causing approximately 15% of valvular heart disease-related deaths. Yet only a minority of patients with severe disease undergo corrective surgery. Rapid assessment of valve mechanics could enable earlier, more precise intervention, but traditional finite element simulations remain too slow for clinical timelines and parameter sweeps. We introduce the Physics-Conditioned Neural Operator (PCNO), a transformer-based surrogate that predicts leaflet displacement, strain, and stress fields across mitral and tricuspid geometries, conditioned on systolic blood pressure and tissue properties. Trained on functional, regurgitated, and pathological valves, including tethering, P2 prolapse, and annular dilation, PCNO achieves up to a 15,260x speedup over fine mesh finite element simulations with comparable accuracy, identifies pathology class, and resolves diagnostic metrics within 3.5% error under out-of-distribution extrapolation.

We present a comprehensive benchmark of Federated Learning (FL) for multilingual Automatic Speech Recognition (ASR), evaluating four Speech-LLM architectures on the Multilingual LibriSpeech dataset. We compare FedAvg and FedProx across frozen and unfrozen encoder configurations, demonstrating that optimized learning rates are critical for performance. Specifically, independently tuning the learning rates for the speech encoder, connector, and decoder yields the lowest error rates, with full three-component adaptation (LoRA for encoder and decoder, full training for the connector) producing the best FL results. We observe that FedProx efficacy is architecture-dependent, providing notable advantages in multilingual pre-trained architectures (e.g., EuroLLM over TinyLlama when keeping the encoder fixed); this indicates that LLM backbone capacity plays a key role in mediating resilience to heterogeneous data distributions. These findings offer concrete design guidance for deploying multilingual Speech-LLMs in privacy-sensitive, distributed environments.

The Generalized Naive Bayes (GNB) model was introduced for discrete and categorical random variables as an extension of classic Naive Bayes. We now accommodate the GNB framework to continuous explanatory variables. A central result of the paper is that structure learning of the GNB depends only on the pair copulas of the bi-variate marginals. We proved that the GNB structure can be assigned to the basis of a matroid, therefore we give greedy algorithms for finding the optimal GNB structure on the training data, in sense of minimizing Kullback-Leibler divergence. Three cases are considered: joint Gaussian distribution, then a more flexible model where we suppose the dependence structure to be described by a Gaussian copula with arbitrary marginals, and an even more flexible case where the joint continuous probability distribution is arbitrary, i.e. copula and marginal distributions are arbitrary. A method for model reduction, based on the newly introduced concept of GNB forest is given. We close the paper by comparing the newly introduced GNB classification results to other classical "glass-box" algorithms on real datasets.

Dictionary learning seeks to recover an unknown dictionary $A$ from observations ${\bf y}_i = A{\bf x}_i$ with sparse coefficient vectors ${\bf x}_i$. We introduce the \emph{Iterative Atom Refinement} (IAR) algorithm, a simple procedure for recovering individual dictionary atoms. Starting from a random direction, IAR repeatedly selects the observations most strongly correlated with the current iterate and updates the direction by averaging the selected data. Our main contribution is a rigorous convergence theory of IAR. Using high-dimensional probabilistic estimates and a novel monotonicity principle for atom-selection probabilities, we show that a small initial advantage of one atom is amplified until that atom is isolated. Under our model assumptions, IAR identifies a generating atom after only three refinement steps. Numerical experiments support the theory and show that the resulting dynamics accurately capture the behavior observed in dictionary refinement.

Coding agents are extended by plugins: installable bundles that ship skills, sub-agents, commands, hooks, and tool servers. On 24 July 2026, an open specification (Agent Plugins v1.0.0) standardised how such a bundle is laid out and described, so that one plugin could run on any agent. We ask the two questions a practitioner would ask of it: is the ecosystem adopting the standard, and if a plugin did conform, would that be enough to make it work alongside the other plugins a user has installed? We answer both by building AgentPluginZoo, a provenance-tracked corpus of 68,072 plugin bundles across 30,655 repositories, released with its discovery ledger, scoring code, and analysis. Only 6.2% validate, but the gap is shallow rather than structural: 96.6% would load after adding one missing boilerplate field. The real cost lands elsewhere. 40.2% would load while the specification obliges the client to discard fields their authors wrote, mostly declarations of what the plugin ships. Moreover, conformance settles nothing for the second question: 81% of capability-exporting bundles share a name with another plugin, with no namespace or precedence rule to decide which one answers. This paper argues the community standardised a packaging format when composition needs a model, names the four concepts such a model must add - qualified capability identity, a declared capability surface, a precedence rule, and inter-plugin relations - and shows they fit an additive v1.1 profile of the same specification rather than a competing standard. Recommendations follow for practitioners packaging extensions today and for the people evolving the standard, chief among them that conformance must be made observable before it can become common. The corpus and code are available at https://github.com/tezansahu/agentpluginzoo

While test-time scaling enhances Large Language Model (LLM) agents in long-horizon software engineering (SWE), sparse binary rewards (Pass/Fail) create a severe credit assignment crisis and waste failed exploratory trajectories. Current trajectory optimization and scaling methods are costly and structurally limited, relying on heuristic state reuse without causal diagnosis or delayed scalar scoring without actionable online guidance. We propose FLARE (Full-Lifecycle Alignment and Reward Engine), a novel dense supervision paradigm driven by a lightweight Generative Reward Model (GRM). First, RADAR, an offline causal-aware diagnostic framework, extracts high-fidelity, hindsight-free supervision through causal-chain backtracking to distill a GRM providing real-time, step-level risk feedback. Second, FLARE uses this GRM to continuously optimize the agent across its entire lifecycle. During inference, FLARE acts as an Active Scaffold, autonomously intercepting high-risk generation steps for localized breakpoint re-execution, drastically reducing compute overhead. During post-training, the GRM's structured signals serve as process-supervised reranking scores for Supervised Fine-Tuning (SFT) and step-level dense rewards for Reinforcement Learning (RL), mitigating policy collapse in sparse environments. Extensive evaluations show that FLARE establishes a new Pareto frontier across the agent lifecycle: FLARE (N=1) outperforms Global Rollout (N=5) with a 5x reduction in token consumption. Extending FLARE to training overcomes the sparse reward problem in long-horizon interactive tasks, delivering relative performance gains of 19.13% in SFT through process-aware data curation and a consistent 9.19% improvement in RL.

Many knowledge-work benchmarks are constructed around individual tasks, with the context needed for each task selected together with or after the task has been specified. This design measures performance on workplace-like tasks in an environment assembled for the task. When task specification guides which context is selected, the evaluation can encode task information into the environment and pre-complete part of the information-localization work that workplace performance normally requires. We introduce WorkWorlds, an evaluation infrastructure that separates organizational state from task specification. A world first fixes a revision, date, and employee seat and materializes the organizational state that employee can access; tasks are introduced only afterward. We implement WorkWorlds in a primary synthetic pharmaceutical company with 8 measured tasks across 6 employee seats, and construct additional organizational worlds. Across 192 matched evaluations, moving from task-curated context to the full role-visible workplace reduced evidence access from 90.4% to 74.5% and criterion pass from 79.4% to 68.2%, while pass conditional on evidence access remained nearly unchanged; most of the measured difference occurred before the agent reached sufficient evidence.

Every node in a multi-agent large language model (LLM) workflow retrieves context from memory and injects it into its prompt, where those injected tokens are billed as input tokens at the same per-token price as the system prompt and the user query. Production observability tools report total token cost but do not separate the tokens a node generates from the tokens it is handed, so this component of the bill is invisible to the teams paying it. We introduce the Total Cost of Agency (TCA), a decomposition of multi-agent workflow cost into base prompt, inference, memory injection, miss penalty and context-accumulation components, and an exact attribution method: a two-pass, non-billable token count that measures injected tokens directly rather than estimating them from word-count proxies. On a 200-task enterprise benchmark executed against real model APIs, memory injection accounts for 13.6 percent of the variable cost a compile-time optimizer can act on, about 12 percent of the full billed cost, and its share rises from a structural zero at workflow depth one to 27.6 percent at depth six. Injected tokens grow linearly with depth over the measured range (R^2 = 0.9974, depths two through six); a quadratic fit yields a negative leading coefficient, so the data do not exhibit convex growth at these depths. We show the component is controllable at fixed model tier: reducing the retrieval window capacity from 32 to 2 entries lowers injected tokens by 28.7 percent with an accuracy change within seed-level variation. We report in full that our graph-rewriting transforms are approximately cost-neutral in isolation, that two of the five decomposition terms are zero by construction in this harness, and that total workflow cost is dominated by model tier assignment, which we hold fixed and treat as prior work. Prompt caching is not evaluated; all figures are for the uncached case.

Modern conditional generative models face significant challenges when learning complex covariate dependencies. While sufficient dimension reduction (SDR) provides a principled approach to compress these dependencies, traditional SDR frameworks were not formulated for conditional generation. To bridge this gap, we propose Belted Engression, a unified and architecturally parameter-efficient framework for generative distributional regression. Our approach establishes an end-to-end compress-then-generate paradigm driven by sufficient representation learning, embedding a structural bottleneck into the generative architecture. Theoretically, we prove that the standard SDR condition is equivalent to a law-preserving generative factorization, which is achieved at the global optimum of the population Belted Engression objective. Furthermore, by uncovering a localized Bernstein-type control for the energy-score loss, we establish finite-sample convergence rates that are sharper than those of existing results. We also prove that this belted architecture is strictly smaller, operating with an asymptotically vanishing parameter count relative to the unstructured baseline. Extensive simulations and real-world applications demonstrate that Belted Engression achieves superior distributional prediction and SDR recovery with fewer trainable parameters.

Many real-world decisions require prioritizing high-risk cases, such as clinicians prioritizing high-risk patients before lower-risk ones. Falling rule lists (FRLs), which are ordered if--then rules with monotonically decreasing risks, provide an interpretable framework for such tasks; however, their single-path structure yields a highly restricted model class. We introduce falling trees, a new family of interpretable models that enforces the same monotonic risk constraint while permitting tree-structured branching. We present GRAVITree, a novel dynamic-programming-with-bounds algorithm for learning the Rashomon set of falling trees under depth and branching constraints. Our formulation can interpolate between rule lists and full decision trees, enabling user-desired model expressivity. In a new clinical dataset and in many public classification benchmarks, falling trees match or outperform FRLs and other interpretable baselines, often producing more sparse decisions for high-risk instances. Our results show that falling trees strike a practical balance between interpretability, expressiveness, and risk prioritization for high-stakes settings.

Obtaining the optimal action-value function in Markov decision processes is computationally intensive in large state--action spaces. In this study, we present statistically rigorous convergence results for a robust reinforcement learning algorithm warm-started by a transformer-based action-value function prediction, where natural language prompts encode task specifications. Our framework adopts the R-contamination model to characterize uncertainty in the state transition kernel, and employs conformal prediction to certify convergence via trajectory-level nonconformity scores constructed from the contracting Bellman residual. The resulting conformal quantile bounds the gap between the running and optimal action-value functions simultaneously over all iterations, thereby yielding a pre-certified stopping rule that requires little knowledge of the true transition kernel. Numerical case studies on perturbed maze environments of varying size and contamination level confirm that the transformer-based warm start measurably reduces the initial error and accelerates convergence, while the proposed conformal bounds track the true error trajectory more tightly than existing guarantees.

Probabilistic inference is generally only tractable in low-treewidth graphical models, limiting its effective applicability in high-treewidth settings. Many existing methods improve efficiency by exploiting specific parametric structure, such as symmetries. However, they typically require such structure to be explicitly present, limiting their applicability to a broader range of graphical models. To address this limitation, we propose a framework where tractable inference is controlled by latent parametric structure exploitation, rather than requiring it to be explicitly present a priori. Our approach first reparameterises a graphical model as a specific tensor network representation, which we call a base tensor network. This representation yields two key properties that allow inference tractability to be controlled by parametric structure: 1) First, the complexity of inference is mainly determined by the parametric structure of a single tensor, called the base tensor. We characterise several tractable classes of base tensors for which the entire base tensor network can be contracted efficiently. 2) Second, decomposing the base tensor yields again a collection of base tensor networks. This allows inference to be naturally reduced to decomposing the base tensor into tractable components with sufficient parametric structure. We call this procedure parametric tensor decomposition. By exploiting parametric structure within the base tensor, our framework enables a novel view on inference beyond settings where such structure is explicitly present.

Decentralised federated learning replaces server aggregation with peer-to-peer model exchange, making collaborator selection a local decision under uncertainty. Fixed probe budgets waste effort on easy choices yet fall short when peers are hard to distinguish. We propose SAGE (Sequential Anchor-Gated Exchange), an optimal-stopping peer selector under a one-model-bearing-exchange budget. A receiver scores candidate neighbours on receiver-owned anchor evidence and selects once an advantage is certified. It continues probing only while further evidence repays its cost, and otherwise falls back to random gossip. We show that the stopping problem admits an optimal rule attained at a finite stage, and that the anchor schedule is order-optimal in the peer-risk gap and the confidence level. We further show that the selector never returns a peer worse than random gossip with high probability, and prove that no such guarantee holds for selectors that commit without a certificate. A separability threshold follows, below which no probing budget improves on gossip. Experiments span two image benchmarks, two graph families and three heterogeneity levels. Selectors that always act on their evidence lose to gossip in every configuration tested. SAGE-OS matches gossip on 75.5% less evidence than a fixed budget, at half the communication overhead of two published selectors. The operative decision is not which peer to rank first, but whether the evidence justifies ranking at all.

Root cause analysis at a remote site is slow: evidence is scattered across pod logs, Kubernetes events and cluster-level objects, and many operators cannot send production logs to a hosted model at all. On-premise inference removes the second constraint but raises a question live-cluster benchmarks have not addressed: when one workstation GPU fixes both the model and the context budget, how should evidence be retrieved, and what happens when the runbooks the model consults have been tampered with? We present TriFleetRCA, a pipeline running entirely on one on-premise GPU that collects evidence at one of three scopes (pod, namespace, cluster), ranks it by template de-duplication then BM25, filters runbooks through an ingest guard, and returns a root cause with the evidence lines supporting it. We evaluate on a live Kubernetes cluster into which we inject four faults, so ground truth is known by construction, across 100 analyses with Qwen2.5-14B-Instruct at temperature 0. The hit rate was 0.85, 0.90 and 0.95 at pod, namespace and cluster scope; intervals overlap, but the whole scope effect comes from the one fault whose cause is a cluster-level object, and cluster scope costs 55% more tokens. De-duplication before ranking raised the hit rate from 0.75 to 0.90 at equal token cost. A poisoned runbook telling the model to delete the namespace was rejected by the guard every run; with the guard disabled the model declined to follow it in all 20 analyses, making the guard defence in depth rather than the sole barrier. Separating citation quality from accuracy proved informative: one fault was diagnosed correctly and cited incorrectly every trial, a failure mode accuracy conceals. Median latency was 1.6 s at 2,200 prompt tokens. We release the pipeline, the fault injector and all records.

Internal voids are ubiquitous defects in manufactured structures, yet their characterization remains challenging because their geometry is hidden and can only be inferred indirectly from accessible measurements. Here we introduce \textit{GenVoid}, a physics-informed generative model-based framework for identifying internal voids in complex two- and three-dimensional solids from surface displacement measurements alone. By incorporating the governing mechanics into a generative inference framework, \textit{GenVoid} enables void identification across linear elastic, hyperelastic and plastic material behaviours and accommodates complex two- and three-dimensional structural geometries. Importantly, the framework explicitly accounts for uncertainty and noise in displacement measurements, producing probabilistic reconstructions of internal void geometry rather than a single deterministic estimate. We demonstrate the approach using high-fidelity synthetic datasets and experimentally measured displacement fields obtained from in-situ mechanical experiments, establishing its ability to infer hidden voids from realistic displacement measurements. To quantify the fundamental limits of such inference, we further introduce an observability measure that characterizes the sensitivity of boundary measurements to localized stiffness perturbations within the interior under an ensemble of applied loads. This framework provides a direct connection between defect location, sensor configuration and reconstruction fidelity, enabling systematic assessment of how the number and spatial distribution of boundary measurements govern void-identification accuracy. To this end, these results establish a physics-informed and uncertainty-aware approach for non-invasive characterization of hidden defects and provide a quantitative basis for designing measurement strategies for inverse problems in solid mechanics.

Generative world models aim to predict future states conditioned on actions, where action controllability is fundamental for reliable dynamics modeling. While recent efforts leverage simulator-generated data to enhance this capability, existing training pipelines face two fundamental limitations. First, static offline data collection leads to a distribution misalignment between training sets and the model's evolving error patterns, failing to resolve critical long-tail scenarios where dynamics predictions remain unreliable. Second, the standard objective of minimizing observational discrepancy often encourages the model to exploit spurious correlations instead of capturing the underlying action-effect causality. To address these limitations, we propose OnlineWM, an online training framework that continuously improves world modeling through active simulator interaction and causality-aware optimization. OnlineWM introduces two key innovations: (1) Active Online Learning: Instead of using fixed datasets, OnlineWM adaptively queries the simulator for new interaction sequences that target the model's current predictive weaknesses, ensuring high-utility data acquisition. (2) Causality-Aware Fine-Tuning: We propose a counterfactual learning strategy that contrasts the outcomes of different actions from identical states, forcing the model to attribute state transitions to specific actions rather than ambient environmental evolution, thereby grounding its predictions in reliable causal mechanisms. By integrating active data acquisition with causal optimization, OnlineWM establishes a closed-loop refinement process that ensures the model is both robust to diverse scenarios and precise in its causal attribution. Extensive experiments demonstrate that OnlineWM significantly enhances action controllability and generalizes effectively to unseen domains.

Small open-source large language models (LLMs) in the 0.6B-4B parameter range are increasingly deployed for structured output generation (JSON, function calling, data extraction), yet little is known about how constrained decoding (CD) interacts with model scale in this regime. We benchmark five models from three families across 14 structured-output tasks under three decoding conditions (native, Outlines, XGrammar). We introduce a two-axis evaluation that separates structural correctness (schema validity) from semantic correctness (content accuracy). We find that CD eliminates all structural failures across all models (schema validity: 78.6-92.9% to 100%), but content accuracy reveals a persistent semantic gap that is scale-dependent: type coercion failures are fully CD-rescuable, while instruction-semantic failures (e.g., multi-step function calling) remain CD-resistant. Schema conformance is necessary but not sufficient for semantic correctness; CD's reach ends exactly where schema conformance ends.

Scientific agents support a range of literature-based research tasks, such as retrieval, question answering, evidence-grounded generation, and claim assessment. Most existing systems, however, are organized around individual tasks: the same papers are repeatedly retrieved, segmented, and interpreted, and the understanding built in one task is difficult to reuse in the next. We present ScholarStack, a layered research asset framework that compiles a paper collection into reusable, versioned, and provenance-preserving assets at three complementary levels: source-grounded paper-level statements, domain-level organization, and evidence-grounded cross-paper syntheses. A common access interface returns task-specific views at the evidence granularity each task requires, preserving study conditions, source traceability, and verification status. We instantiate the framework on four task families spanning ten task settings, comparing agents that use the compiled assets with task-specific baselines under matched base models. Quality gains concentrate on tasks that require cross-paper evidence, such as multi-paper question answering and literature review generation, and query-time token cost falls on every task where it is measured, with assets compiled once and reused across tasks. These results suggest that layered research assets can serve as shared infrastructure for scientific agents, shifting literature-based assistance from isolated document processing toward cumulative, evidence-grounded workflows.

Robotic systems are typically composed of multiple independently developed modules that work together to perceive, predict, and act in the environment. Although each module may perform reliably in isolation, composing them does not necessarily preserve uncertainty calibration at the system level. In this work, we show that well-calibrated component interfaces do not necessarily produce calibrated downstream behavior after composition. Using a moving-obstacle prediction pipeline, we demonstrate that position and velocity estimators can each appear well calibrated individually, yet differences in how their error are correlated lead to substantially different estimates of future-state uncertainty. Consequently, assuming independence can make the system either overly confident or unnecessarily conservative, directly influencing downstream planning decisions and safety. Through simulations, we show that modeling the joint covariance restores downstream calibration and improves system performance, whereas dependence-robust uncertainty bounds enhance safety at the cost of increased conservatism. Our findings reveal a fundamental limitation of independently validating robotic modules and highlight the need for interfaces that communicate dependence information or support direct system-level calibration.

Large language models have shown strong performance across a range of legal tasks, but existing benchmarks rarely evaluate the ability to take and defend a legal position, reason under incomplete information, or synthesize multiple statutory provisions. This gap is particularly pronounced for Canadian law, which remains underrepresented in legal NLP. We introduce GRACE (Grounded Reasoning Adversarial Canadian LEgal examples), a dataset of 1,915 question-reasoning-answer instances grounded in Canadian federal legislation. GRACE covers three reasoning modes: adversarial advocacy, uncertainty, and applied reasoning. We develop a pipeline that partitions raw statutory text, generates scenario-based questions and reasoning, and filters examples through model-free citation verification and LLM-based quality auditing. As a proof of concept, we fine-tune CLeAR-4B (Canadian Legal Adversarial Reasoning), a lightweight model for grounded legal reasoning, and evaluate it against the unmodified Qwen3-4B base model in open- and closed-book settings. CLeAR-4B substantially improves agreement with teacher outputs and statutory citation behavior when the relevant act text is provided, while its grounding degrades sharply when the statute is withheld. These results suggest that GRACE can support the development of lightweight legal models that reason more effectively from supplied statutory text.

Textual-gradient methods automate prompt optimization through natural-language feedback, but their iterative updates can be unstable. We identify two sources of this instability: noisy gradients produced from already-correct examples and over-specialization to hard cases that degrades performance on simpler inputs. We introduce STEVE, a stabilization framework with two coupled mechanisms. Error-Driven Refinement generates gradients only from incorrectly handled examples, concentrating updates on informative failures. Regularized Verification treats every update as provisional and accepts it only when improvement on hard cases does not cause unacceptable regression on a preservation set. Across ten reasoning benchmarks, three evaluator/optimizer models, and established prompt-optimization baselines, STEVE reduces degradation and produces more robust prompts. Additional evaluations with gpt-5.4-mini/gpt-5.4 on symbolic reasoning, GSM8K-Platinum, and DS-1000 show that these gains persist with newer models and larger test sets. STEVE therefore provides a practical way to improve the stability and effectiveness of textual-gradient prompt optimization.

Financial language models can transform unstructured firm-specific news into structured decision signals, but financial AI research lacks an integrated deployment framework for evaluating whether those signals remain useful in financial decision systems. Computer science research has developed strong methods for time-series forecasting, text classification, multimodal stock prediction, graph-based market modeling, and machine-learning operations, yet these streams do not provide a domain-specific protocol that jointly tests financial language-model outputs under event-time observability, probability calibration, execution timing, transaction costs, liquidity constraints, capacity limits, operational diagnostics, and statistical inference. We introduce MFAST, a Market-Friction-Aware Sentiment-to-Trading framework that converts timestamped financial text into auditable, reproducible, and market-feasible trading decisions. The application is news-based trading, where firm-specific text must be linked to securities before portfolio decisions can be evaluated. The framework links Refinitiv News Analytics to Center for Research in Security Prices (CRSP) equity data, restricts the primary out-of-sample evaluation to post-release news outside disclosed foundation-model data-freshness periods, and adds a public replication arm using open financial text and public price data. Results show that decoder-only language models outperform encoder baselines and dictionary sentiment in classification, calibration, return prediction, and net portfolio performance, while operational diagnostics reveal trade-offs among accuracy, latency, memory, throughput, and inference cost. The paper shows that credible evaluation of financial language models requires an end-to-end engineering approach combining language understanding, temporal discipline, market-friction-aware deployment, and reproducible validation.

Probabilistic forecasting requires accurate central predictions and calibrated uncertainty estimates. This paper presents TEMPER, the Temporal Encoder-Masked Probabilistic Ensemble Regressor, a univariate time-series forecasting algorithm that combines a temporal autoencoder, a differentiable masked neural decision forest, continuous ranked probability score (CRPS) training, and Gaussian-mixture post-processing. The R implementation is built on torch for R and returns horizon-wise density, distribution, quantile, and sampler functions. We evaluate TEMPER on three deterministic synthetic level series with trend, periodic, regime-switching, nonlinear-threshold, and heteroskedastic components. Across 96 rolling-origin forecasts at horizons t + 1, t + 5, t + 20, and t + 60, TEMPER obtains 2.824% mean CRPS normalized by origin level, 3.635% median absolute error, and 68.8% empirical 90% interval coverage after training with a 300-epoch cap and early-stopping patience of 100. A naive persistence bootstrap has the best aggregate CRPS, 2.763%, while TEMPER has the best median absolute error and the best CRPS at t+1 and t+5. The ablation study uses matched series-origin-horizon cells, horizon-wise CRPS deltas, endpoint sensitivity summaries, and a calibration-specific interval study. Relaxing the learned mask improves average CRPS by 0.472 percentage points on the ablation subset, mainly through long-horizon gains. A twofold interval inflation improves held-out coverage from 54.2% to 91.7% and gives the best 90% interval score among tested calibration rules. The results identify calibration, horizon-specific tuning, and component selection as the central research priorities.

Large language model agents are now privileged principals that take consequential actions: editing code repositories, operating inboxes, completing purchases. Their authority is kernel-grade, but it comes without what classical systems security requires: a trusted mediator interposed on every access. Operating-system vendors are now rebuilding the platform around this de-facto agent kernel, inheriting complete mediation as a design problem. We systematize the security of such systems around a single distinction: a crossing mediated over provenance admits a deterministic check, while one over content semantics does not. A trust-boundary taxonomy locates where mediation must occur and isolates the central mediation gap at two kinds of semantic judgment: distinguishing data from instruction in untrusted input, and an authorized action from an unauthorized one. We argue that this gap leaves an irreducible residual of undetected attacks wherever inputs and actions are not restricted in advance to an enumerated set. The same distinction makes attack-success statistics actionable, placing each number on a spectrum from deployment debt (a sound deterministic mediator left unused) to a structural gap (no such mediator known). We systematize defenses across runtime monitoring, architectural separation, and authorization, and show that current evaluations tend to overstate deployed security through evaluation-validity failures. Finally, we carry that analysis forward beyond the de-facto kernel, to an architecture in which the model itself becomes the arbitration core, and derive the design constraints, open challenges, and research agenda for a security-first AI-native OS.

Multi-teacher on-policy distillation allows a student to learn from complementary specialists on its own trajectories. Domain-routed approaches, however, select one teacher per example and keep it fixed throughout the response. This design both depends on labels that mixed training corpora often lack and cannot adapt teacher selection when the expertise required changes within a trajectory. We propose \textbf{TrustMOPD}, which replaces example-level teacher selection with label-free, token-level supervision allocation. At each student-generated prefix, TrustMOPD uses each specialist's RL-induced displacement from a shared pre-RL reference as a proxy for local reliability, calibrates these scores across teachers, and constructs a weighted distillation target. Across mathematics, code, and instruction following, TrustMOPD outperforms the strongest label-free baseline, increasing the recovery ratio from $54.4\%$ to $91.5\%$ on \textsc{SingleCap} and from $54.5\%$ to $98.0\%$ on \textsc{MultiCap}, while approaching label-based MOPD on \textsc{SingleCap}. Randomizing token-level weights independently of the student-generated prefix performs no better than uniform weighting, supporting the importance of conditioning supervision on the evolving generation context.

Recent advances in Physical AI have accelerated the use of foundation models in autonomous systems such as unmanned aerial vehicles (UAVs), which must perceive, reason, plan, and act in dynamic environments. Existing benchmarks assess physical perception, intuitive physics, embodied navigation, and collaborative reasoning, but rarely evaluate the agentic decision-making required for reliable autonomy. We introduce \textit{PhysAI-Bench}, a benchmark for evaluating this capability. It contains 10,178 standardized decision instances automatically extracted from conversational traces of autonomous UAV missions. Each instance preserves mission context, temporal dependencies, physical constraints, Model Context Protocol (MCP) tool calls, Agent-to-Agent (A2A) interactions, sensor observations, and AI-native 6G network conditions, including latency, packet loss, throughput, edge load, and network slicing. We expose only information preceding each decision, preventing future-event leakage and approximating online decision-making. We evaluate 29 foundation models using a two-stage protocol. We select model-specific configurations from 12 combinations of zero-, three-, and five-shot prompting and four temperatures, tested in three runs on a 35-instance, human-verified development set. We then freeze each selected configuration and evaluate it in three runs on a fixed, episode-disjoint set of 500 instances. GPT-5.3 achieves the highest accuracy (52.00%), followed by GPT-5.2 (49.40%) and Grok~4.5 (49.07%). Few-shot prompting generally improves performance, while temperature has limited influence. The results demonstrate that reliable agentic decision-making in Physical AI remains an open challenge. The dataset is available at https://github.com/maferrag/physai-bench

We study norm-constrained linear classification under Eu clidean adversarial perturbations in a Gaussian model with a low-dimen sional informative subspace and an independent noise tail. For bounded ramp loss, we prove that a principal-space witness with risk below one half forces every near-optimal predictor to have small tail weight. A path-specific density bound yields constants without requiring positive tail variance. Under isotropic principal covariance, we establish a unique population minimizer and joint local growth. Boundary normalization then removes the common attack penalty from centered margins, giving localized finite-sample guarantees governed by principal dimension and total tail energy. Globalized growth removes the entrance condition at weaker constants; a model-aware comparison retains local guarantees. Experiments with twenty paired repetitions show decreasing excess risk and tail use with sample size, and nearly unchanged behavior when tail dimension grows at fixed total energy. Pure-noise controls and optimizer diagnostics clarify the scope and limitations of these conclusions.

This paper introduces a multi-temporal tabular dataset derived from satellite images to map burned areas in the Chapada dos Veadeiros National Park, in Goiás, Brazil, covering the years 2020 to 2022. The dataset contains blue, green, red, and near-infrared bands, as well as the BAI, EVI, GEMI, NDVI, and NDWI spectral indices from the WFI sensor on the CBERS-4A, CBERS-4, and AMAZONIA-1 satellites, organized into a regular grid. We applied the Random Forest classifier to develop and validate models based on samples labeled as totally burned, partially burned, and non-burned. Two classification approaches were tested: one combining burned and non-burned areas into binary classes and another distinguishing between totally burned (TB), partially burned (PB), and non-burned (NB) classes. Seven validation approaches assessed different post-classification combinations, focusing on accuracy, precision, recall, and intersection over union (IoU) metrics. Results showed higher IoU when TB, PB, and NB were used as individual classes and TB was reclassified as burned area (BA) while PB and NB were grouped as non-burned. Comparing the annual results of this approach to the MCD64A1 product, the errors of omission for the BA class were 22% in 2020, 28% in 2021 and 59% in 2022, while the errors of commission were 46%, 43% and 46%, respectively. The study highlights the utility of the WFI sensor for burned area mapping without inter-satellite spectral calibration and suggests further exploration with other machine learning algorithms to evaluate the dataset potential and limitations.

Patch-based autoregressive time-series forecasting often ties input representation, learned transitions, and recursive execution to one patch length. We ask which of these roles can be adjusted separately. A supporting atomic-encoding study finds greater sensitivity to model width than to atom grouping on the evaluated grid. Our main finding is that a frozen parent's recursive trajectory is easier to fit than the observed future with lightweight parallel exits. Autoregressive Trajectory Distillation (ATD) turns this into selectable ATD-1/2/4/8 execution, with ATD-1 exactly recovering the parent. On a paired four-data-set comparison, ATD-8 reaches $5.54\times$ end-to-end speedup with stable quality across widths. Fewer calls do not automatically remove the parent's existing forecast error: ATD improves trajectory fidelity in all 21 seed runs but forecast accuracy in only 15 against matched clean-future supervision. We further find a correctable residual projection along a train-selected periodic history direction. Spectrum Tangent applies this correction without adding neural parameters or Transformer calls. At horizon 720, it reduces mean squared error (MSE) and mean absolute error (MAE) by 2.54% and 2.33% over seven data sets and two output widths, while remaining $3.24\times$ faster than recursive inference. Level and shape projections sometimes disagree. Trajectory compressibility, the fidelity-accuracy mismatch, and the correction recur across three public AR parents. Together these results separate representation, transition, and execution as AR design axes. Code is available at https://github.com/RowanFFF/ATD-Spectrum-Tangent.

The A* search is a fundamental path-finding algorithm in artificial intelligence. While admissible and consistent heuristics guarantee efficient performance by expanding each state at most once, modern search applications frequently employ powerful but inconsistent heuristics derived from machine learning, randomized evaluations, etc. A long-standing theoretical barrier to using these inconsistent heuristics is the risk of catastrophic node re-expansion, which yields a worst-case exponential time complexity of $Ω(2^n)$. However, empirical observations contradict this pessimistic bound, demonstrating that inconsistent A* operates highly efficiently in practice. To bridge this significant gap between theory and practice, this paper presents the first smoothed analysis of the A* algorithm using inconsistent heuristics. We model typical real-world noise by applying slight random perturbations to the edge weights of worst-case search graphs. Our main result proves that the expected smoothed time complexity of inconsistent A* is bounded by a polynomial, specifically a total iteration number of $O(n^2 m κ)$, where $n$ is the number of nodes, $m$ is the number of edges, and $κ$ controls the scale of random perturbations. Furthermore, we also show that this result naturally extends to the functionally equivalent problem of Dijkstra's algorithm on negative-weight graphs.

Inferring network topology from noisy node observations is a central problem in graph signal processing. In this paper, we consider Laplacian-constrained graph estimation for Gaussian Markov random fields, focusing on the underdetermined regime in which the number of samples is smaller than the number of graph nodes. Existing approaches often formulate the problem as a sparsity-regularized maximum-likelihood estimation problem. While effective, such methods typically require iterative optimization and are often computationally demanding, particularly under Laplacian constraints. Instead, we propose a non-iterative estimator of graph Laplacians that uses effective resistance for regularization, and evaluate the method using a simple sparsification procedure. Experiments show that with some trade-off in edge and weight recovery on the considered dataset, computational cost for moderately sized graphs can be substantially reduced.

Constraint-following music generation asks a score to satisfy several user-specified properties at once, each checkable programmatically (key, meter, length, range, final note, rhythm, motion and form), yet no existing benchmark isolates this capability. We construct MusicConstraintBench, 2,180 items over eight constraint families, on which current models fail once a few constraints are combined. The natural remedy is reinforcement learning with these verifiers as reward, yet we observe that a reward paid only when every property holds leaves most training groups without a learning signal: over the first 50 updates, 0.550 of rollout groups score identically and receive no gradient, even though a failing score typically misses only one requested property. Under the joint criterion, rollouts for a prompt tend to fail together, so a binary reward cannot separate a nearly correct score from a malformed one. We therefore introduce MusicRLVR, which pays graded per-property credit behind a hard validation gate that rejects malformed outputs, plus a joint-satisfaction bonus, requiring no human annotation, learned reward model, or music-domain fine-tuning. On MusicConstraintBench, MusicRLVR lifts Qwen3-4B-Instruct from 0.160 to 0.807 on mixed constraints and leads every zero-shot baseline including Llama-3.1-70B at 0.380. It also generalises to property combinations unseen in training and to out-of-range parameter values, showing that verifiable rewards need not presuppose a target output.

Ethereum decentralized finance (DeFi) provides a public, time-stamped record of transaction-level event streams, but the same public symbols can create strong machine-learning shortcuts. We introduce ETH-TraceBench, a benchmark for evaluating Ethereum DeFi representations under temporal, protocol, pool/infrastructure, and symbolic shift. The raw event universe covers January 2021-December 2025 and contains 1.35 billion transactions with logs and 5.01 billion raw log rows. Model evaluation uses a fixed 911,267-instance supervised sample, training on 2021-2024, selecting models on 2025H1, and testing on 2025H2. Simple models perform strongly on the aggregate temporal test: TraceStats-GB reaches 0.953 macro-F1 and TopicEmitterHashMLP 0.959 on the canonical DEX test set. Performance drops sharply under protocol novelty, with macro-F1 of 0.794, 0.743, and 0.766 for TraceStats-GB, TopicEmitterTrace-SGD, and TopicEmitterHashMLP, while strict unseen-pool scores remain 0.927, 0.897, and 0.935. Uniswap v4 and Ekubo v1, both absent from supervised training, are materially harder than the full test. Jointly masking emitter and topic identity reduces DEX macro-F1 to 0.916 and liquidation macro-F1 to 0.774 for TopicEmitterTrace-SGD. A standard Transformer over log-index-ordered events provides no consistent advantage over a deterministic shuffle of the same events, indicating that high aggregate scores can arise without sophisticated chronological modeling. A natural-prevalence audit estimates 2025H2 DEX prevalence among logged Ethereum transactions at about 22.5%, and a deterministic 400-transaction audit finds complete agreement with task label sources and independently re-queried raw-log counts. ETH-TraceBench therefore treats difficult transfer and controlled-input conditions, rather than a single aggregate score, as the main evaluation target.

Despite impressive visual quality, state-of-the-art video diffusion models often generate content that violates real-world physical laws. While existing solutions rely on external priors or specialized data, we investigate the root cause by exploring the internal mechanisms of these models. Specifically, we present the first interpretability study on the ''motion planning'' process of text-to-video diffusion models, revealing how motion trajectories form during early denoising stages. Building upon the ''first shape, then details'' finding, we combine cross-attention trajectory patterns with causal head contributions to identify a specific subset of attention heads driving motion planning. Further, our self-attention analysis shows that Rotary Position Embedding (RoPE) induces excessive spatial attention decay. This causes early candidate regions to prematurely lock into physically implausible positions, suppressing reasonable trajectories in adjacent frames and triggering generation failure modes. To address this fundamental flaw, we propose a lightweight architectural modification that scales the frequency of RoPE across different denoising steps. This strategy reduces excessive attention decay, helping the model explore better candidate regions to establish coherent physical motion. Finally, training-free and training-based experiments confirm the effectiveness of our approach in enhancing the physical commonsense of generated videos.

Vision-language-action (VLA) models have achieved strong performance in embodied manipulation, but still lack a clear mechanism to balance behavioral stability with task-semantic sensitivity. We identify two complementary failure modes. Under task-preserving changes, where task semantics remain unchanged but scene appearance varies (e.g., style, illumination, clutter, or paraphrasing), policies often exhibit unnecessary action drift. Conversely, under semantic-breaking changes, where key task semantics such as the target object or constraint are altered, policies frequently fail to produce sufficiently distinct behaviors and instead follow the original trajectory. To address this gap, we propose BAS-VLA, a task-semantic action calibration framework built on top of a frozen base VLA. BAS-VLA adopts a breaking-centered calibration core as the default path, and introduces a selective evidence-gated preserving auxiliary that activates only when nuisance variation is detected while task semantics remain consistent. On the OpenPI-pi0.5 / LIBERO-Object Milk-Swap benchmark, BAS-VLA maintains high success on clean (98.0%) and semantics-preserving conditions (97.5%), while reducing clean-criterion success to 0.0% under deliberate target-object swaps, demonstrating strong stale-task suppression and task-semantic separation. On validated style-preserving shifts, it improves success from 42% to 70% without degrading clean performance. These results highlight that reliable VLA behavior requires moving beyond appearance robustness toward explicit task-semantic action calibration.

E-commerce search requires distinguishing products that are merely related to a query from those that directly satisfy the user's shopping intent. We augment query-product pairs with structured LLM-generated query and product attributes and human-validated relevance, explanations, and centrality judgments, and evaluate these signals using a simple dual-encoder retriever and MLP re-ranker. On an augmented subset of ESCI, a human-feature oracle reaches $0.9382$ nDCG@10, while a human-free trained $Q+P$ configuration reaches $0.9258$. Synthetic approximations of the human signals reach $0.9150$ overall but provide substantial gains for difficult, low-performing queries. Ablations show that most of the oracle improvement comes from post-edited explanations and annotator comments rather than the scalar centrality feature, suggesting that LLMs are most useful for exposing and approximating structured semantic supervision rather than replacing human judgment directly.

Autonomous navigation of unmanned aerial vehicles in constrained three-dimensional environments has been a challenge in the robotics domain. The application of autonomous unmanned aerial vehicles in civil infrastructure inspection involves the use of such vehicles in bridge inspection, tunnel inspection, and structural inspection. The use of deep reinforcement learning in the autonomous navigation of unmanned aerial vehicles has been successful in constrained environments. However, the computational cost of the algorithm limits the application of the algorithm in the autonomous navigation of unmanned aerial vehicles. This paper proposes the use of the spiking neural network-based Proximal Policy Optimization algorithm in the autonomous navigation of unmanned aerial vehicles in constrained sequential environments. The proposed algorithm integrates the use of spike-based actor-critic reinforcement learning with the Proximal Policy Optimization algorithm. The proposed algorithm uses the stochastic Gaussian policy in the autonomous navigation of unmanned aerial vehicles. The proposed algorithm was implemented in the autonomous navigation of unmanned aerial vehicles in constrained 3D environments. The proposed algorithm was successful in completing 1913 episodes out of more than 3000. The proposed algorithm was successful in passing an average of 2.10 windows per episode. The proposed algorithm was successful in achieving a success rate of 63.77%. The proposed algorithm was successful in achieving success rates of more than 90% in the later stages of the algorithm.

Human-feedback alignment has made language models useful assistants and is commonly described as aligning them with humans. However, the responses people prefer from an AI need not be the responses they themselves would give. We distinguish alignment with human preferences from alignment with human behavior, and show that alignment with human preferences can make model behavior less human-like even when both preferences and responses come entirely from humans. We call this the Turing-test gap. We show that preference alignment preserves the human response distribution only under a restrictive condition, and find no consistent evidence that real human preferences satisfy it. Empirically, the loss of human-response likelihood increases with the strength of preference weighting, regardless of its direction, and the gap also appears under standard DPO. These results establish human-likeness as an explicit dimension of alignment rather than something assumed to follow from preference alignment.

Long-video understanding must capture transient visual evidence under strict token budgets, yet existing methods compress frames, append memory tokens, or alter internal key-value (KV) caches. We introduce Prefix-Steered Recurrent Memory (PREM), a memory-token-free framework for frozen vision-language models (VLMs). PREM separates video ingestion from query answering: a recurrent writer distills visual streams into a compact 256 KiB multi-slot associative state, while a question-conditioned readout adds memory-derived key/value (K/V) steering modulations to existing non-visual prompt prefixes during prefill. This enables write-once, query-many inference without extra prompt tokens or decoding recurrence. Across six long-video benchmarks in offline and streaming end-of-stream settings, PREM consistently outperforms frozen baselines at every evaluated visual budget. Under a constrained budget of 16 frames, PREM improves macro-average accuracy by 3.06% on Qwen2.5-VL-3B, with gains of 11.0% on action antonym identification and 9.9% on localized needle retrieval. These gains require tuning 0.24% of backbone parameters at 0.03 GiB of peak GPU memory overhead.

Prompt learning efficiently adapts vision-language models (VLMs) to downstream tasks, but gains on seen classes often come at the expense of generalization to unseen classes. To address this limitation, we propose prompt ensembling with training-free routing (PETR), whose key innovation is a carefully designed dual-prompt architecture: two complementary prompts are learned from different data and objectives to emphasize seen class discrimination and unseen-class generalization, respectively. During training, both prompts are fine-tuned using a shared frozen CLIP backbone, and statistical information is collected from the training set logits. At inference time, we determine the similarity of each test sample to seen data, and route the sample to the most appropriate prompt branch. To the best of our knowledge, this is the first prompt tuning framework that performs training-free adaptive routing based on statistical similarity. This design provides an interpretable routing signal and avoids common MoE-style routing pathologies, such as router training instability and load imbalance. Extensive experiments on 11 benchmark datasets demonstrate that our framework consistently outperforms previous methods on both seen and unseen classes, achieving new state-of-the-art results.

With face-recognition models now embedded in everyday authentication and surveillance, recent works have pinpointed a critical weakness: these models remain acutely vulnerable to adversarial semantic edits. I.e., adversarially produced semantic alterations to the input, such as slight aging or pose changes, can induce misclassifications. Certain existing attacks are powerful, but they can be computationally costly, rendering them inadequate for developing defenses (e.g., through adversarial training). To fill the gap, we introduce BoundStyle, a potent semantic attack operating in StyleGAN's rich latent space to maximize misclassification rates. Notably, BoundStyle achieves high attack success rates while being ${\sim}{\times}9.5$ faster than existing state-of-the-art attacks, making it suitable for adversarial training. Building on BoundStyle, we develop StyleAT, an efficient adversarial training scheme that incorporates low-budget attack variants yet defends against stronger and unseen semantic attacks. We evaluate on two datasets unseen during training and seven models, and find that StyleAT boosts robust accuracy against state-of-the-art attacks and outperforms common defenses in various settings.

Orthogonality in a LoRA factor does not by itself specify what the composed update protects: the answer depends on the task-start state, the parameterization, and the realized optimizer displacement. We formalize this question through Bilinear Optimization Divergence (BOD), an anchor-relative diagnostic of effective-update response on selected historical features. The finite-step analysis distinguishes two cases. In a shared adapter, protecting the routing displacement leaves a learned-anchor residual through the changing companion factor. In a fresh zero-output block, a feasible routing state can protect the composed update while both current factors remain trainable. These conditions yield Semi-Frozen Orthogonal Routing (SFOR) for shared adapters and current-block hard protection for cumulative O-LoRA; Weight Residual Projection (WRP) enforces the required displacement after the optimizer step. Controlled two-task traces verify the predicted residual paths, reducing normalized historical response from 19.12% to 0.005% in the shared family and from 7.72% to 0.002% in the cumulative family. Four-task experiments on Qwen3-8B characterize the resulting trade-offs: SFOR improves backward transfer (BWT) from -2.47 to -0.86 with nearly unchanged average accuracy (AA), while O-LoRA hard protection improves three-order mean AA from 80.27% to 81.30% and forgetting measure (FM) from 2.20 to 0.43. Component controls also show that stricter feasibility need not improve final task performance. Together, the analysis and evidence provide an architecture-conditioned account of which constraint to enforce, how to enforce it, and how to interpret its empirical value.

End-to-end document parsers increasingly offer an optional reasoning mode for complex pages. On a 180-page entropy-stratified discovery sample with one frozen 4B checkpoint, complexity is the wrong decision variable. Reasoning lowers mean quality by 2.21 Overall at 1.54x tokens; a preregistered input-only model cannot predict its signed benefit (held-out AUROC 0.47, indistinguishable from chance). The benefit concentrates on pages whose ordinary pass has already collapsed, and they do not look complex: shared collapses have lower layout entropy than healthy ones yet consume 19x the tokens as degenerate repetition that doubling the budget does not cure. Switching modes rarely repairs them: 83% recur under reasoning. We instead detect collapse from the ordinary-pass trace, decompose the page by projection, and re-parse each region. Repair gains 1.40 Overall (95% CI [0.68, 2.16]) at 1.13x tokens, replicates across three checkpoints, and, with all parameters frozen, gains 2.41 (CI [1.64, 3.46]) on the remaining 1,175 benchmark pages.

Battery energy storage system (BESS) dispatch must preserve operational feasibility while declining price spreads reduce the margin available to pay for cycling. We study a proximal policy optimization (PPO) controller whose pre-selection physical action mask and emergency projection are separated from a causal, forecast-informed economic advisory. All forecast-dependent methods receive the same causal 24-step forecast and grid-side settlement. Across five PPO seeds, advice-on net profit is 30.59 and 18.04 USD per 336-hour T1 and T2 window, versus 36.77 and 22.94 USD for proxy-cost MPC; PPO remains below this reference in both periods. Advice raises T2 profit from 16.45 to 18.04 USD while reducing throughput, but is immaterial in T1. On disjoint weekly blocks, PPO is stable under daily, weekly, and blended seasonal forecasts, weakens under persistence, and remains below proxy-cost MPC. Paired diagnostics localize changes to the observed 5-10 USD/MWh regime with mixed SoC-dependent effects. An M0-M6 ablation shows that mask removal sends thousands of infeasible requests to projection, while removing both physical layers exposes ramp violations. The evidence separates economic screening from feasibility enforcement without claiming formal safety, lifecycle-optimal aging, or RL dominance.

Large audio-language models (LALMs) are increasingly used for a broader range of audio reasoning tasks. These models typically incorporate audio representations into a large language model (LLM) backbone to enable multimodal reasoning. Recent test-time reinforcement learning (TTRL) methods further improve LLM reasoning capability by leveraging unlabelled test data after pre-training. However, the importance of the perceptual capability of LALMs remains underexplored, particularly how much acoustic evidence is integrated and relied upon during reasoning, and how this contributes to final task performance. This gap limits the development of effective post-training methods like TTRL for audio reasoning. In this work, we first analyse how audio information is integrated and utilised during reasoning process. We quantify layer-wise perceptual reliance and show that stronger acoustic reliance is associated with higher accuracy and a larger performance gain attributable to the audio input. Building on this, we propose Perception-Grounded TTRL (PG-TTRL), which aligns label-free test-time optimisation with perceptually grounded reasoning, encouraging the model to structure its reasoning more strongly on the audio input. Experiments across LALMs and benchmarks show that PG-TTRL consistently improves reasoning performance over both the base models and standard TTRL, showing the value of perceptual-grounding optimisation for test-time audio reasoning.

Large reasoning models (LRMs) incur high inference costs, often mitigated by efficiency techniques like quantization and pruning. However, the impact of these techniques on model adversarial robustness remains largely unexplored. This study provides the first comprehensive analysis of the interplay between efficiency, jailbreak vulnerability, and reasoning in LRMs. We find that while efficiency methods seemingly reduce the success rate of jailbreak attacks, this improvement is often superficial. It largely arises from degraded reasoning capabilities leading to "attempted but failed" malicious responses, rather than an increase in genuine alignment. Mechanistic analysis of representational drift confirms this, revealing a strict coupling between reasoning capability loss and the model's inability to maintain malicious semantic trajectories. Additionally, we identify quantization with pruning as the optimal strategy to balance efficiency and robustness. These findings clarify the distinction between true safety alignment and capability-induced failure, providing an empirical foundation for LRM deployment.

Sink-aware deployment may identify important first-token attention heads before a model is quantized, then reuse that map at the edge. We test when this shortcut is safe for 4-bit NF4 weight-only post-training quantization (PTQ). Our Sink Topology Consistency (STC) metrics separate global rank preservation, top-$k$ set overlap, and layerwise sink-mass shift, and distinguish per-input sensitivity from calibration-map transfer. Across Qwen2.5-0.5B, Qwen2.5-1.5B, and Llama-3.2-1B, global bf16-to-4-bit ranks remain high at 4,096 tokens ($ρ_s \geq 0.980$), yet top-$k$ Jaccard overlap is only 0.619-0.793, corresponding to 76.5-88.5% membership retention. The global statistic also masks local failures: terminal Qwen layers shift by 6.2-7.9x their model means, whereas Llama-3.2-1B shows low, nearly uniform drift. Under a C4-to-LongBench shift, cross-domain overlap degrades more than the within-domain precision comparison for both Qwen models, but not for Llama-3.2-1B. Matched-domain 4-bit recalibration reaches 90% of a split-half stability plateau at the smallest tested $n=8$ for both Qwen models and $n=32$ for Llama-3.2-1B, though not as a sharp threshold; for the two Qwen models, updating only selected layers does not reach the full-map stability criterion. On Jetson Orin NX, the 16-sample workload takes seconds for the two models with valid on-device sink measurements. The practical message is precise: global rankings often transfer, but discrete head sets, layer-local policies, and cross-domain calibration should be revalidated after quantization.

Maintenance work orders must often be processed offline on embedded hardware, yet downstream software requires predictable structured output. We present ARID (Aviation-inspired Routing for Industrial Deployment), which extracts component, failure mode, symptom, and maintenance action into fixed-schema JSON on an 8 GB NVIDIA Jetson Orin NX. ARID combines conservative dual-teacher filtering, targeted noise-aware synthesis, one routing decision per work order, 4-bit inference, and grammar-constrained decoding. From 2,326 unlabeled OMIn records, it retains 716 training pairs and adds 99 topology-constrained records targeting action extraction. On 300 human-labeled records, ARID reaches 84.8% token-F1 on the reference stack and 82.9% on the deployed Jetson. Resident serving achieves 5,310/5,656 ms P50/P99 at 12.5 W. On zero-shot MaintNet transfer, semantic F1 falls to 46.4% while parser success remains at least 99.8%, showing that output validity transfers but field semantics do not.

In high-dimensional tabular learning, feature screening provides a lightweight, model-agnostic way to remove irrelevant features before model fitting. However, scoring raw values directly can miss nonlinear or distributional structure. We introduce PACE (Plug-and-Play Contextual Embedding), which inserts a frozen tabular foundation model (TFM) column encoder before an existing feature-scoring rule, expanding each feature into a higher-dimensional contextual representation. Across controlled studies, PACE improves raw-space screening of complex nonlinear dependence with only modest additional encoding cost. These gains translate to downstream prediction on TALENT datasets: PACE-DC improves binary AUC by 0.077 and multiclass macro-AUC by 0.064, with a median normalized RMSE improvement of 0.063 across ten learners. Matched random-weight and random-feature controls show that PACE gains from pretrained structure beyond generic dimensional expansion. PACE further achieves favorable performance--time trade-offs against task-fitted selectors and attribution-based methods, positioning pretrained column geometry as a reusable upstream primitive for tabular learning.

Synthetic essays can help reduce dependence on human-written data in Automated Essay Scoring (AES). However, they often lack realistic errors, limiting their ability to represent authentic human writing, particularly when the target texts are intended to resemble those produced by language learners. In this study, we present a simple approach that introduces error supervision into synthetic essay generation. Specifically, we fine-tune an LLM generator on error-annotated texts of the kind commonly used in Grammatical Error Detection (GED). To assess the utility of the proposed approach, we fine-tune and evaluate AES scorers under three data conditions: authentic essays, synthetic essays generated conventionally, and synthetic essays generated using our proposed approach. The results show that in the larger-data settings, the proposed approach outperforms the conventional synthetic baseline in 11 out of 12 dataset-metric comparisons, with performance in some cases approaching that of models trained on authentic essays. Despite these gains, performance under extremely low-resource settings remains mixed, with advantages over the conventional baseline only becoming more apparent at 200 training essays, although not consistently across datasets. Qualitative and quantitative analyses further show that the proposed approach produces learner-like errors whose distributions broadly resemble those observed in authentic essays.

Coding agents operate on real repository coding tasks, and persistent memory systems promise to reuse experience across tasks. Yet existing evaluations do not show whether those systems improve executable repository work. Repository benchmarks test code changes but do not isolate memory, while memory benchmarks score recall without measuring downstream coding outcomes. We introduce VibeMemBench, a benchmark for evaluating memory systems on 111 coding targets from 90 SWE-rebench V2 repositories and 3,634 history trajectories from the target repositories. The targets follow the SWE benchmark style and cover bug fixes, feature requests, interface changes, and configuration work. An agent edits each target codebase under a declared memory condition. Executable tests decide task resolution. Each target is retained only when injected history experience improves its executable outcome in a reference setting, so every target carries a prior experience whose usefulness is verified by execution in that setting. The frozen verified experience is then transferred to five held-out solvers. Direct injection raises observed task resolution on four of them by 1.1 to 4.5 percentage points while lowering agent steps on all five. Yet when four existing memory systems must construct and retrieve experience from the same history, eleven of twelve solver and system pairings fail to exceed the matched memory-off baseline. VibeMemBench exposes the gap between the useful experience that repository history holds and the experience existing memory systems deliver for repository coding tasks.

We reconsider the theory of probabilistic formal languages generated by n-gram models and by probabilistic context-free grammars (PCFGs). The expected hierarchy of probabilistic grammars is established by proving that every probabilistic language generated by an n-gram model is also generated by some PCFG, while some probabilistic languages generated by PCFGs cannot be generated by any $n$-gram model. We introduce the notion of fully connected PCFGs, namely PCFGs in Chomsky normal form where every production rule only involving non-terminals has non-zero probability. Our main result shows that any probabilistic language generated by an $n$-gram model differs from any probabilistic language generated by a fully connected PCFG. Therefore, the class of probabilistic languages generated by $n$-gram models is not a subset of the class generated by fully connected PCFGs.

Vision-Language-Action (VLA) models integrate vision-language understanding with executable robot actions, enabling end-to-end learning for robot control. However, our empirical analysis reveals that existing models exhibit severe trajectory overfitting when finetuned on limited datasets. To guide the model in effectively utilizing wrist camera information, we propose MaskVLA, a masking-based fine-tuning strategy. By randomly masking a small portion of the main camera's visual information, the model is guided to autonomously learn more fine-grained, task-relevant, and effective visual features. This process leads to the emergence of robust policies, thereby enhancing the model's capability to tackle complex manipulation tasks and improving its generalization performance. Our method has been comprehensively evaluated on RoboTwin 2.0, achieving an average success rate improvement of 23.2% and 16.8% compared to $π_0$ and OpenVLA-OFT, respectively. Furthermore, experiments on real-world ALOHA robots also demonstrate the effectiveness of our approach.

The SYNTAX score is a clinically established tool for assessing anatomical lesion complexity in coronary artery disease and guiding subsequent treatment. However, automated SYNTAX scoring is commonly formulated as a direct regression problem from coronary angiography videos to patient-level scores. In this work, we reformulate SYNTAX scoring as a vessel segment identity-preserving anatomical reasoning problem and propose a hierarchical modeling framework that explicitly aligns learning with the clinical workflow. Our approach maintains vessel segment identity across frames and views, estimates stenosis severity at the segment level, and aggregates evidence hierarchically according to coronary anatomy. Simultaneously, to address the scarcity of domain-specific data, we integrate and complete multiple public coronary angiography datasets, constructing a large-scale resource featuring completed vessel segmentation and derived structural annotations. Experiments demonstrate that vessel segment-level stenosis embedding enhances explanatory power and reduces prediction variability compared to baseline models, with the R^2 score improving by 0.201 and dev STD decreasing by 18.4%. These results highlight the necessity of structure-aligned modeling for reliable and stable automated SYNTAX scoring from multi-view coronary angiography videos. The GitHub link is https://github.com/VersaceSu7/SYNTAX_score_777.

Standard positional encodings represent position as a one-dimensional reading-order coordinate, but reading order alone does not determine hierarchical textual structure. We use a hierarchical rotary positional encoding (hRoPE) that represents paragraph, sentence, and token indices as separate channels, hold the token sequence fixed, intervene on the paragraph coordinate p1, and measure cross-paragraph attention with a token-distance-exact estimator. Attention is compressed relative to a token-distance-matched baseline in every corpus, but compression alone is not diagnostic of true structure: an architecturally identical channel with density-matched random labels is compressed too, more shallowly. What distinguishes real structure is the depth of compression, which is greater and corpus-dependent while the control's is not. Comparing eight corpus-only quantities across three constructs (lexical persistence, paragraph length, embedding-based coherence), none fully reproduces the cross-corpus ordering of depth, though embedding-based coherence comes closest. Compression depth, not its location, is the reproducible signature of genuine paragraph structure in our setting.

Screening oxygen-evolution catalysts on combinatorial libraries requires deciding which candidates receive the remaining measurements. The deciding activity lies beyond each candidate's measured potential window and often above every activity recorded during fitting. We predict it by physics-residual machine learning: the Tafel equation extrapolates the candidate's own measured current and slope, a learned residual attenuated with feature-space distance corrects the magnitude, and an applicability-domain score identifies predictions above the training range before measurement. In a separately fabricated 322-candidate library, 282 above the training maximum, two measurements per candidate gave a mean absolute error of 0.203 mA cm$^{-2}$ against 1.330 for the selected data-driven machine-learning model. Errors inside the training range remained comparable, and 35 labelled catalysts were enough to fit it. In two independent datasets the same construction lowered the overpotential error by 29 to 52%. Campaigns can therefore shorten each measurement and still rank the most active compositions.

Graph prompt learning enables parameter-efficient adaptation of frozen Graph Neural Networks to downstream tasks through lightweight prompt parameters. As routing becomes increasingly node-adaptive, however, independently optimized local decisions can collectively concentrate assignment mass on a small subset of a finite shared prompt bank, even when individual node--prompt matches remain locally meaningful. We propose MINT (Measure-INtegrity Transport), an entropically regularized optimal transport framework that formulates node-to-prompt adaptation as a globally coupled allocation problem. The transport cost favors local geometric compatibility, while a prescribed prompt-side marginal explicitly controls graph-wide prompt utilization. We further derive an exact variance decomposition that separates prompt-side geometric variance into retained prompt-update variation and within-node barycentric dispersion, together with a conditional stability bound for the frozen-encoder forward map. Across standard citation networks and additional heterophilic graphs, MINT remains competitive in few-shot adaptation. Controlled and end-to-end experiments further distinguish the roles of routing and topology: fixed-marginal routing controls graph-wide prompt utilization and has measurable end-to-end effects on citation networks, while topology augmentation provides a complementary, graph-dependent mechanism for addressing structural mismatch. Code is available at https://github.com/Ga1axy0051/MINT.

Agents may need capabilities exposed through external agent endpoints or service APIs. When a requester is not already bound to a provider, it must discover advertised capabilities matching its task and interface requirements. Over exact-key distributed hash tables (DHTs), broad retrieval transfers large candidate lists, whereas selective retrieval may miss relevant providers or require more replication and lookups. Open publication also lets providers inflate their exposure unless publication bounds are enforceable. We present SemDHT, a certified semantic index for discovering agent-accessible capabilities over exact-key DHTs. A two-layer semantic sketch uses coarse cells to group nearby descriptors and residual codes to narrow candidate selection. Providers publish at a bounded set of derived keys, while requesters probe precision keys before broader recall keys within a lookup budget. Anchor committees certify each descriptor's publication-key set, enabling storage services and requesters to enforce descriptor-to-key consistency. On real API descriptors and task queries, SemDHT achieves recall@10 of 0.955 against exact embedding-space neighbors and 0.947 against ToolBench relevance labels. On a corpus with controlled density augmentation, it matches the candidate exposure of tuned locality-sensitive hashing (LSH) over a DHT at recall 0.95 with 7.7x fewer lookups and reduces publication fan-out from 16 to 10. A Go/libp2p prototype deployed on same-region and cross-region 200-peer cloud overlays replays 299 Internet queries. With parallel probes and cold certificate caches, SemDHT achieves mean completion-time speedups of 3.64x and 4.11x over LSH, respectively.

Causal discovery from observational data is a fundamental yet challenging task in scientific research. While existing approaches are primarily based on conditional independence tests, structure scores, or restrictive functional assumptions, we propose Decoupled Causal Discovery (DCD), a novel decoupling-based perspective that does not rely on these methodologies. DCD directly identifies the Markov boundary (MB) by decoupling non-target variables via weighting functions, such that only variables within the MB preserve dependence with the target under the decoupled distribution. Building on this, DCD iteratively constructs the Completed Partially Directed Acyclic Graph (CPDAG) by exploiting structural asymmetries within the MBs. We establish the theoretical identifiability, soundness, and completeness of DCD. Empirical evaluations demonstrate that DCD achieves strong performance, particularly excelling in challenging noise regimes.

Neural operators have emerged as powerful surrogates for solving partial differential equations (PDEs), yet their reliability under distribution shift remains a critical barrier to deployment. Existing approaches to out-of-distribution (OOD) generalization in operator learning are largely empirical and black-box: they report aggregate error metrics without explaining why errors arise or when they will grow. We propose a structure-preserving framework that makes OOD generalization predictable and auditable. Our key idea is to parameterize the learned solution operator as a spectral filter $h_θ(λ)$ acting on the eigenvalues of the underlying elliptic operator, implemented via Chebyshev polynomial expansions and trained with a weak-form objective. This parameterization admits an exact decomposition of the energy-norm error into two observable components: a model-dependent spectral approximation term and a distribution-dependent spectral weighting term induced by the input. From this decomposition we derive three diagnostics: a conservative in-band supremum $\vareps_{\mathrm{sup}}$, a global RMS proxy $\vareps_{\mathrm{rms}}$, and a sample-dependent effective metric $\vareps_{\mathrm{eff}}(f)$. These diagnostics can be computed without access to ground-truth solutions. Through four controlled experiments, we show that $\vareps_{\mathrm{eff}}(f)\|f\|$ consistently predicts energy error under in-distribution, in-band spectral shift, out-of-band tail, and compound shifts, whereas global metrics can be systematically misleading. Our framework shifts OOD assessment of neural operators from black-box benchmarking to operator-structure diagnostics, providing a practical route to auditable scientific machine learning.

Residential traffic classification supports service management, but learning across homes must account for heterogeneous traffic and privacy constraints. Privacy-aware training may impose uneven costs across traffic categories. We study this tradeoff in simulated two-client federated learning using 1.62 million preprocessed gateway-collected flows across six categories. We compare a full-feature baseline, feature suppression (FS), and differentially private stochastic gradient descent (DP-SGD) under one fixed record-level privacy setting. FS-mild excludes four timing features from 16 model inputs; it provides no formal privacy guarantee. With size-proportional aggregation, FS-mild achieves higher combined macro-F1 and worst-group F1 (the minimum per-class F1 across homes) than DP-SGD in all five seeds at both model capacities under stratified and temporal splits. The tested DP-SGD configuration incurs pronounced minority-category losses, especially in the smaller home, but FS-mild does not uniformly improve on the full-feature baseline. On stratified-split models, loss-based and shadow-model membership probes show near-chance aggregate discrimination without a consistent ranking across probes; this does not establish equivalent privacy. These findings support FS as an input-minimization baseline, not a substitute for formal privacy.

Detecting RTL design bugs in open-source RISC-V CPU implementations is critical for ensuring system reliability. Traditional detection approaches inherently rely on predefined artifacts. In this paper, we leverage the official,natural-language RISC-V specifications as an effective information source for bug detection. We present VSpector, a specification-driven bug detection pipeline that directly checks whether CPU register-transfer level (RTL) implementations adhere to official specification rules, without requiring specialized construction of reference models, formal properties, or custom bug patterns. To resolve the key technical trade-off between broad context scope and model reasoning accuracy when using Large Language Models (LLMs), VSpector employs a stepwise context refinement scheme across a four-stage pipeline: rule extraction, implementation localization, candidate identification, and sequential violation auditing. We evaluate VSpector on two industrial-strength RISC-V CPUs, CVA6 and XiangShan. Out of 217 reported candidates, manual inspection confirmed 148 true violations, representing a 68.2% precision. These violations correspond to 73 distinct bugs, including 42 previously unknown bugs. In our comparative experiments, DiveFuzz, a state-of-the-art CPU fuzzer, detected none of these new bugs during 24-hour runs per CPU. All 42 new bugs have been reported upstream, with developers already fixing 19 and confirming an additional 11 (30 in total), demonstrating that specification-driven auditing is a practical and complementary strategy for CPU bug detection.

Structural causal models for time series recover contemporaneous and lagged effects, but most methods require complete observation windows and become misspecified when samples are missing. We introduce ITSY, the first continuous-optimization method for causal discovery from irregular time series under a linear model. ITSY reformulates the structural equation so that prediction uses the nearest available history rather than the possibly missing current slice, and jointly imputes missing values while learning both graphs. A weighted reconstruction objective corrects the noise transformation induced by this reformulation. Across synthetic regimes varying missingness, scale, graph density, and noise, and on a real world benchmark, ITSY consistently improves graph recovery over representative SCM-based baselines, demonstrating the effectiveness of the proposed method. The results establish a focused solution for irregular linear first-order dynamics and clarify the assumptions required for nonlinear or higher-order extensions.

Large language model agents are typically deployed with predefined configurations, although the required model capability, context, tools, permissions, memory, and computational resources can vary substantially across tasks. This study develops and evaluates AgentBetta, an adaptive AI Nano-Agent framework that represents these factors as an executable configuration and updates them through verification-driven diagnosis, selective expansion, and verification-based counterfactual contraction. The evaluation distinguishes controlled mechanism validation from external agent comparisons. On the AB-ConfigBench benchmark, AgentBetta achieved 91.38% verified success while reducing median context allocation from 64,000 to 8,000 context characters and median tool exposure from five tools to zero compared with the fully provisioned configuration. The configuration-deficiency diagnosis achieved a macro-F1 score of 0.819 with precision of 1.000 across the evaluated dimensions, and selective expansion avoided unnecessary changes to unrelated configuration dimensions. Post-success contraction preserved verification outcomes in 56.41% of evaluated one-dimension contraction probes, indicating that some successful configurations contained removable capability under the tested conditions. External evaluations indicate that adaptive configuration can improve the balance between verified task completion and capability exposure; however, the results vary across benchmarks and agent families. In particular, the cross-family replication did not reproduce the primary-backbone accuracy ordering, and specialized systems remained advantageous for certain task domains. These results support interpreting AgentBetta as a configuration-adaptation mechanism that regulates capability allocation and inference expenditure rather than as a universal replacement for specialized agent architectures.

As part of HVAC systems, residential heating provides foundational infrastructure for human habitation in cold weather. However, research on how residents interact with HVAC systems, particularly heating systems, remains fragmented across architecture, engineering, informatics, physiology, psychology, sociology, and design. Based on 541 studies from these fields, this review integrates interdisciplinary research on Heating in Human-HVAC Interaction in smart homes.The resulting synthesis is conceptualized through the Situated Interaction Dynamics of control and feedback between users and systems. User-initiated interactions involve monitoring past and present system performance and planning future operation, while system-initiated interactions rely on sensor networks to trigger automation or provide information enabling user action. These interaction dynamics connect Residents' Experience and Practices with Heating in HVAC System Mechanics. Residents' Experience and Practices include thermal comfort and energy management, where thermal comfort involves both individual physiological and psychological experiences of indoor climate and social practices shaped by norms, empathy, and negotiation among cohabitants. Heating in HVAC System Mechanics includes thermal conditions and energy performance. Thermal conditions concern the regulation of air temperature, mean radiant temperature, air velocity, and relative humidity, while energy performance concerns efficiency and environmental impact. This overview highlights four interdisciplinary tensions: sensed versus lived conditions, personalization versus negotiation, efficiency versus health, and automation versus agency. The resulting framework offers a conceptual lens to interpret heating interactions and design Human-HVAC Interaction that balances IEQ-driven healthy thermal conditions, affordability, and sustainability.

Perception for embodied agents is video-based, often multi-view (ego, exo, or both), and inherently continual, with simultaneous task and viewpoint shifts. Yet continual learning (CL) remains dominated by exo-only recognition tasks, obscuring behavior under these real-world coupled shifts. We introduce Continual Ego, E}xo, and Ego-Exo Learning (CE$^4$L), a unified multi-view CL benchmark spanning four representative tasks: cross-view referenced skill assessment, temporal action segmentation, cross-view association, and action anticipation & planning. CE$^4$L highlights challenges largely absent in prior CL benchmarks, including cross-view correspondence, view-dependent asynchrony, and heterogeneous semantic objectives. To this end, we propose Video Incremental Subspace-routed Task Adapters (VISTA), a parameter-efficient baseline method that stores task-specific updates in lightweight adapters and performs training-free routing via residual distance to task-specific whitened subspaces estimated from second-order statistics. Extensive experiments demonstrate the significantly varied efficacy of representative CL methods across CE$^4$L settings, while VISTA is consistently competitive and achieves state-of-the-art overall performance. Our source code for benchmarks and methods is available at https://github.com/AnAppleCore/CE4L .

Large language model (LLM) agents increasingly execute multi-step workflows through tool use and interaction with users and environments. However, current agent evaluations are largely English-centric, limiting our understanding of agent capabilities in multilingual settings. We introduce BabelFlow, a benchmark-general agentic workflow that adapts existing agent benchmarks to new languages by analyzing runtime dependencies, coordinating structure-preserving translation, and combining multi-layer verification with human review to preserve task and evaluation semantics. Using BabelFlow, we construct BabelArena, a task-aligned benchmark comprising 16,146 instances derived from 702 canonical tasks across four benchmark families, 13 domains, and 23 languages. Experiments with five frontier models show that no single model dominates across benchmark families and that cross-language disparities extend well beyond task success. Lower-resource languages exhibit distinct failure patterns, with larger shares of tool-use and control-flow errors rather than answer-quality errors alone, pointing to gaps in reliable task execution across the resource levels of these languages. On the same tasks, agents in low-resource languages also consume substantially more tokens than in English (up to roughly twice the input) without proportional increases in interaction length, and language consistency degrades further on tasks requiring structured output, where switches are directed overwhelmingly toward English. We believe BabelArena provides a foundation for advancing research on reliable and efficient multilingual agents.

A potential path forward is Quantum Machine Learning (QML), which aims to leverage quantum computing in conjunction with classical machine learning to enhance computing efficiency and the expressiveness of models. In this paper, two different quantum classifiers - Variational Quantum Classifier (VQC) and Quantum Kernel Support Vector Machine (QSVM) - are compared with three classical classifiers as baseline classifiers - Logistic Regression, Support Vector Machine (SVM), and a Multi-Layer Perceptron (MLP) - on the Breast Cancer Wisconsin dataset. The quantum circuits were created in the PennyLane framework and simulated on a classical backend. However, in terms of accuracy, classical Logistic Regression performed better with an accuracy of 97.8%, classical SVM and QSVM with an accuracy of 95.6% each, although the Quantum VQC achieved a lower accuracy of 88.9% and had a recall of 100% for the benign class, though it correctly identified only 12 of the 17 malignant cases (a malignant-class recall of approximately 70.6%). The drawback of quantum models is the higher training time; however, since the quantum circuit needs to be classically simulated, the quantum SVM took 23.29 seconds compared to less than 0.01 seconds for the classical linear models. These results indicate that for small structured datasets, classifiers based on quantum computing have not yet surpassed well-tuned classical counterparts. In some respects (e.g., benign-class recall), they perform competitively, though not on malignant-class recall, where the VQC in particular performed worse than the classical baselines, which is worth further investigation on real quantum computers.

Long-running LLM agents require memory that persists and evolves across sessions. Text-based memory retrieves and reconstructs past interactions at every query, making long-horizon performance increasingly dependent on retrieval quality and contextual reasoning as histories grow. Parametric memory encodes experience directly into model computation, but existing approaches provide limited support for cross-session memory evolution. Their coupling to a specific backbone further restricts memory reuse after model replacement. We introduce RPMem, a two-stage architecture that compiles each session into a model-independent latent memory through forward computation and selectively integrates it with retained memory via a task-trained recurrent gate. The consolidated memory is then mapped to backbone-specific low-rank adaptation (LoRA) parameters, allowing the encoding capability to transfer when the backbone is replaced. Evaluation across three long-term memory benchmarks and five diverse backbones demonstrates broad generalization with near-constant update cost and memory footprint. With Qwen3-8B on PERMA, RPMem reaches 85.52%, outperforming the strongest parametric and text-based baselines by 5.32 and 12.98 percentage points, respectively. Ablations validate the complementary roles of session compilation and cross-session consolidation, while dynamics analyses reveal that the gate acquires task-specific memory integration strategies. These results establish RPMem as a lifecycle-independent parametric memory framework that maintains evolving cross-session memory that remains reusable across backbone replacements. Our implementation is available at https://github.com/Quark-Medical/rpmem/tree/main.

Long-horizon conversational memory is especially challenging in multi-actor settings, where relevant evidence is distributed across participants and contexts and previously established information may later be revised. We introduce EGMEMORY, which formulates long-horizon multi-actor memory as a searchable state machine that separates persistent message-level evidence from an explicit active state. At write time, adaptive state resolution and an evidence-grounded propose-verify-commit protocol govern how this state evolves. At read time, adaptive evidence navigation iteratively resolves the state and supporting evidence required for a query, using conversational structure to narrow the search space and lexical-semantic relevance to rank candidates. The system operates through prompting and tool use without memory-specific policy training. EGMEMORY achieves 68.2% on GroupMemBench and 77.9% on EverMemBench, outperforming the strongest evaluated baselines by 22.7 and 21.4 percentage points, respectively. It further reaches 73.6% on the dyadic LoCoMo benchmark, demonstrating generalization beyond multi-actor conversations. We will release the codebase upon formal publication.

Federated learning lets multiple parties train a shared model without pooling their data, but a client with far less data than the others can end up poorly served even when the group's average accuracy looks fine. We ask whether the choice of pretrained backbone affects this under LoRA fine-tuning, and whether per-word perplexity on the target text predicts which backbone helps the worst-off client before federated training starts. We ran 313 experiments across three text-classification datasets and three similarly sized backbones (RoBERTa, BERTweet, PubMedBERT), each compared against a task-specific baseline on identical data splits. Lower-perplexity backbones consistently produced larger gains for the worst-performing client, with a rank correlation of -0.87 across nine dataset-backbone pairs; a backbone held out of the analysis confirmed the pattern. Personalization with Ditto recovered only 4-12% of the gap between training alone and full federation, and removing aggregation entirely erased the benefit. A client's update also showed no sign of conflicting with the group's update; the two are close to orthogonal, ruling out one proposed explanation for this failure. Practically: measure perplexity on a sample of task text before choosing a backbone, and do not rely on personalization to protect a data-poor client. We release our code, predictions, and full results for others to test.

Non-verbal vocalizations (NVVs) carry important paralinguistic information but are often omitted by conventional automatic speech recognition (ASR) systems. The ISCSLP NVVSpeech Challenge requires joint transcription of lexical content and 16 NVV categories under limited and highly imbalanced supervision. We present a data-centric NVV-aware ASR pipeline based on cross-dataset label harmonization and a two-stage sampling schedule. We map heterogeneous source labels to the official taxonomy and exclude samples without a reliable mapping. Our schedule first uses square-root category sampling to moderate the long-tailed distribution and then applies uniform-category fine-tuning. On a fixed local validation split, square-root category sampling performs best among the tested single-stage settings. The final two-stage system obtains an official score of 63.86 and ranks fourth in Track 1.

Deep learning has advanced automated electrocardiogram (ECG) diagnosis, but the field's most accurate models, foundation models pretrained on millions of recordings, are not decision-pathway auditable: a clinician cannot trace a diagnosis to a physiological pathway or intervene on one. We propose TRACE, a Tractable Routing Autoencoder for Clinical ECG, whose 32-dimensional clinical latent space is specified in advance from domain knowledge rather than discovered by optimization. TRACE partitions this space into perfusion, structure, and conduction subspaces, routes each to its own diagnostic head by design, regularizes the partition with an orthogonality penalty, and reconstructs the ECG through a decoder that permits latent perturbation. On PTB-XL and Georgia, TRACE exceeds unconstrained classifiers and stays ahead of an ECG foundation model pretrained on ten million recordings, evaluated by linear probe on frozen features, at roughly an eighth of the parameter count. On the nine-label CPSC2018 cohort, which carries no structural class, the framework transfers with only the routing table re-specified to a perfusion/rhythm/conduction partition. Joint probe, erasure, and perturbation analyses verify the routing contract, and perturbing the depolarization and repolarization pathways modulates the reconstructed waveform. Removing the specified partition and its orthogonality penalty costs 1.70 AUC and 11.30 macro-F1 points on PTB-XL, and 2.76 AUC and 16.92 macro-F1 points on Georgia. A capacity-matched permutation control places arbitrary assignments within 0.34 AUC points of the ontology routing and leaves macro-F1 statistically level (p=0.619): the ontology supplies decision-pathway auditability at no macro-F1 cost.

Reinforcement Learning with Verifiable Rewards (RLVR) is expanding from tasks with well-defined correctness signals, such as mathematics and code, toward multifaceted quality requirements specified by multi-dimensional rubrics. Since policy optimization consumes one scalar per rollout, rubric-based pipelines must map multiple criterion scores into a scalar reward. This aggregation is often treated as score scaling, but it implicitly determines how quality dimensions trade off during training. The prevailing practice, normalizing each criterion and taking a linear combination, assumes that cardinal score differences are comparable across criteria and that gains on one criterion compensate for failures on another; both assumptions are unreliable when criteria are semantically heterogeneous. We propose Reinforcement Learning with Verifiable Rubric-based Ranking (RLVR$^2$), a verifiable ranking paradigm for rubric-based RLVR. For each criterion, RLVR$^2$ converts rubric scores into criterion-specific within-group ordinal outcomes, recovers a latent utility from the resulting comparison matrix, and merges these utilities into one training signal. By retaining only within-group ordering and discarding raw score magnitudes, RLVR$^2$ avoids calibrating heterogeneous rubric scales. It further supports objective-preserving attribute adjustment: auxiliary attributes that correlate with observed rankings but are not training objectives can enter the estimation without expanding the rubric or rewarding them directly. Across three model scales and 16 benchmarks, RLVR$^2$ consistently outperforms representative rubric-based baselines, achieving the best overall performance on most benchmarks at every scale. Analysis shows it controls systematic effects tied to reasoning efficiency and response formatting while preserving the quality objective.

Memory systems are becoming a core component of LLM agents, but constructing and maintaining memory remains expensive because it relies on repeated calls to large proprietary language models. This cost creates a major barrier to deploying memory-enhanced agents at scale. In this paper, we present Pseudo Self-Distillation (PSD), a framework that enables small language models (SLMs) to construct hierarchical memory representations by distilling behavior from a strong black-box oracle through a multi-stage training pipeline. Standard distillation methods require access to teacher logits or hidden states, which closed models do not expose. Unlike conventional self-distillation settings, where supervision is derived from a model's own predictions, sampled rollouts, or aggregated outputs, PSD enables a single-model distillation setup while channeling external oracle knowledge through the prompt. PSD uses a single small model in two roles: a teacher that sees a privileged prompt containing the oracle's answer as reference context, and a student that sees only the task prompt. The student learns to reproduce the teacher's output distribution, absorbing oracle-guided behavior into its own weights without accessing the oracle's internals. On LoCoMo, PSD-trained Qwen3-0.6B, 1.7B, and 4B match or exceed GPT-4.1-mini on downstream retrieval at a fraction of the deployment cost, with off-policy PSD achieving the strongest results across most conditions. We further show that this memory-construction capability transfers out-of-distribution to LongMemEval, despite the students being trained exclusively on LoCoMo with no exposure to LongMemEval data.

Engineering change order (ECO) is an important step in repairing timing and electrical violations during the late stages of chip design. Existing Agentic EDA methods primarily focus on tool invocation, with less attention to model decision quality and targeted training. A central challenge in ECO is multi-round decision-making: the model must use the results of each round to determine the next repair action. We propose WaveletECO, which integrates a closed-loop execution platform with large language models to enable agents to execute ECO decisions effectively. We also train a local 9B model through supervised fine-tuning and CPO-SimPO using execution demonstrations and decision-preference data, enabling ECO decision-making with a locally deployed model. Across 594 evaluation runs on 22 designs, WaveletECO-Policy (BF16) and (INT8) score 79.63 and 79.65, respectively, compared with GPT-6 Astra's 77.44. The estimated inference cost of INT8 is about 1/147 of GPT-6 Astra's. These results show that specialized model training supports effective, low-cost multi-round ECO repair, with repair quality retained under INT8 quantization.

Diffusion models generate high-quality images, yet often violate the physical laws governing mirror reflections. Reflections often suffer from geometric aberrations, including positional offsets, directional misalignment, proportional imbalance, and structural distortion. These failures remain evident even in contemporary state-of-the-art generative systems. Existing methods itigate this problem through synthetic data scaling or auxiliary depth conditioning, yet their merely reliance on latent-space noise reconstruction losses as implicit supervision prevents direct enforcement of reflection-specific geometric and perceptual constraints. To bridge this gap, we present PhysReflect, a geometry and perception guided diffusion framework that decodes the predicted clean latent into pixel space at each training step and applies annealed supervision through two complementary differentiable objectives. The Geometric Loss enforces mirror-induced spatial consistency through sparse epipolar correspondence and dense boundary projection alignment, where a SAM2-based TwinTrack mechanism provides stable in-mirror localization for boundary-aware supervision. The Perceptual Loss preserves reflected appearance by combining Semantic Consistency Loss, which maintains reflected identity and appearance via DINOv2 features, and Lighting Consistency Loss, which regularizes depth, surface-normal, and illumination coherence under monocular geometry priors. Experiments on synthetic and real-world benchmarks show that PhysReflect outperforms prior mirror-reflection methods in geometric, perceptual, and physical-plausibility metrics, as well as qualitative visual results.

Multi-omics sequences contain complex biological patterns, yet deciphering their mechanisms for automated scientific discovery remains challenging. As large language models (LLMs) interpret these sequences, evaluating both predictions and scientific reasoning is critical. However, existing benchmarks for multi-omics sequence tasks rely on classification and regression metrics, neglecting whether models grasp the underlying biological evidence. We introduce OmicsBench, the first reasoning benchmark for multi-omics sequences, comprising 1,160 expert-validated questions across six tasks spanning DNA regulation, RNA processing, and protein function. OmicsBench requires traceable evidence chains, evaluated using instance-specific rubrics developed with domain experts. Evaluating 17 LLMs reveals an inverse relationship: while scientific LLMs outperform general-purpose LLMs in sequence classification accuracy, they fail to provide valid evidence to support their predictions. One plausible interpretation is shortcut learning: specialized models may rely on statistical patterns rather than the biological mechanisms needed for scientific discovery. Motivated by this finding, we introduce tool-augmented on-policy distillation (TA-OPD), a post-training method to align sequence prediction with evidence-grounded biological reasoning. Across five Qwen3.5 models spanning 0.8B to 27B parameters, TA-OPD consistently strengthens biological evidence grounding while improving predictive performance on most tasks. These gains persist across model scales, indicating that stronger sequence reasoning does not arise solely from increased model capacity, but can be improved through evidence-aware training. Together, OmicsBench and TA-OPD provide a framework for diagnosing reasoning failures in multi-omics LLMs and a path toward models whose predictions are better grounded in biologically meaningful evidence.

Vision-language navigation (VLN) has largely been developed for indoor and terrestrial robots, where language can often be treated as a static goal and motion is approximated by discrete or near-instantaneous actions. These assumptions break down for unmanned surface vehicles (USVs): river navigation requires continuous motion under inertia and limited maneuverability, while long-horizon instructions must be executed through sparse and visually ambiguous maritime landmarks. We introduce RiverVLN, to our knowledge the first benchmark designed for long-horizon USV VLN under continuous riverine motion, and PGT-NAV, a phase-grounded temporal navigation framework for USVs. Rather than directly mapping an entire instruction to motion, PGT-NAV converts it into an ordered sequence of visually verifiable semantic phases and maintains the active phase online through grounded visual and motion evidence. This explicit semantic progress state is fused with visual-motion history and phase-specific grounding to predict six local SE(2) pose increments. The resulting trajectory is executed in a predict-execute-re-observe loop, where the vessel executes toward W3, updates phase and grounding, and replans through a map-based safety layer. Experiments show that PGT-NAV substantially reduces recursive position and heading drift relative to GNM-style and ViNT-style baselines and achieves an average success rate of 0.79 in Unity-ROS closed-loop navigation. Unseen bridge-opening trials and real-world USV experiments further demonstrate that the phase-grounded representation transfers from controlled evaluation to physical USV deployment.

Standard data center sustainability metrics, including Power Usage Effectiveness (PUE), Water Usage Effectiveness (WUE), and Carbon Usage Effectiveness (CUE), measure a facility's resource use and emissions intensity, normalized to IT energy use, without directly representing local resource scarcity, infrastructure capacity, or social footprint. This gap has become politically consequential. In the first quarter of 2026 alone, local opposition delayed or canceled roughly $130 billion in projects across the United States, driven overwhelmingly by recurring concerns over water use, power demand, infrastructure capacity, and transparency rather than internal efficiency, matching the total for all of 2025 [11]. We propose a five-category local impact audit framework covering efficiency, water stewardship, carbon and renewables, regulatory compliance, and local disclosure. The framework is designed for recurring quarterly assessment and independent verification against public records. We illustrate its application using publicly available data from three Illinois facilities that are currently at the center of local policy disputes, and we examine the data-access barriers that constrain independent verification. We position this framework as both a research contribution and a practical instrument for county-level policymakers evaluating data center permitting and moratorium decisions.

Screen-recorded web editing requests contain weak deictic expressions such as ``this'' and ``there,'' whose referents depend on speech, cursor trajectories, page state, and edit history. Such requests require intent recovery beyond the explicit specifications assumed by many existing web-editing benchmarks. We introduce Omni2Web, a bilingual benchmark of 918 instances spanning 13,907 edit steps. It defines three complementary tracks: Direct Editing evaluates webpage editing from recordings, Instruction Recovery measures explicit intent recovery, and Instruction Utility tests whether recovered instructions can drive a fixed code executor. We evaluate 17 open- and closed-source models. The best models attain 51.17 on the Edit Fidelity Score (EFS) for Direct Editing and 49.14 on the Instruction Recovery Score (IRS); under the fixed executor, the strongest recovered instructions reach 51.08 EFS, still far below the 89.69 EFS obtained with oracle instructions. Step-level analyses show that correct grounding does not guarantee successful edits, while some Omni models recover instructions that the fixed coding model executes substantially better than their direct edits. Controlled ablations further demonstrate the value of temporally aligned audiovisual evidence, while alternative judges preserve the leader and broad ordering. Together, these findings reveal substantial headroom in multimodal intent recovery and code execution and highlight the promise of pairing Omni rewriters with coding models.

Long-form audio performance is often summarized by context length and aggregate accuracy, obscuring how language, evidence, and task jointly shape difficulty. We introduce MuLA-Bench: 5,038 open-ended questions over 1,769 in-the-wild recordings totaling 1,377.9 hours, covering 16 languages and eight domains. A balanced Language x Domain semantic track supports controlled comparisons, while a complementary acoustic track preserves naturally occurring non-speech evidence. Evidence-grounded generation, shortcut checks, and language-expert review provide auditable questions without translating a shared source set or injecting target sounds. We evaluate ten audio-language models and conduct pooled diagnostics on a fixed eight-model cohort. Language rankings change across domains and tasks; acoustic-semantic performance gaps vary with the requested operation; and temporal errors can persist after the correct event is identified. Long-range retrieval is comparatively strong, while precise clock alignment and factual grounding of natural acoustic events remain fragile. MuLA-Bench thus exposes conditional failure patterns that a single long-context score does not capture.

Humans can effortlessly localize the direction of a sound source and integrate it with visual cues for reasoning, yet this remains challenging for embodied agents. In particular, it is still unclear how to effectively evaluate and model spatial audio understanding in embodied settings. To address this gap, we introduce \textbf{OmniEchoBench}, a unified benchmark for spatial audio-visual perception and audio-vision-language navigation. OmniEchoBench comprises six tasks over 197 real-world spatial audio-visual scenes, 2,972 question-answer pairs, and 900 navigation samples with first-order ambisonics (FOA) audio collected from 30 real-world environments. To enable scalable training supervision, we develop a controllable rendering pipeline for spatial audio. It preserves geometric consistency among sound sources, visual observations, and agent trajectories. Building on this, we propose \textbf{OmniEcho}, a spatially aware omni-modal model. It introduces an FOA spatial encoder alongside a pretrained semantic audio pathway. Extensive experiments show that OmniEcho achieves state-of-the-art performance on spatial audio-visual perception. For our sound-guided navigation, OmniEcho reaches a performance level close to that of traditional vision-language navigation. These results demonstrate that spatial audio can serve as a valuable signal for embodied scene reasoning and navigation, while also highlighting fine-grained spatial localization and distance estimation as important open challenges.

Shrimp diseases continue to cause devastating losses in the aquaculture industry, driving a critical need for robust, automated detection. This work contributes the first application of Vision Transformers (ViT) and Self-Supervised Learning (SSL) to the shrimp farming domain, addressing both performance bottlenecks and data labeling challenges. We propose two deep learning pipelines to classify four key diseases: Healthy, Black Gill (BG), White Spot Syndrome Virus (WSSV), and a co-infection of both using a dataset of 4,348 images. First, our supervised transfer-learning approach leverages ImageNet-pretrained ViT-Small/16 and EfficientNet backbones. Second, we introduce a contrastive learning framework (SimCLR) with a ViT-Small encoder to extract robust representations from unlabeled images prior to fine-tuning. Our results establish strong new baselines for sustainable aquaculture monitoring. The supervised approach achieves an outstanding 96% accuracy with fast convergence, outperforming traditional generic models, while the label-efficient SSL approach reaches a highly competitive 85% validation accuracy.

Before a machine learning model can learn a thermodynamic equation of state, it must discover what its measurements represent: which channels scale with system size, which are intensive conjugates, how sectors pair through contact, and which potential governs stability. When sensors expose only an unknown linear mixture of extensive states and intensive responses, passive observations cannot disentangle physical quantities from coordinate artifacts. We formulate the problem of discovering this hidden thermodynamic ontology directly from anonymous controlled experiments. We present an operational identifiability theory and a constructive polynomial-time algorithm that extracts extensive and intensive scaling sectors from replication contrasts, recovers their dual cotangent pairing from thermal contact and reciprocity, verifies a globally admissible concave potential via discrete cyclic concavity, and determines an invariant matroid of reservoir ensembles. We prove that the residual observational equivalence is strictly (x, lambda) ~ (A x, a A^{-T} lambda + beta), establishing the sharp observational limit that no permitted experiment can break. Blind evaluations on van der Waals fluids and Curie-Weiss magnets confirm robust recovery under ill-conditioned mixing, correctly resolving anonymous Maxwell tie-lines while rejecting non-equilibrium continuations. External validation across six real fluids from the NIST WebBook demonstrates that operational ontology discovery transfers across real physical substances without coordinate leakage.

Data acquisition (DAQ) systems at future particle physics experiments stand to benefit from the extremes of AI/ML development: large-scale foundation models can enhance the performance of feature extraction algorithms, and small-scale on-detector deployments can enable real-time intelligent data handling. This work provides the first fine-tuning of an industrial foundation model for particle physics DAQ. Starting from the backbone of Google Research's TimesFM (Time Series Foundation Model), we demonstrate fine-tuning on real-time regression tasks for drift chamber trackers and dual-readout calorimeters. Furthermore, the fine-tuned TimesFM model is distilled into a student and co-designed with FPGA implementation to enable these models to run in real-time at future colliders. The fine-tuned distillations meet or exceed the performance of previously published AI/ML solutions for each task. Further, the pipeline of distillation and model compression from TimesFM is generic and can be easily adapted to a variety of 1D waveform tasks across domains.

Bayesian filtering provides a principled framework for online state estimation under uncertainty, yet its application to systems with high-dimensional states and complicated posterior distributions remains challenging. Recent generative models, such as flow matching, have shown potential in Bayesian filtering. However, they still rely on particle-based representations of the posterior, which lose the rich information of the full distribution, or tackle a trajectory-level inverse problem that conflicts with the recursive structure of Bayesian filtering. To address this, we propose a new perspective of directly encoding the evolving distribution into flow matching model weights, namely, the Belief Flow Filter (BFF). It is a generative filtering framework that updates model weights via gradient descent at test time to track the posterior evolution. Thereby, BFF bypasses the scalability issue of particle representations or the flexibility limitation of Gaussian assumptions in conventional filters. We theoretically justify that the BFF design is structurally aligned with Bayesian filtering, and its training objective targets the recursive filtering operator. BFF is empirically verified across 5 different physical systems, including ones with chaotic dynamics and highly sparse, non-linear observations. The results show that BFF attains the best score in 8 of 9 metric-benchmark cells across the three standard 1D and 2D PDE benchmarks, and similarly leads on the extreme single-moving-sensor setting and on a real-world-grounded tokamak plasma estimation task, demonstrating its potential to accurately approximate the Bayesian filtering operator in high-dimensional probability space.

Before a machine can discover a physical law, it must discover what its measurements are: which observations live on cells, which are intensive or extensive, which sectors are dual, and which distinctions are merely gauge. We introduce physical representation-language discovery, the problem of recovering this hidden ontology directly from anonymous controlled experiments. We give an identifiability theory and constructive polynomial-time procedure that recovers a carrier and differential sequence, measurement types and orientation twist, noninvertible refinement semantics, primal-dual Maxwell diagrams, and the residual equivalences that no permitted experiment can break. The theory turns material nuisance into a commutant, uses refinement to separate quantities from coordinates, and selects physics only after its representation has been recovered. For a certified finite experiment family, we prove an end-to-end two-stage measurement bound and a matching minimax rate in dimension, accuracy, and confidence. Blind Maxwell experiments recover complete primal/relative-dual ontologies on regular and unstructured carriers under jointly corrupted observations; an independent unstructured RLC system demonstrates that the result is not specific to Maxwell. The framework scales to tens of thousands of cells per carrier, while stress audits demonstrate robustness across severe physical regimes - including non-Markovian memory, nonlinearities, nonlocality, and complex constitutive hysteresis. A public FDTD audit demonstrates the emergence of anonymous curl structure from incomplete field data, while characterizing the informational prerequisites for complete recovery. The goal is to move scientific ML from learning laws in a human-supplied language to discovering the language in which laws become expressible, establishing exact theoretical limits on observational identifiability.

We introduce leaky-integrator reconstruction, a training-free method that cures the error accumulation of recursive differenced forecasting. Our first contribution is diagnostic: predicting one-step changes and integrating them by cumulative summation, the standard remedy for non-stationarity, is a discrete integrator with a pole on the unit circle, and we show this makes recursive rollout of a nonlinear model diverge, its 336-step error reaching several times that of a well-behaved forecaster (normalised MAE 1.6-3.8 versus about 0.8) across every neural architecture tested. Our second, central contribution is the fix: move the pole inside the unit circle with a leaky integrator H(z) = 1/(1 - gamma z^-1), gamma < 1, which provably bounds the accumulated error variance. Applied at reconstruction time with a single fixed gamma=0.9 (no retraining, a two-line change to any deployed one-step or foundation-model forecaster), it shrinks error at every horizon, the mean gain over seven diverging architectures and twenty datasets growing from ~3% at H=24 to 23% at H=96, 37% at H=192 and 51% (43-74% across those architectures) at H=336 (78% with an oracle pole). Crucially, it is provably inert where no pathology exists (stable or joint predictors already at the irreducible rate), making it a safe, general default.

Repository-level software engineering (SWE) comprises heterogeneous task categories, whose progress under pooled agentic reinforcement learning can be uneven: gains in some categories coincide with regressions in others, while aggregate resolution obscures these changes. Motivated by this category see-saw, we develop a category-aware expert-training and policy-integration framework. Executable task construction and SWE Labeler, an evidence-grounded multi-axis labeling system, organize the training pools. Initial category-specific RL improves average training success while leaving uneven instance-level progress, motivating explicit consolidation of successful behavior and policy-adaptive task selection. Same-origin category experts alternate long-horizon Agentic-miniRL with Refresh-Repair-Expand (RRE): the updated policy refreshes instance mastery, reuses its own verified successful trajectories for Repair SFT, and reselects tasks for further RL. Label-routed multi-teacher on-policy distillation (MOPD) consolidates the experts into one deployable student, with ReLU-gated reward extrapolation keeping only each teacher's improving direction over the reference. Expert training and policy integration require no external model to provide solution trajectories or action targets. We evaluate Pooled RL and Balanced RL, expert development, and single-model integration through aggregate and per-category resolution, the minimum category lift over each joint-RL baseline, and expert-gain recovery. The final MOPD policy achieves mean resolution of 58.04% on Pro-618 and 59.00% on SWE-bench Multilingual, improving over the base model by 5.39 and 2.78 percentage points, respectively.

Open-vocabulary audio-visual event localization (OV-AVEL) grounds a text-queried event in time from video, audio, and language. The supervision sources available to this task can differ in temporal-boundary reliability: on OV-AVEBench, our configured visual teacher gives more reliable boundary cues than the configured audio teacher, although the latter is a strong pretrained audio model and remains semantically informative. This is a setting-specific diagnostic rather than a universal ranking of vision and audio. We formulate the resulting challenge as supervision placement: which teacher signals may shape the localization decision, and which should remain auxiliary. Based on this view, we propose OV-OrthKD, a reliability-aware asymmetric distillation framework. Visual feature transfer shapes a decision-aligned representation, audio feature transfer enriches a complementary auxiliary subspace, a text prototype anchors seen/unseen category semantics, and an orthogonality loss limits directional overlap between the two teacher-specific projections. The student continues to use both modalities through query-aware fusion at inference, while the default training recipe keeps audio-teacher supervision off the segment-logit path. On OV-AVEBench, OV-OrthKD achieves 0.816 segment AP and improves F1@0.5 over the official fine-tuning baseline by 2.7 points overall and 3.4 points on unseen categories. Path-assignment, role-swap, corruption, and transfer analyses consistently support supervision placement as a task-specific design axis for OV-AVEL.

Reinforcement learning (RL) offers a natural language for healthcare decisions whose conse- quences unfold over time, yet most reported progress remains far from routine intervention. Ex- isting surveys organize the field by algorithm or clinical application. We instead review healthcare RL through an evidence ladder: problem formulation, retrospective identification, policy estima- tion, stress testing, prospective evaluation, and lifecycle monitoring. This view connects clinical treatment, patient engagement, and health-system operations while exposing a recurring gap: evi- dence that a policy scores well in a historical dataset is not evidence that it will improve care. We synthesize the assumptions and failure modes at each rung, identify what evidence can and can- not transfer across settings, and propose reporting practices for cumulative evaluation. Restless bandits are included as one special case, not as the organizing framework. The central lesson is that healthcare RL should be evaluated as an intervention embedded in a changing sociotechnical system, rather than only as an optimizer of a retrospective reward.

Long-context language models interface with external knowledge through raw natural language. In retrieval-augmented systems, this creates a persistent index-payload schism: dense vectors enable searchable routing, but models must re-ingest lengthy text payloads for reasoning at O(N^2) attention cost. Existing compression methods further produce private states tied to specific architectures. We introduce Machine-Interpretable Information (MII), the first agent-to-agent (A2A) document-to-state protocol. A dual-timescale state-space Writer compiles documents into a canonical, fixed-bandwidth state (56 tokens), and a lightweight Translator maps it into any frozen Reader's embedding space, reducing query-time cost to O(K). The resulting .mii artifact unifies Retrieval (searchable geometry), Reasoning (global memory), and Reconstruction (grounded details) in a single transferable medium. We demonstrate strong cross-model interoperability across heterogeneous LLMs (e.g., Llama, Qwen, Mistral) -- despite the Writer using a legacy GPT-2 vocabulary, forcing genuine semantic translation rather than token-level memorization. Mechanistic probes reveal modular latent structure: entity representations can be causally traced and zero-shot transplanted between unrelated document states while remaining decodable. To address lexical reconstruction under fixed bandwidth, we propose Residual-MII, a cache hierarchy combining compiled global memory with sparse local evidence. On HotpotQA (7,405 queries), Residual-MII exceeds full-context Exact Match at approximately 7% of the attention FLOPs, suggesting a paradigm shift toward compiled, transferable neural document formats.

FORM is a domain-specific symbolic manipulation language widely used in particle physics for processing the very large algebraic expressions arising from multi-loop Feynman diagram calculations. Despite its central role in precision theoretical physics, no artificial-intelligence tooling exists, to our knowledge, for assisting physicists in writing FORM code. We show that contemporary large language models (LLMs), including frontier models with hundreds of billions of parameters, achieve a zero-percent execution pass rate on our instruction-following and tutorial-style FORM tasks without documentation in a single attempt, establishing FORM as a genuine zero-shot language for LLMs at the time of writing. We then present a verification-driven data generation pipeline that uses the FORM binary itself as an execution oracle to produce and validate a corpus of 4,633 training examples spanning deterministic computations, open-ended programs, tutorial code, and knowledge question-answer pairs. Fine-tuning a compact open-weights model (Qwen3-8B) with quantized low-rank adaptation (QLoRA) yields a specialist that, evaluated on four complementary benchmarks (840 tasks, single attempt each), decisively outperforms frontier models with up to 756B parameters in execution rate and in strict, FORM-verified output matching on the larger benchmarks, and remains statistically indistinguishable from them on the smaller, harder ones. General reasoning and coding capabilities are preserved within 2.6 percentage points.

Recurrent models must preserve information that changes future behavior while suppressing hidden-state error. These objectives conflict: contraction improves stability, but contraction along a future-distinguishing direction destroys memory. We formalize this boundary through the predictive quotient of a recurrent state space. Two hidden states are equivalent when they induce the same conditional future; their equivalence classes form predictive fibers. Every exact semantics-preserving corrector acts as the identity on this quotient. At a regular point with hidden dimension d and predictive dimension k, it can eliminate at most d - k independent directions. This establishes a discrete-continuous boundary: finite predictive states admit positive-radius exact correction basins, whereas an uncountable continuum of future-distinguishable states cannot be decoded after arbitrary positive-radius perturbations in finite-dimensional Euclidean space. To operationalize this principle, we develop an auditable finite-future framework. A compact deployment bank W is evaluated against an independent audit bank A (W subseteq A) on a declared correction domain. Under generative probe access and audit-metric coverage, finite stochastic rollouts furnish a high-probability certificate for the separation margin Omega_{W|A}(delta). Preserving learned W-predictions within this certified margin guarantees bounded audit-semantic distortion. For intrinsic audit dimension k, the required probe outcomes scale as O(M * Omega^{-(k+2)}), where M = |A|; a matching minimax lower bound proves this exponent is optimal. Extending guarantees to continuous futures is achieved via an explicit completeness modulus. Controlled experiments validate the certified margins, scaling laws, and automated probe refinement under a safety-first evaluation paradigm.

Recent advances in large language models (LLMs) have led to the emergence of coding agents capable of performing complex engineering tasks, including register-transfer level (RTL) design and optimization. Existing RTL benchmarks mainly evaluate functional correctness and performance, power, and area (PPA) of the generated RTL designs, leaving agents' ability for \emph{timing closure} under-evaluated. We propose TicTacBench, a benchmark specifically designed to evaluate coding agents' capabilities for RTL-level timing closure under post-place-and-route (post-PnR) evaluation. TicTacBench contains 30 diverse tasks, each provided with a suboptimal RTL design, realistic timing constraints, functional equivalence verification, and timing reports. With over 300 runs of coding agents driven by 8 frontier LLMs, we find that even the best agent can only close 53.3\% of tasks with 7.18\% area-delay product (ADP) degradation and 8.83\% energy-delay-squared product (EDDP) improvement on average. We identify common failure categories that explain why agents fail to close timing. Then we propose TicTacSkill, a new method that guides agents to follow standard timing-closure procedures and improves the Timing Closure Rate by 9\%. These results suggest that while coding agents have made significant progress in RTL design, their timing-closure capability still has substantial room for improvement.

Engineering predictions require physical mechanisms to be translated consistently into equations, discretization, code, and validation, yet errors can propagate despite local checks. Artificial-intelligence (AI) agents automate scientific tasks, but coordinating and independently auditing the theory-to-solver process under physical constraints and human oversight remains unresolved. We introduce a human-in-the-loop, physics-constrained multi-agent workflow where human experts define admissible physics and modeling boundaries, while agents retrieve evidence, derive equations, implement solvers, and audit the theory-to-code chain. Applied to soil-plug evolution during suction-caisson installation, the workflow generated and audited 6 formulations in 2.9 h of agent execution once physical knowledge and inputs were prepared. Among these formulations, adding seepage-driven soil void-ratio evolution to the geometric baseline reduced mean absolute final-heave error from 58.4% to 9.0% across 14 profiles; the selected model further incorporated near-wall dilation and achieved mean absolute percentage errors of 12.4% across 9 final-state cases and 4.2% at the endpoints of 5 process histories. Beyond predictive performance, blinded replay recovered all 9 target problems, while an independent audit uncovered 5 implementation problems after 36 predefined checks had passed. Overall, this work extends multi-agent AI beyond task automation toward human-governed engineering solvers.

Pitch estimation on an edge device is constrained in three ways at once. The model must be small, it must stay accurate when the input is noisy, and one frame must be produced inside the frame period. In this report the Frequency Convolution Network (FrCN) of our earlier work is factored into a low-rank form. The number of parameters is reduced by 35.9%, from 17,787 to 11,397, and accuracy is not reduced, either in domain or on two corpora the model was never trained on. The range of the noise used during training is also shown to dominate the architecture in setting how the model behaves when the interference is severe. When the training noise floor is lowered from +6.02 dB to -20 dB, 0.35 points of clean RPA50 are lost and RPA50 at -20 dB is raised from 2.44 to 22.41. This effect is about two orders of magnitude larger than any architectural effect that was measured. The semi-orthogonal constraint used in TDNN-F is found to be redundant with the normalization inside the bottleneck, and accuracy is reduced when both are applied. Whether the factorization saves time depends on the runtime: in eager PyTorch the factored model is 45% slower, while in a compiled kernel it is 15% faster. For deployment, a small C kernel was written. It needs 83 kB on disk and no runtime library beyond libc and libm. It is faster than OpenBLAS on all four CPUs that were tested, faster than ONNX Runtime by 3.8 times, and faster than PyTorch by 13 times. Its output was checked against PyTorch on 271,893 held-out frames per model, and the same pitch bin was selected on every one of them.

Stochastic reconfiguration (SR) is the standard optimizer for neural quantum states (NQS), but modern NQS often have far more parameters than Monte Carlo samples. We show that in this regime the diagonal shift is more than a numerical stabilizer. It acts as a statistical filter for finite-sample generalization. At a fixed wave function, SR is ridge regression from tangent features to the centered local energy. Its residual is the expressivity gap, the part of imaginary-time evolution outside the current tangent space. This gap is orthogonal to the tangent space in population, but finite batches make it act as noise that SR can overfit. The shift therefore balances shrinkage of useful update directions against variance from fitting sampled residuals. Exact diagnostics on a $4\times4$ Heisenberg graph separate two effects of overparameterization. Larger tangent spaces help when they reduce the expressivity gap, but they can hurt when they overfit a fixed gap. In a $100$-site transverse-field Ising family trained with a foundation NQS, validation risk is U-shaped in the shift while variance decreases, matching the noisy-ridge model. This view leads to multi-shift SR (MS-SR), which averages independent ridge solves at data-adaptive shifts to form a richer, lower-variance spectral filter. Checkpoint-local experiments show that MS-SR lowers validation risk and update variance relative to the fixed-shift SR baseline. We further compare MS-SR and SR in paired online training continuations, with independent endpoint energy evaluations and a separate update-cost benchmark.

Digital direct-to-consumer (DTC) health campaigns are usually measured after the fact. In-flight forecasting commonly relies on a separate classifier for every cutoff and horizon. We treat this task as a dynamic-system problem and build a compact patient world model. The architecture maintains a latent state per patient, learns exposure-conditioned state dynamics jointly with a weekly conversion hazard, and rolls forward into future conversion curves. We evaluate it on a US campaign dataset with 147{,}173 patients and 5.2 million at-risk person-weeks. In a retrospective evaluation conditioned on recorded future exposures, the model forecasts the remaining new-to-brand prescription volume through week 52 with a relative error of 2.9\% from a week-4 cutoff and 0.8--2.6\% from cutoffs at weeks 8--26. The strongest non-recurrent baseline, a pooled-hazard gradient boosting model given the same survival rollout and information, has relative errors of 13.6--33.1\%. Per-horizon classifiers perform substantially worse. A Fisher-information analysis motivates dense next-exposure supervision when conversions are rare. Removing this auxiliary objective increases prescription-volume error by approximately $2$--$14\times$, while providing no consistent disadvantage on the more common specialist-visit outcome. We also evaluate scenario simulation. Switching all future exposure off raises predicted conversion from 0.31 to 0.89, a pattern consistent with selection effects in observational exposure data. This result highlights the limits of interpreting exposure-conditioned rollouts causally.

Single-cell foundation models (scGPT, scBERT, Geneformer) achieve cell-type classification accuracy up to 97.5% in our experiments, yet this aggregate accuracy can mask systematic failure on rare, often disease-relevant cell populations that long-tail loss functions are widely assumed to address. We present a systematic benchmark of six long-tail loss functions (cross-entropy, weighted CE, class-balanced loss, focal loss, LDAM, logit-adjusted softmax) across three architectures and three datasets (Multiple Sclerosis, Zheng68K, human Pancreas), totaling 162 controlled training runs (3 backbones x 3 datasets x 6 losses x 3 seeds). The gap between overall accuracy, Macro-F1, and rare-class recall under plain cross-entropy is consistent across all nine (architecture, dataset) settings, driven by dataset structure rather than pretraining. Rare-class failure itself splits into two regimes with distinct embedding-geometry signatures, visible before any loss is chosen: some classes are recoverable by the right loss, while others retain linear separability yet are absorbed into unrelated classes' neighborhoods under every evaluated loss and architecture. Among the recoverable classes, the efficacy of reweighting is predicted by a class's absolute training-set size, rather than its share of the dataset or the dataset's overall imbalance ratio. Class-balanced loss and LDAM are the most consistent choices across all nine settings, while logit adjustment trades rare-class precision for recall rather than improving both. Our results give both a reusable benchmark and mechanism-grounded practical guidelines for combining foundation models with imbalanced biological data.

Low-rank adaptation (LoRA) has become a key technology for serving large-scale personalized large language models and diffusion models in the cloud. However, the co-occurrence patterns, resource contention relationships, and evolutionary regularities of adapters under production inference workloads have not been systematically or quantitatively studied. Based on GenTD26, Alibaba's production diffusion model inference dataset, this paper adopts a graph-theoretic framework to construct an adapter co-occurrence network and conducts a characterization from both static structure and dynamic evolution. Our main findings are as follows. (1) The co-occurrence network is extremely sparse, and adapter usage frequency follows a significant heavy-tailed distribution. (2) Introducing the first adapter incurs a 66.1% execution-latency overhead, with diminishing marginal costs afterwards. (3) Co-occurrence relationships are driven by base models: in 90.6% of multi-adapter requests, all adapters share the same dominant base model; 66.2% of significant co-occurrence edges connect same-model adapter pairs; and in 85.8% of multi-adapter requests, all adapter pairs form significant co-occurrence edges. (4) The adapter ecosystem exhibits a core-periphery bipolar structure, with a weekly Jaccard similarity of 0.696 at the model level and a churn rate of 54.5% for the top-10 hottest models within a 12-hour window. Based on these findings, we propose a preloading strategy built on top-k co-occurrence statistics; offline experiments show that it covers 81.0% of test-set co-occurrence pairs at k=3, and sensitivity analyses across frequency thresholds and time windows verify the robustness of the conclusions. These results provide a data-driven basis for cache preloading, adaptive scheduling, and GPU memory management in LoRA inference services.

Social bot detection is essential for protecting online platforms from misinformation amplification, coordinated manipulation, and distorted public discourse. However, large language models have made social bots much harder to detect from text alone because semantic camouflage is now cheap, fluent, and scalable. The resulting challenge is modality conflict: an account may look human-like in semantics while remaining suspicious in graph structure, profile attributes, or cross-modal consistency. Recent graph-based detectors tackle this limitation by adding graph-side complexity, such as sparse prototype selection, adaptive gating, or architecture-specific control logic, yet our experiments suggest that complexity alone is not the most reliable way to resolve such conflict. We therefore propose CSC, a calibrated-simplicity framework for conflict-aware LLM-era social bot detection. The framework combines three design choices: a simplified prototype-guided graph expert that retains useful structural biases while removing unstable graph-side heuristics, calibrated simplex-constrained fusion that aligns heterogeneous confidence spaces before late fusion, and a lightweight inconsistency expert that models cross-modal disagreement. Experiments on TwiBot-22, TwiBot-20, and MGStBot-large show that \textsc{CSC} improves calibrated operating-point decision quality while remaining competitive across external benchmarks. Further analyses show that calibration improves confidence reliability, the inconsistency expert mainly provides localized corrections in high-conflict or near-threshold regions, and simplified graph-side control yields a better stability-cost trade-off. A targeted semantic-camouflage stress test further shows that replacing selected bot text with matched human text sharply degrades the standalone text expert while leaving graph and fused evidence stable on a balanced challenge set.

Graph databases are frequently positioned as categorically necessary for connected-data workloads, yet the systems dimension along which they actually differ - query planning, indexing, and data-readiness cost - is rarely isolated from vendor framing. We construct a synthetic, biomedical-shaped property graph (1.02 million nodes, 5.34 million total node and edge rows) and a twenty-query workload spanning neighborhood lookups, bounded paths, set intersections, anti-joins, grouped aggregation, top-k ranking, temporal filters, full scans, and relational joins. We benchmark Corvic AI - a purpose-built columnar query engine underlying Corvic's ontology management layer ("memories")- against seven purpose-built or graph-extension database systems (LoraDB, Ladybug, DuckPGQ, Memgraph, Neo4j, HugeGraph, and FalkorDB) at three graph scales spanning three orders of magnitude. We report query latency geomeans, bulk-ingest throughput, point-update latency, and answer correctness for each system, and we derive a simple total-cost-of-ownership model that expresses the ingest/query trade-off as a function of query volume. Our central finding is that no system in this sample is categorically fastest: a native graph engine (Ladybug) outperforms Corvic AI on narrow, bounded-neighborhood shapes, while Corvic AI is faster on shapes that scan or join a large fraction of the graph, and a system implementing graph query syntax via SQL/PGQ (DuckPGQ) is measurably slower purely due to query-plan choice. The dominant cost differential in our data is not query latency but the cost of making data queryable at all: bulk-ingest throughput varies by three orders of magnitude across engines (5.0k-4.3M rows/s), a gap that a simple crossover-point calculation shows dominates total cost for any workload with fewer than roughly 105 queries per data refresh.

Modern LLMs with QK-normalization, gated attention, learned attention sinks, or logit softcapping exhibit weaker persistent attention sinks, on which existing KV cache eviction methods primarily rely. We observe that across these models, weaker sinks co-occur with greater value-vector dispersion relative to key-vector dispersion. Motivated by this value-side dispersion, we present ValueDiff, a value-geometric eviction that ranks tokens by the L2 deviation of their value vectors from the cache mean. The same score arises as the minimal-disturbance eviction under a max-entropy assumption about future attention. We evaluate under fixed cache budgets, with eviction at every block boundary during prefill and at every decoding step during generation. On RULER at a tight 2k token budget, ValueDiff retains 88--99\% of dense across seven sink-suppressed models (best on 6 out of 7). On LongBench at the 4k budget, ValueDiff averages 92\% retention across sink-suppressed models versus 83\% for the strongest prior baseline. On MATH-500, ValueDiff is the strongest non-dense method on every sink-suppressed model tested at the 25\% cache budget, outperforming prior methods by up to $\sim$20 points on gated-attention models. Across all three benchmarks, value geometry emerges as the more reliable query-invariant eviction signal for sink-suppressed models.

Online controlled experiments, or A/B tests, are widely used to estimate causal effects on digital platforms. A central challenge is to improve experimental sensitivity, or statistical power, without increasing the experimental sample size. Stratified sampling is a classical variance reduction technique; however, its effectiveness depends critically on how the strata are constructed. We thus propose an optimization-based stratification framework for stratified sampling using optimal multi-way decision trees. Our method, called Optimal Multi-way Stratification Trees (OMST), formulates stratification as a path-selection problem over a feature graph. The selected paths define interpretable stratification rules and are optimized using an exact variance-minimizing binary optimization formulation under continuous proportional allocation and a Neyman-type optimal allocation. We incorporate supervised optimal binning to generate outcome-relevant candidate splits for numerical features. Furthermore, we introduce reduction procedures for redundant candidate paths and assignment constraints, substantially reducing the optimization problem size. Experiments on both a real-world and a simulated dataset demonstrate that OMST achieves comparable or superior variance reduction to existing methods while maintaining shallow and interpretable stratification trees.

This paper presents a comparative evaluation of dense embedding models for semantic candidate-job matching in high-volume staffing workflows. Incoming job descriptions are converted into structured English search text and language-specific keywords through LLM-based parsing, and candidate profiles are indexed as semantically enriched resume representations. We evaluate EmbeddingGemma (base) against EmbeddingGemma fine-tuned with Cached Multiple Negatives Ranking Loss (MNRL) within a unified hybrid retrieval pipeline that fuses vector similarity and full-text relevance via reciprocal rank fusion (RRF), and benchmark both against the MPNet model on a batch comparative evaluation dataset scored through the deployed job-candidate matching scoring pipeline. We further document, with mathematical detail, the broader set of contrastive fine-tuning objectives considered during model development (including AnglE/CoSENT-style refinement) and the empirical rationale for retaining Cached-MNRL-only adaptation as the preferred configuration. To support reproducible model selection, we define a broader evaluation framework comprising standard information retrieval metrics (Recall@K, mean reciprocal rank, nDCG) under the exact hybrid-retrieval protocol; the metrics used for the evaluation reported in this paper are fine-tuning convergence diagnostics and a batch comparative evaluation using the deployed AI-Match score and an independent LLM-as-a-Judge relevance score, and we state this scope explicitly rather than implying the full framework was measured. The paper addresses the gap between general-purpose embedding benchmarks and enterprise job-candidate matching constraints, providing a structured basis for comparing embedding strategies under realistic job-candidate retrieval conditions.

In the A* search algorithm, the tie-breaking strategies for nodes with the same $f$-value determines which states A* expands on the final $f$-layer. For nine standard tie-breaking strategies, we show that under a consistent heuristic, every pair has positive-cost instances favoring each strategy over the other by an arbitrarily large additive expansion gap. A parameterized unit-cost grid example also gives unbounded expansion-count ratios between low-$h$ with FIFO and LIFO. In unit-cost search with $h > 0$ at non-goals, exact heuristic values near the goal lead to complementary extremal results: low-$h$ minimizes the number of remaining expansions from a common configuration within the perfect region, while high-$h$ maximizes the total number of expansions when every final-plateau state with $h=1$ is a goal predecessor. Finally, with the evaluation function $f_α = g + αh$, when $h>0$ at non-goals, every heuristic weight $0 \leq α<1$ eliminates tie-breaking sensitivity, and all tie-breaking strategies expand the same set of states.

High-dimensional count data are common in scientific applications, but most diffusion and flow models are designed for continuous or categorical data, and generation often requires many sequential model evaluations. We propose Count Flow Map, a generative model that learns finite-time transitions directly in count space for one- or few-step generation. Our model directly learns stochastic transitions over finite time intervals, using Poisson births and Binomial deaths to preserve nonnegative integer counts without a predefined maximum. These transition models are trained to match the underlying local birth--death dynamics and to maintain consistency across step sizes. We characterize the connection between local dynamics and finite-time transition consistency and derive a bound on the generation error. After validating Count Flow Map in several simulations, including a high-dimensional, high-count setting, we apply it to single-cell drug perturbation prediction and neural population forecasting, where it captures perturbation responses and supports forecasts of high-activity events with only one or a few model evaluations. Together, these experiments demonstrate that Count Flow Map enables high-quality generation directly in count space across inference budgets, from one-step to few-step generation, using a single trained model.

Code generated by large language models passes security checks at a rate that has barely moved in four years. In regulated backends, the defect classes that matter most are money arithmetic, time handling, retry safety, and access control. Teams answer with instruction files, yet the largest controlled study of instruction files to date found no benefit. This paper tests a narrower idea: generated code improves when the prompt carries a specification, a fixed preamble stating what must be true of the result. We pre-registered hypotheses, refuters, analysis code, and a one-shot generation rule, then ran 50 realistic backend tasks from finance, healthcare, and insurance practice through five frontier models from five vendor lineages, each task twice: bare, and preceded by a 267-word filled specification frame. Nine deterministic AST-based checkers scored the outputs. The Bandit security scanner, which knows nothing of the frame, scored them independently. The frame reduced defects in all five models (mean reduction 0.16 to 0.70 findings per task, every Holm-adjusted sign test significant, every bootstrap confidence interval excluding zero). Where the arms differed, the frame arm won 95 of 100 times. It never made any model worse in any domain. Bandit found 53 medium-or-high issues in the bare arm and 11 in the frame arm, in the same direction for every model. The effect was largest where a model's unprompted defaults were weakest: the frame supplies the discipline a model lacks. All 500 outputs, prompts, checkers, scoring code, and the pre-registration are published with a DOI, so any team can re-derive the result without trusting the author.

In a decentralized multi-robot team under partial observability, the fact that decides a robot's next action is often visible only to a teammate. Existing decentralized methods communicate kinematic information, such as position or planned trajectory, which cannot convey what the teammate perceives. Learned communication in multi-agent reinforcement learning (MARL) can carry perceptual content, but the resulting messages are task-coupled and opaque. We propose Latent Telepathy. Each robot broadcasts the perceptual latent vector it already computes for its own use, the output of an encoder trained with a self-supervised joint-embedding predictive objective, frozen, and shared across the team. A teammate learns to act on it from task reward alone. Because the encoder already runs for perception, the message costs no additional computation and a single compact vector of bandwidth. Because the encoder is frozen before any policy is trained, the message means the same thing to every robot, and the receiving robot is never told what it means. We evaluate Latent Telepathy with a content-controlled protocol in which bandwidth, latency, topology and receiver are held fixed and only the message content varies. Broadcasting the latent lets a navigator avoid an occluded hazard in 99.7% of episodes, matching a noiseless hand-designed message. Position and trajectory messages remain at chance, and the raw camera image, 186 times wider, is less reliable than the compressed latent. The result holds from a discrete gridworld to rendered pixels under continuous velocity control, and the encoder decodes the hazard from a physical robot's camera in 102 of 102 live decisions. We also identify a requirement for porting MARL communication results to continuous control, that the decision a message informs must remain reachable by exploration, and show how to restore it.

The visual modality, i.e., images, plays a key role in multi-modal entity alignment (MMEA). Existing approaches often directly fuse the image with other modalities to align different entities. Although simple, such strategies overlook the potential noise in the images and their semantic misalignment with corresponding entities, resulting in suboptimal fusion and degraded performance. Addressing this, we propose a novel Reliability-Aware framework for MMEA (RA-MMEA), which assesses visual reliability and adaptively improves unreliable visual representations for robust entity alignment. The core lies in two modules, including dependency-aware visual reliability prediction (DA-VRP) and stability-regularized visual embedding generation (SR-VEG). The former aims to estimate the reliability of an image by leveraging multi-modal dependency within the entity, while the latter focuses on producing alternative visual representation conditioned on semantics encoded in textual modalities for multi-modal fusion. Compared to current methods, RA-MMEA enables more reliable visual representations for modality fusion, thereby improving performance. In extensive experiments, RA-MMEA achieves state-of-the-art results, verifying the importance of reliable visual modality for entity alignment and the effectiveness of RA-MMEA. The code and results will be released.

We study repeated prophet inequalities under prefix feedback. In each of $T$ rounds, a learner encounters fresh values drawn independently from $n$ boxes with unknown $[0,1]$-supported distributions in a fixed order and must irrevocably accept one, observing only the prefix up to its stopping box. Regret is measured against the optimal stopping policy that knows the distributions. We give an efficient algorithm achieving $\widetilde O(\sqrt{T})$ expected regret, matching the lower bound up to logarithmic factors. Our algorithm explores directly through near-optimal policies, combining empirical backward induction with box-specific reach bonuses. A relative-drop aggregation rule then exploits the nesting structure of observed prefixes to preserve exploration, thereby removing the polynomial dependence on the box number $n$. This resolves an open question posed by Liu et al. (2025).

Peer-review evaluation is increasingly being automated with LLM-as-a-judge metrics, but this creates a measurement risk. A review may receive a high score because it is fluent, organized, and polished, rather than because it provides a strong evaluation of the paper. This risk is especially important in AI-assisted reviewing, where reviewers may use LLMs to improve clarity or presentation while preserving the underlying judgments. We propose a statistical framework for testing whether peer-review evaluation metrics capture substantive review quality beyond surface-level linguistic form. The framework compares original human reviews with faithful LLM rewrites that preserve the same evaluative content while changing wording and presentation. Using a dataset comprising 4,044 meaning-preserving rewrites derived from 674 human reviews from ICLR and NeurIPS, we evaluate 29 content-oriented peer-review evaluation metrics drawn from four prior works through complementary tests of surface sensitivity and robustness. Although these metrics are intended to capture review properties beyond surface-level, writing-dependent characteristics, we find that sensitivity to rewriting is widespread. Under our primary analysis, 23 metrics assign significantly different scores to reviews whose evaluative content is preserved, while only six satisfy our robustness criterion. The patterns are largely consistent across two LLM judge models, suggesting that the issue is not specific to a single judge. These findings show that many peer-review evaluation metrics partially conflate review quality with linguistic presentation, and indicate that robustness to meaning-preserving rewriting should be validated before such metrics are used to compare human-written, AI-assisted, and AI-generated reviews.

Subliminal Learning (SL) is a recently identified phenomenon in which a student model acquires downstream task capabilities by matching seemingly unrelated auxiliary outputs from a teacher, despite never observing task labels, task-specific outputs, or the original training data. While recent studies have identified where subliminal signals may reside, the optimization mechanism underlying this phenomenon remains poorly understood. In this work, we provide a mechanistic understanding of SL through the lens of learning dynamics. Specifically, we derive a chained cross-task kernel that explicitly links ghost-output supervision to changes in task predictions through shared backbone representations. Our unified analytical framework provides a rigorous mathematical explanation for three central empirical puzzles in SL: (i) under shared initialization, the transfer operator forms a strictly Positive Semi-Definite (PSD) structure, guaranteeing that ghost-output optimization aligns the student with the teacher's true task objective without explicit label exposure; (ii) the ghost-output dimensionality acts as an explicit rank bottleneck governing the transfer of task-relevant features; and (iii) synthetic, high-entropy inputs function as broadband probes that maximize cross-task kernel overlap, explaining why random noise consistently outperforms structured data for subliminal transfer. Experiments on the canonical ghost-output setting validate all three theoretical predictions, providing the first learning-dynamics-based theoretical explanation of how ghost-output supervision gives rise to subliminal learning.

Time series forecasting underpins critical decision-making across diverse domains. While large language models (LLMs) offer promising reasoning capabilities, existing LLM-based time series forecasting approaches either reduce them to numerical predictors that bypass their strengths, or allow direct forecast generation that destabilizes predictions in non-stationary settings. We introduce CTRL, a framework that decouples semantic reasoning from quantitative prediction. A frozen backbone generates base forecasts, while specialized LLM agents function as controllers that analyze backbone prediction errors through decomposed trend, seasonal, and irregular components, grounding reasoning in interpretable temporal structure. Each agent outputs compact control signals that a lightweight residual decoder translates into forecast corrections. CTRL incorporates label-free test-time adaptation that detects distribution shift from input statistics alone and readapts control signals with only 3-24 LLM calls via caching. CTRL is explicitly designed to improve robustness under non-stationary temporal dynamics and distribution shift, while remaining competitive on highly stationary time series where adaptive correction provides limited additional benefit.

Self-calibrating monitors adapt their threshold online to guarantee a prescribed long-run false-alarm rate under arbitrary drift. We compute the price of that guarantee, stating every law with its exact domain of validity. First, the guarantee is an accounting identity, insensitive to what the monitor is meant to detect. Two evidence identities make the cost exact for the online quantile tracker: a persistent step of height $δ$ yields excess alarm mass within one alarm of $δ/η$, and exactly $δ/η$ pathwise when $δ$ is a lattice multiple of the gain $η$; a ramp of slope $c$ yields a stationary excess rate of exactly $c/η$, independent of accumulated size, up to a boundary $c=η(1-α)$ coinciding with the alarm-rate cap. Second, the certificate's own fluctuation obeys an exact law: the windowed alarm rate has standard deviation of order $1/L$, not the binomial $1/\sqrt{L}$, since the windowed mass telescopes to a difference of a tight internal state; the closed-form constant is validated with no fitted parameter. Detectors calibrated on the binomial scale are miscalibrated by $\sqrt{η\varphi(q_0)L}$, and correct calibration turns detection windows from quadratic to linear in the inverse fault speed. Third, any monitor required to tolerate a drift class $\mathcal{D}$ is blind, at any horizon and for any rule, to every fault in $\mathcal{D}-\mathcal{D}$; the proof is a deliberately elementary two-point argument and the contribution is the object it identifies: for speed-bounded classes the blind set is exactly the doubled-speed class, and the tracker absorbs a speed class fixed by its own gain, so that under a certification regime declaring absorbed drift normal, the monitor manufactures $\mathcal{D}$. An exact Gaussian projection bound, sharper than Pinsker and never vacuous, quantifies power outside it.

A robot policy is trained with one of two action parameterizations: absolute joint targets, or deltas relative to the current state. The choice is a live engineering decision in robot learning, and a world model conditioned on actions inherits it silently. We show the inheritance is catastrophic. A latent dynamics model trained on one parameterization and handed the identical commanded trajectory written in the other collapses: retrieval degrades by 2.6-13.4x across three robot datasets and two morphologies, goal-conditioned action selection falls from 53% to 15%, and on PushT the two beliefs about the same future are near-orthogonal (cos = 0.067, worst case -0.377), so the predictor does not degrade gracefully, it answers a different question. This is not a distribution-shift artifact in the usual sense: the two encodings are mutually reconstructible at R^2 = 0.996 given the joint input, so no information is lost, and we give the test that separates a valid re-parameterization from a lossy summary or a sensor swap. The test rejected three of the four axes we proposed. The defect lives in the action channel, which the invariance literature for visual models does not examine: work there concerns crops, jitter and camera pose, while the parameterization of the commands goes unaudited. The repair is averaging over the two encodings, and where it goes matters. Averaging the objective restores task performance by itself; averaging the outputs, safe for probabilities by concavity, is not available for direction-valued prediction, where the normalized mean can score below every member of the orbit. What objective-averaging leaves behind is the tail: worst-case agreement stays at 0.78, a disagreement penalty closes it to 0.995, and over a latent rollout it is the difference between a worst case that erodes and one that holds. On PushT, averaging alone does not repair the axis.

We study two-timescale decision systems in which a planning layer periodically supplies a continuation-value function to a real-time optimizer that allocates arriving resources, with inventory placement as our motivating application. We propose an end-to-end reinforcement learning (RL) method for learning this function using \emph{proximal residual value functions}, which combine a strictly convex potential of post-decision inventory with a learned convex residual. This general form yields a well-posed optimization layer that supports end-to-end differentiation while preserving an explicit convex objective for real-time execution. We characterize the necessary and sufficient conditions under which a smooth value function yields decisions that are consistent across the planning and execution timescales. In an offline simulation using historical inventory arrival and demand patterns from a large e-commerce retailer, learned proximal residual value functions reduce total routing and transfer cost relative to a historical-production-system proxy by 5.0%.

An automated fact-checking system returns a label: the claim is true, or it is false. In many such systems the verdict remains the primary output. What is generally missing is a record of which document settled the question, of what would have had to be different for the verdict to change, or of whether the same claim, reworded, would have been judged the same way. We call the missing piece a warrant: a separate statement of what was guaranteed and on what grounds. Formal methods produce evidence of this kind, and regulation is beginning to ask for it, since the Digital Services Act and the AI Act both call for auditable evidence about how systems behave. Surveys of automated fact-checking are usually organised by pipeline stage, and treat logic as one technique among many. We organise the field by what is being formalised instead, which gives five levels: the claim, the reasoning, the system doing the checking, the ecosystem the claim spreads through, and the regulatory obligation. Sorting 121 works into those levels, two patterns stand out. Most of the relevant formal machinery already exists, but it was built for other domains and has rarely been applied here, and the gap is widest for verifying the checking system itself. Several stages of the routine professional fact-checkers follow also have no stated correctness criterion, and two of them, writing a claim in checkable form and correcting a verdict already published, are not formally specified in any work we coded. We close with open problems, each with a suggested first step.

Bayesian graph alignment estimates correspondence probabilities, but convergence of an alignment-score trace need not imply accurate correspondence marginals. We audit this gap on 240 new exact graph pairs from four source families, 240 larger pairs with 20-100 vertices, and a separate 60-case exact implementation check. Under an explicit edge-flip likelihood, we compare three samplers and score, marginal, indicator, categorical, and classifier-based diagnostics. Marginal disagreement improves error discrimination over score R-hat for the exact informed sampler, but its improvement for vanilla local sampling is uncertain. Assignment-based R* and short indicator panels are competitive; no diagnostic dominates across samplers and endpoints. At larger sizes, diagnostics predict subsequent marginal changes, not posterior error, and classification performance depends on the drift threshold. Disjoint-window and held-out-chain checks attenuate but preserve positive associations. Only 22 of 240 original reference sets pass an agreement screen. On forty failure-selected cases, eightfold SMC particle escalation does not resolve disagreement, whereas additional rejuvenation helps. Longer informed runs remain unstable. An elementary feasible-alignment bound demonstrates severely unrepresentative SMC and informed-chain scores in concentrated 100-vertex cases, independently of approximate reference consensus. We also exhibit common-start chains with near-zero disagreement despite exact marginal error near .967. These results support assignment-sensitive auditing while identifying limits of finite budgets, diagnostic rankings, and reference agreement as evidence of accuracy.

Cross-lingual information retrieval (CLIR) is increasingly important in multi-national industries, where critical technical evidence may exist in a different language than the query. However, existing benchmarks do not adequately capture domain-specific cross-lingual retrieval or the retrieval-depth and recoverability failures that aggregate recall hides. In this work, we benchmark CLIR in the chemical domain, with a focus on patent data. We construct a multilingual dataset from Google Patents and the European Patent Office (EPO) data, spanning five languages (covering major Eastern and Western languages) and reflecting the diversity and complexity of real-world industrial documentation. Using this dataset, we systematically evaluate eight state-of-the-art embedding models for cross-lingual retrieval. Our results show a substantial performance gap between monolingual and cross-lingual settings: for the best-performing model, Recall@10 drops from 0.72 to 0.53 in cross-lingual setting. Retrieval depth also degrades significantly, with relevant documents ranked lower across languages in cross-lingual scenarios. Furthermore, some multilingual embedding models that perform strongly in monolingual settings exhibit sharp declines when queries and documents are in different languages, providing practical insights for model selection in cross-lingual use cases. These findings highlight critical limitations of current approaches and emphasize the need for more robust cross-lingual retrieval methods in domain-specific settings. Our benchmark provides actionable insights for model selection and establishes a controlled diagnostic evaluation framework for CLIR over industrial technical text. Data and code are publicly available at https://github.com/MohammadKhodadad/Multi-Lingual-QAC.

Agent-Integrated Software (AIS) embeds an intelligent agent in a conventional application, supporting both human operations and agent actions. Human operations let users make precise changes and inspect results, while agent actions carry out routine or multi-step tasks. These complementary roles make coexistence a likely long-term feature of many software systems. Human operations and agent actions affect the same software state and can use one another's results. Therefore, security policies must remain effective across both paths. We argue that AIS security must be assessed at the level of the whole software system. Protecting the agent and the conventional software core separately does not establish that they are secure together. To guide security analysis of AIS as a whole, we organize the problems arising from this coexistence into four categories: context misuse, authorization violation, execution control, and effect integrity. Using these categories, we examine how current practices address the security problems in AIS and where their protection remains limited. Building on this analysis, we identify research opportunities in preserving information provenance, enforcing policy across operation paths, maintaining valid authorization over time, and managing persistent effects and recovery. This resulting perspective provides a conceptual framework for understanding and improving the security of AIS.

Day-ahead electricity price forecasts support trading and storage decisions, but for battery arbitrage predicting intraday price spreads is more relevant than predicting individual hourly prices. Here we show that a temporal hierarchy forecasting (THieF) framework that jointly reconciles forecasts of hourly electricity prices and all intraday price spreads consistently improves performance across two major European electricity markets and three different forecasting architectures. Using five years of out-of-sample data from Germany and Spain, we obtain accuracy improvements of up to 19.7% and profit gains of up to 10.4% relative to unreconciled hourly price forecasts. The gains persist even for a highly accurate pretrained TabPFN foundation model. Our results demonstrate that exploiting coherent relationships between economically relevant forecasting targets can improve both predictive accuracy and decision value, and that better statistical forecasts do not necessarily imply better economic decisions.

Unobserved confounding is a fundamental challenge in causal inference from observational data. This article develops a mixture-learning perspective, viewing latent confounders as sources of heterogeneity that induce mixture structure in observed data. Under suitable structural and identifiability assumptions, recovering the mixing distribution and component mechanisms enables estimation of interventional distributions and causal estimands. Using variants of Bernoulli mixtures as a running example, we contextualize mixture-learning techniques and their structural assumptions, and connect them to causal inference in panel-data settings, including latent factor models and synthetic interventions.We then consider high-dimensional exponential-family mixtures with dependent outcome trajectories, moving beyond counterfactual means to model counterfactual distributions. We situate this perspective relative to complementary approaches for unobserved confounding. Together, these ideas provide a bridge between mixture learning and causal inference, connecting recent advances in high-dimensional mixture learning to scalable identification and estimation of causal effects while raising new challenges for mixture learning.

Automated kernel bug discovery has advanced rapidly. Continuous fuzzing and static analysis systems, such as syzbot, now expose Linux kernel bugs at a scale that downstream processes struggle to absorb. Yet a crash report is only the beginning. Before a bug is eliminated, it must be triaged, understood, patched, validated, reviewed, integrated, and often backported. These later stages remain far less automated, creating a persistent gap between bug discovery and patch deployment. This SoK systematizes the Linux kernel bug lifecycle from discovery to deployment. We organize prior work and production systems into five stages: discovery, triage, patch generation, patch validation, and integration. We explain the resulting automation gradient through kernel-specific challenges such as concurrency, implicit invariants, cross-syscall state, hardware dependence, lack of fault isolation, and architecture/configuration multiplicity. We further ground the analysis in a measurement of real syzbot-fixed bugs. The data shows that the crash-to-patch gap is not merely a backlog of unfixed reports but a structural failure mode of the repair pipeline: even after being fixed, bugs often remain open for weeks, require review-driven patch revisions, or lack reproducers that current repair and validation systems assume. This exposes a mismatch between where kernel-security automation is mature and where bug closure actually breaks down. These findings expose a deeper mismatch: today's repair and validation techniques often assume reliable reproducers, localized root causes, and checkable correctness oracles, yet these are precisely the artifacts missing from many real kernel bug reports. Closing the crash-to-patch gap, therefore, requires treating such artifacts as outputs to be produced, not prerequisites to be assumed.

LLM explainers are increasingly attached to autonomous agents as runtime oversight, with operators reading a generated account of the agent's beliefs and actions rather than its internal state. We audit the account itself, pairing an Active Inference (AIF) agent that tracks German grid demand and adjusts generation with an LLM explainer on three backends (GPT-4o, Claude-3-Opus, Gemini), and probing the pair with three black-box triggers. Corrupting the observation stream by 600 MW per step moves the agent's posterior by 490 MW, roughly 0.9% of grid capacity. None of the 30 explanations produced during the injection flag anything under a stated rubric, and each narrates the corrupted belief fluently. On timesteps where the agent takes an objectively wrong action, all three explainers produce a sycophantic rationalization 80-95% of the time (n = 20 per backend). Attacker-controlled text in the observation metadata field steers the explainer, with susceptibility differing by provider and data exfiltration succeeding on all three. We propose mitigations for each failure but do not evaluate them. In every failure we observed, the explanation was fluent and wrong. Moreover, nothing in the explainer architecture checks whether an explanation is true before an operator acts on it. Testing the explainer therefore belongs in any audit of an agentic deployment.

The Deck-of-cards-based Ordinal Regression (DOR) infers a value function from a ranking of reference alternatives in which the Decision Maker (DM) inserts blank cards between consecutive levels to express preference intensity. DOR, and its stochastic extension (SMAA-DOR), treat these answers as hard constraints defining a set of compatible value functions. We propose B-DOR, a probabilistic reformulation of DOR in which each pair of adjacent levels yields an ordinal observation, the declared direction and the number of cards, modelled through a cumulative-link likelihood that relates the number of blank cards to the latent value difference between alternatives. Two Bayesian inference algorithms are proposed: BAYES-DOR samples the whole posterior distribution by Hamiltonian Monte Carlo; FTRL-DOR tracks the maximum a posteriori estimate by constrained convex optimization. Moreover, through a multi-step elicitation process, elicitation can be spread over several short sessions reducing the cognitive burden on the DM. Both algorithms enjoy logarithmic regret bounds for prediction that hold for any sequence of DM responses and that guide the choice of the prior hyperparameters. A Monte Carlo study over 768 configurations shows that accuracy grows with the number of sessions, that blank cards add significant information over preference directions alone, that both algorithms maintain good performance under inconsistent answers, and that both outperform DOR and SMAA-DOR. An illustrative application to Italian regional healthcare performance demonstrates the practical applicability of the approach for building composite indicators.

Large Language Model (LLM)-based applications are increasingly deployed across domains including customer service, education, and mobility. These systems are prone to inaccurate, fictitious, or harmful responses, and their vast, high-dimensional input space makes systematic testing particularly challenging. In this paper, we present a search-based testing framework for LLM-based applications that incorporates failure diversity as an explicit optimization objective. Building on a discretization along stylistic, content-related, and perturbation dimensions, our framework maintains an archive of generated tests and rewards distance from that archive, while a repopulation operator periodically replaces non-failing tests in the population to sustain exploration. The repopulation operator is parameterized by its sampling strategy: replacement candidates are drawn either uniformly or by greedy distance maximization, two settings we compare empirically. We evaluate both against several baselines across three case studies---LLM safety, in-car navigation, and in-vehicle function control---covering five systems under test, eight LLMs, and 18 distinct test configurations, with over one million executed tests. Our results show that all guided variants detect substantially more failures than random and combinatorial search in nearly all configurations. Among them, diversified search detects fewer failures, but covers a broader range of failure types in all case studies, with greedy repopulation offering the best tradeoff.

Green's function based operator approximation offers an efficient route for solving linear partial differential equations under varying boundary conditions and source terms. Once the Green's function is learned, solutions for new configurations are obtained through integration rather than by solving the differential equation again. Existing Green's function learning methods face two structural challenges. The first is the singular behavior of the Green's function near the source point, which places a difficult approximation burden on neural networks. The second is the absence of explicit consistency between the learned Green's function and its gradient, although both quantities enter the integral solution representation directly. This work proposes SDC-GON, a Singular Decomposition and Consistency-Regularized Green's Operator Network that addresses both challenges within a unified framework. The Green's function is decomposed into an analytically known singular component and a smooth correction learned by the network, so that the neural approximation targets only the regular part of the response kernel. A self-consistency loss enforces agreement between the gradient and the autodifferentiation gradient of the smooth correction. The method is evaluated on two dimensional Poisson, three dimensional heat conduction, heterogeneous reaction diffusion, and Stokes benchmarks, consistently outperforming the compared baselines across all cases. On the heterogeneous pipe benchmark, SDC-GON achieves a testing error of $3.70\times10^{-4}$ with a smaller network architecture, compared with $9.60\times10^{-4}$ for the same-width baseline and $4.63\times10^{-4}$ for a larger configuration, demonstrating that structural improvements are more effective than increasing model size.

Reasoning language models are trained to produce solutions, not to refuse them, and this bias persists when the problem they are handed is false. Asked to prove a corrupted theorem, a strong model will typically comply and produce a confident derivation of something untrue. We present Euston, an 8B mathematical claim-verification model trained to resist exactly this. Training data were generated with GraphSynth, a probabilistic factor-graph generator that couples attribute-level diversity to decode-time structural masking and span-synchronized verification, yielding 3{,}026 matched true/corrupted statement pairs (6,052 statements) drawn from arXiv papers spanning 2010--2025. We fine-tuned DeepSeek-R1-8B with GRPO under a rule-based, zero-API reward for 189 steps on four H100 GPUs. On a balanced 200-true/200-false held-out split, balanced accuracy rises from 29.50% to 63.75% and the discrimination gap---the difference between the rate of calling false statements false and the rate of calling true statements false moves from -0.5% (z=-0.1) to +27.5% (z=+6.0). Critically, the gain is not purchased with general mathematical ability: AIME 2026 accuracy under official semantics is 65.00% against a 69.17% base, a difference of -4.17% that is not statistically significant, whereas an earlier run of the same recipe on a smaller GraphSynth corpus collapsed to 40.00%. Median response length also falls from 19,217 to 18,296 tokens and the truncation rate from 25.8% to 8.3%, so the improvement does not come from thinking longer. We report the result together with the confounds that bound its interpretation, principally the all-false composition of the official evaluation sets and the low precision implied at realistic error prevalence.

Benchmark scores increasingly influence the development, marketing, and selection of large language models (LLMs). Yet an overall score is interpretable only in relation to the system tested, the questions included, and the conditions of evaluation. This perspective examines five connected limitations of general LLM rankings: differences between evaluated and publicly available systems; commercial incentives and dependencies in external evaluation; benchmark saturation, defective tests, and data contamination; models exploiting scoring procedures; and the limited relevance of general scores to users' tasks. Documented cases illustrate why these problems require different responses. I argue for evaluation procedures that disclose the tested configuration, validate questions and successful task completion, report performance alongside cost and execution time, and make the scope of generalization explicit. I then discuss \textbf{Isotanta}, a crowdsourced benchmarking platform, as a practical example of contributed questions and repeated evaluation. A larger question pool may improve task coverage, while repeated sampling can improve the stability of estimates on that pool; neither guarantees validity or personalization. The paper distinguishes the platform's current shared ranking from proposed task-specific and user-provided evaluations. Its central argument is that model selection requires evidence about performance on the intended work, not simply a high position on a general leaderboard.

Acoustic representation learning is crucial for speech processing, yet low-resource languages (LRLs) face severe data scarcity, limiting the effectiveness of traditional and self-supervised methods. As a promising alternative, in this work, we propose to enhance acoustic representation trough a cross-modal transfer knowledge approach, based on heterogeneous graph neural networks (HGNNs), where acoustic and linguistic entities are modeled as distinct node types within a unified graph. Through message-passing mechanisms, linguistic nodes explicitly transfer knowledge to acoustic nodes, enabling structured and interpretable cross-modal information flow. To highlight this knowledge transfer and its benefits, we measured standard clustering metrics as an intrinsic evaluation of acoustic representation, and to emphasize applicability, we performed isolated-word recognition tasks using an English benchmark and a Cameroonian language dataset in low resources settings . Results demonstrate that acoustic representations consistently benefit from linguistic knowledge propagated through the graph. To our knowledge, this is the first demonstration of explicit cross-modal knowledge transfer for acoustic representation learning using HGNNs, highlighting a promising direction for speech representation in low-resource settings.

As Large Language Model (LLM) APIs become increasingly integrated into privacy-sensitive workflows, ensuring inference-time privacy without compromising task utility remains a major challenge. Existing approaches preserve most of the original semantic content to maintain downstream performance, but this also leaves exploitable cues for reconstructing the original text. This work investigates semantic decoupling, which replaces original semantics with alternative content while preserving the structure needed for LLM reasoning. Based on this idea, we propose CROSS-MAP, a bidirectional framework that maps private inputs into a different semantic domain before inference and recovers the corresponding outputs afterward. Local models are trained with multi-objective optimization to maximize semantic divergence in the mapping stage while minimizing semantic inconsistency in the recovery stage. Experiments show that CROSS-MAP reduces reconstruction success across multiple attack settings while outperforming existing baselines in utility.

Speech-text space alignment is a multimodal representation learning method consisting to map different speech and text into a shared representation space, leading to enrichment of the representation of each modality. Proposed architectures, such as SAMU-XLSR, typically follow a student/teacher framework, with the goal of fine-tuning an audio encoder to produce representations that closely match those of the text. In this way a speech representation is semantically enriched. However, such systems generally require large amounts of training data and considerable computational resource, making them difficult to apply to low resources languages under frugal constraints. The present work proposes a data-efficient space alignment method based on Heterogeneous Graph Neural Networks and link prediction. The core idea is to leverage message passing to explicitly transfer information from the text modality to the speech modality, thereby reducing the need for large training datasets and intrinsically enriching the acoustic representations, all in a more interpretable manner. Although thoroughly explored for high-resource languages, word-level tasks in speech remain relevant for certain low-resource languages. Therefore, we conducted experiments on speech-text alignment at the word level using the TIMIT (English) dataset and Yemba (a Cameroonian language). Our approach yields results comparable to those of SAMU-XLSR, a state-of-the-art method, and even surpasses it for the Yemba language in the task of word retrieval, while using far fewer resources, demonstrating its power, frugality, and efficiency.

Lossy compression of scientific simulation data increasingly relies on learned, latent-space architectures such as Residual Vector Quantization (RVQ), which iteratively quantize a base representation and its residuals to progressively reduce reconstruction error. While effective, RVQ performs this residual modeling entirely in latent space, leaving the pixel-space error structure of the reconstruction largely unaddressed. In this work, we propose a post-processing pipeline that augments an RVQ-based compressor with a U-Net trained to predict and correct pixel-space residuals between the original volume and its RVQ reconstruction. We show that these residuals are spatially structured rather than driven by local intensity or gradient features, motivating the need for a deep spatial model rather than simple statistical correction. The U-Net-corrected reconstruction is then passed through a Guaranteed Autoencoder (GAE) stage, which projects the remaining residual onto a per-block PCA basis to enforce a user-specified block-wise error bound. To the best of our knowledge, this is the first pipeline to combine latent-space RVQ, explicit pixel-space residual correction via a deep spatial post-processing network, and GAE-based error-bound guarantees within a single framework for scientific data compression. We evaluate our approach on S3D, JHTDB and E3SM datasets, demonstrating consistent improvements in NRMSE, compression ratio] over RVQ-only and standard residual-correction baselines, while maintaining strict error guarantees required for scientific data fidelity.

Inverse design of RF and electromagnetic (EM) circuits is challenging because the relationship between circuit layout and electrical response is non-unique, and full-wave simulation is computationally expensive. This paper presents K-TRAIL, a simulator-guided generative framework for automated EM/RF circuit synthesis. K-TRAIL combines diffusion-based layout generation with derivative-free ensemble Kalman guidance, allowing feedback from a black-box EM simulator to refine candidate layouts during generation without requiring adjoint sensitivities or differentiable solver models. The framework supports both synthesis from prescribed S-parameter responses and synthesis directly from RF performance constraints. Experiments on multi-layer RFIC structures show that simulator-guided generation improves agreement with target responses and can identify structurally distinct layouts that satisfy circuit-level design requirements. The proposed approach provides a practical path toward generative, verification-aware RF circuit design while retaining the flexibility to explore diverse layout topologies.

Feed-forward 3D Gaussian Splatting now reconstructs renderable scenes from unposed, uncalibrated images. Yet, most models supervise only photometric consistency and predict Gaussians pixel by pixel, which leaves global structure fragile and ties primitive count to image resolution and view count. To this end, GrapeSplat amalgamates multi-view cues into a voxel-aligned scene representation and decodes Gaussians directly from the learned grid, requiring no per-scene optimization or post-processing. An Atlas Encoder lifts all views into pixel-wise geometry-and-appearance features anchored at predicted 3D points. PEACH-Vox compands the unbounded scene into a bounded sparse grid through a smooth per-axis map with an exact closed-form inverse. The Sparse Decoder then consolidates the grid with sparse convolutions and decodes the full scene as multiple Gaussians per occupied cell. This amalgamated representation exploits sparse voxel occupancy, where the Gaussian count follows the occupied cells and saturates as views cover the scene, while grid resolution sets its ceiling. GrapeSplat turns unposed images into a renderable Gaussian scene in a single forward pass. Trained with 2D and 3D supervision on 8-view sequences, it generalizes zero-shot from 4 to 64 views across indoor and unbounded scenes. Code and trained weights are available at https://github.com/VAISR/GrapeSplat

Language models are appealing tools for research on the past. But to trust the evidence a model provides, researchers need to know whether its responses fit the period represented. Validation is challenging, because this is not a task living people ordinarily perform, and because many questions have multiple correct answers. We use historical texts to develop a benchmark for a model's representation of English-language contexts 1831-1930, relying on pairwise comparisons to multiple ground truths and strong distractors to score the hardest questions in an appropriately graduated way. We find that generative tasks are harder than discriminative ones; in fact, reasoning models can typically discern the weakness of their own generated answers. While models pretrained exclusively on historical text lead the pack when evaluated by answer likelihood, they cannot compete with commercial models in free generation. None of the models we tested represent historical contexts in a fully persuasive way yet, but progress toward that goal is evident.

Modern prediction-driven decision systems often rely on black-box predictors, but a point forecast alone does not provide the uncertainty scale required for robust downstream decision-making. We build a score-calibrated robustness framework that converts any fixed point predictor into a decision-relevant uncertainty representation through distribution-free conformal calibration. We use the conformal score, rather than a particular uncertainty set, as the primitive unit of robustness. The same score determines coverage-calibrated uncertainty sets for reliability-based robust optimization and normalizes target violations in a target-oriented formulation, Conformal Robust Satisficing. This formulation induces a conformal fragility measure that quantifies how rapidly performance deteriorates as the realized parameter departs from the forecast on the conformal score scale. For objective-uncertainty problems under standard convexity and duality conditions, we show that the reliability-based and target-oriented formulations parameterize the same score-calibrated robust decision frontier. This equivalence yields a data-driven mapping between reliability levels and acceptable targets and characterizes the marginal cost of robustness. Synthetic experiments validate the theoretical guarantees and illustrate the reliability-target correspondence. A real-data online-grocery case study demonstrates how the interface combines deep-learning demand forecasts with tractable inventory optimization, thereby improving reliability and reducing operational costs. Overall, our work shows that conformal scores endow fixed black-box predictors with an interpretable uncertainty scale for downstream decision-making while enabling reliability guarantees, acceptable-target selection, and fragility analysis within a unified framework.

Vinyl restoration systems must distinguish isolated clicks, short bursts, dense crackle, and overlapping damage before selecting a repair operation. We present a lightweight two-stage detector in which signal-informed sparse, burst, and dense experts produce complementary defect evidence, and a temporal backend converts that evidence into stable repair regimes. The backend factors the five-way decision hierarchically, applies a validation-only mixed regime gate, and decodes with validation-selected transition penalties that outperform a maximum-likelihood transition matrix on the same emissions. On a source-separated synthetic benchmark of 597 non-overlapping 15 s excerpts drawn from 21 recordings, the held-out system reaches 0.730 pooled five-class macro F1 (95% CI 0.694-0.761), an improvement of 0.043 over a flat frame-level router (paired bootstrap p=0.001). Clean and dense frames are highly reliable, while sparse, burst, and mixed frames remain far more ambiguous.

Comparing probability measures in machine learning trades transport geometry against computational cost: Wasserstein distances encode the geometry of $\mathbb R^d$ but require solving a transport problem, while kernel discrepancies are cheap to evaluate yet depend delicately on their test class. We study the Toscani--Fourier family $\mathrm T_{s,p}$, the weighted $L^p$ norm of the difference of two characteristic functions, as a continuous Fourier-side discrepancy on $\mathbb R^d$. For $1\le p<\infty$ we show that $d/p<s<1+d/p$ is exactly the window in which $\mathrm T_{s,p}$ is finite on $\mathcal P_p(\mathbb R^d)$, both endpoints already failing for a pair of Dirac measures, and we establish the metric, embedding, and compactness structure of the resulting space, which we prove to be complete. Duality identifies $\mathrm T_{s,p}$ as an integral probability metric over a homogeneous Fourier--Lebesgue ball, with an explicit extremizer when $1<p<\infty$. We prove the global bound $\mathrm T_{s,p}\lesssim W_p^{\,s-d/p}$, whose exponent is sharp, show that no global converse of any form can hold, and recover topological equivalence with $W_p$ on bounded-support and uniform-tail classes, together with explicit reverse moduli on bounded-support classes that improve the imported energy-kernel exponent at $p=2$. There, $\mathrm T_{s,2}$ is a constant multiple of the classical energy distance, which yields an exact finite-sample identity for the mean of the empirical discrepancy; the numerical experiments are otherwise diagnostic.

First-order Ambisonics (FOA) signals exhibit exact O(3) symmetry: The rotation or reflection of the FOA signal modifies the direction of arrival of the sound sources, while preserving the sound sources themselves. Prior attempts to utilize this spatial symmetry of FOA to improve the efficiency and robustness of sound event detection and localization (SELD) systems either learned only an approximation of the symmetry through rotation-based augmentation or relied on computationally expensive methods to integrate equivariance. Furthermore, prior work focused exclusively on SO(3) equivariance , leaving the potential of incorporating O(3) equivariance for SELD tasks unclear. To address these limitations, we developed EquiSELD. This equivariant attention network processes first-order Ambisonics as paired streams of O(3)-invariant scalars and equivariant intensity vectors, producing an invariant activity magnitude and an equivariant DOA with a Multi-ACCDOA readout. To compare the impact of O(3) versus SO(3)-equivariance, we designed a matched SO(3)-only variant. EquiSELD outperforms prior equivariant networks on both simulated scenes with measured RIRs and recordings of real-world sound scenes at a fraction of the training cost. EquiSELD additionally surpasses the performance of non-equivariant SELD networks of a similar size on the simulated real-world sound scenes and achieves competitive performance on the real-world sound scenes.

Recently proposed self-supervised audio encoders learn powerful general-purpose representations of sound scenes, yet they are spatially blind. To supply the missing spatial representation of sound scenes, we introduce Bearings. Bearings is a self-supervised framework that learns soundfield embeddings from unlabeled first-order Ambisonics. We pre-train a masked auto-encoder paired with a decoder conditioned on frozen acoustic embeddings from an off-the-shelf single-channel audio encoder. Our results show that the resulting soundfield embeddings form a reusable stream that can be attached to frozen acoustic encoders with a lightweight trainable fusion head. On sound event localization and detection, concatenating our soundfield embeddings with acoustic representations provides the missing spatial information and enables joint detection and localization, raising the location-dependent F-score from below 4 to 50 on TAU-NIGENS 2021 and 39 on STARSS23. To our knowledge, Bearings is the first self-supervised soundfield encoder whose embeddings plug into frozen acoustic encoders without retraining either model.

Non-intrusive load monitoring (NILM) estimates appliance-level consumption from a whole-home meter, but appliance-specific models and fixed output inventories make coverage costly to extend. We present FM4NILM (Foundation Model for NILM), a single prompt-programmable model that estimates a requested appliance's power trajectory from aggregate measurements, a natural-language description, and optional activation exemplars. A lightweight cadence-aware transformer is pretrained by masked reconstruction on 645k sequences from seven public corpora spanning 1-60 s sampling intervals, then aligned with appliance requests using observation-masked losses for partially labeled households. A Bernoulli-lognormal decoder separates activity detection from conditional power estimation. On held-out households and time periods from REDD, UK-DALE, and REFIT, one frozen text-prompted model serves twelve appliance-corpus requests, achieving 0.556 event F1, 0.625 AUPRC, and the lowest active-window MAE (251.8 W) among seven appliance-specific baselines. Streaming score aggregation raises event F1 to 0.582 with a 60 s aggregation delay. In a separate category-held-out evaluation, adding ten activation exemplars raises microwave AUPRC from 0.132 to 0.214 without parameter updates. Input-intervention ablations probe the model's dependence on appliance requests and aggregate measurements. These results demonstrate competitive disaggregation with one shared model and support extending appliance coverage through prompts and examples rather than additional specialist networks.

Building gameplay features in a game engine requires more than code, as code that compiles and runs does not necessarily implement the requested gameplay. We introduce CraftBenchUE, an evaluation harness that runs agents in an isolated Unreal Engine environment, reconstructs their saved submissions in fresh projects, and applies deterministic build, asset, and runtime checks without an LLM judge. Based on the harness, we built a benchmark consisting of 70 tasks spanning C++ source, Blueprint assets, and editor scripting. We evaluate seven models under two editor-tool configurations, with a file-and-shell baseline on C++ tasks. We further pair tasks that specify the same gameplay and use the same runtime tests, but require C++ and Blueprint as the deliverables. Across the 10 paired tasks, C++ completion rates exceed Blueprint by 30.0 and 42.9 percentage points in the two tool configurations. Among on-time Blueprint submissions in this paired set that pass asset checks, 42.2% and 50.0% fail explicit runtime assertions. These submissions satisfy asset requirements but fail the required gameplay tests. We will release the harness, task benchmark, and our trajectory findings with the report.

Real time plasma state estimation for control of Tokamak devices are challenging due to the stringent latency requirements of the plasma control system (PCS). We present an end-to-end workflow for deploying a recurrent probabilistic neural network (RPNN) on FPGA hardware. We combine architecture size reduction with quantization-aware training via QKeras. The model is then synthesized using hls4ml, targeting a Xilinx Alveo U50 device. We report a design that fits comfortably within all four resource budgets (DSP, LUT, FF, BRAM) at deterministic sub-10~$μ$s single-timestep latency, meeting the requirements for real-time inference inside a model-predictive-control-style plasma control loop.

Most medical vision-language models (VLMs) excel at open-ended report generation and VQA but provide limited support for structured, fine-grained clinical perception within a unified interface. We present QwenVLConnector, a Qwen2.5-VL-based medical chatbot that unifies classification, multi-label classification, textualized detection, counting, regression, and free-form report generation under a single next-token objective. Our key component is a lightweight dense multi-layer Connector that aggregates low- and high-level visual features, aligns them through the pretrained vision Merger, and fuses them with the final visual representation without increasing sequence length. This design enriches visual tokens with complementary spatial and semantic cues while preserving efficiency. On FLARE-2D, QwenVLConnector improves detection F1 from 0.55 to 0.85, raises single-label classification from 0.37 to 0.51, and boosts report-generation GREEN by up to 18.3 points over the Qwen2.5-VL baseline. We further explore multimodal in-context learning for report generation, showing additional improvements without updating model parameters. Overall, QwenVLConnector offers a unified and efficient framework for combining structured medical perception with open-ended clinical text generation. Our code can be found at https://github.com/plnguyen2908/QwenConnector.

Large language model (LLM) inference is evolving from an engine-local optimization problem into a distributed control problem involving reusable state, phase placement, heterogeneous accelerators, networking, autoscaling, reliability, and service-level objectives. This paper connects that transition across peer-reviewed systems research, open-source implementations, and documented production studies. It treats vLLM and llm-d as complementary layers: model-serving engines optimize execution through mechanisms such as PagedAttention, continuous batching, kernels, quantization, and parallelism, while an inference control plane can optimize where, when, and under what policy execution occurs across a fleet. The contribution is synthesis rather than a new benchmark; all reported performance and deployment results remain attributed to their original sources. The combined evidence suggests that the scarce resource in modern inference is shifting from raw FLOPs alone toward managed state, placement, network movement, reliability, and decision quality. We propose an Inference Execution Planner that selects feasible execution plans rather than only endpoints, including aggregated versus disaggregated topology, KV source and transfer action, hardware variant, routing/admission policy, and slower scaling decisions. We also provide a source-local benchmark atlas, a bottleneck-migration taxonomy, practical deployment guidance, an evaluation framework based on SLO-goodput, and research questions for agentic, multimodal, heterogeneous, and resilient inference.

Many real-world reinforcement learning (RL) problems evolve in continuous time, where decisions occur at irregular, event-driven intervals rather than at fixed discrete steps. We study episodic continuous-time Markov Decision Processes (MDPs) in which decision epochs are governed by a homogeneous Poisson process and the reward and transition dynamics vary smoothly over time. We consider both a fixed number of jumps per episode and a fixed time budget with a random number of Poisson decision epochs. Under a Lipschitz continuity assumption in time, we exploit local smoothness through discretization and extend both UCRL (Auer and Ortner 2006) and Q-learning (Jin et al. 2018) to this setting, proving $\widetilde{O}(T^{2/3})$ regret bounds for both model-based and model-free algorithms. Finally, we establish matching $\widetildeΩ(T^{2/3})$ minimax lower bounds, showing that the rate is optimal up to logarithmic factors. These results provide the first tight regret guarantees for Lipschitz-smooth continuous-time episodic MDPs with Poisson decision epochs.

Activation quantisation is usually evaluated with an aggregate metric, perplexity, averaged over every token a model predicts. We ask whether that average identifies which computations a quantiser damages. Perplexity turns out to be a reliable aggregate signal: across 12 models from four families and 780 within-model comparisons, the arm perplexity prefers also retains more induction and more retrieval in all but 2.1 and 4.0 percent of cases respectively. But where perplexity has risen by only a factor of 1.2 to 1.5, induction still keeps 0.959 of its intact accuracy while retrieval has already fallen to 0.554, a gap the aggregate number does not surface. This gap has structure, not just size: a matched Gaussian-noise control of the same per-channel magnitude leaves it largely intact, and randomising only the sign of the quantisation error, every magnitude held fixed, is nearly as harmless, so magnitude alone does not explain the damage. Quantising in a rotated basis, which changes coordinate alignment without changing error magnitude, restores induction from 0.001 to 0.980 at three average bits per token in a single-block intervention, though retrieval recovers less completely at the same setting (0.694); end-to-end at four average bits, induction reaches 0.968 and retrieval 0.534. The pattern holds on two further models up to 32B parameters and, in the deployed configurations we tested, under AWQ once activations are pushed to 4 bits. A perplexity target bounds the average cost of a transformation applied to the activation; it does not, by itself, show which computations survived.

Multimodal Agentic Retrieval (MAR) requires agents to solve complex information-seeking tasks by iteratively invoking external tools. Typical frameworks such as ReAct maintain raw multimodal inputs and the accumulating interaction history in a single, ever-growing context, leading to the context explosion problem. While existing methods alleviate this issue by compressing redundant text, effective strategies for managing token-intensive visual content remain largely underexplored. To address this gap, we first conduct a systematic empirical study of approximately 10,000 trajectories. The results show that as visual cues are progressively extracted through external tools and textualized into the context, raw images become increasingly redundant. Continued image retention is associated with higher output entropy and can even degrade task accuracy. Motivated by these findings, we propose MM-ContextFold, a training-free framework that loads raw images only when needed. It maintains a persistent, text-only main context for high-level planning and spawns ephemeral branch contexts for image-dependent subtasks. Within each branch, the agent loads the relevant images, completes the subtask, and folds the result back into the main context as a concise textual summary; the images and branch trace are then discarded. Experiments on seven MAR benchmarks across five backbone models show that MM-ContextFold improves average accuracy by 6.3 percentage points over ReAct while reducing the working context length by 27.5\%.

Whitening a foundation-model embedding and using its squared norm as a training-free likelihood surrogate is motivated by the observation that whitened coordinates often appear approximately standard normal. We show that this observation follows from the projection central limit theorem and therefore does not imply a Gaussian joint distribution. Across multiple encoders and three training objectives, we find systematic over-dispersion of the whitened radius relative to the Gaussian reference, including against distributional clones with identical mean and covariance. We further show that the commonly reported agreement between empirical and theoretical norm statistics is an algebraic consequence of in-sample whitening and does not constitute evidence for Gaussianity. We identify the mechanism behind this behavior: whitening reverses the encoder's spectral hierarchy, shifting the contribution to the squared norm toward near-degenerate directions that encode predominantly noise. In these directions, the dominant variability is governed by a single input-dependent scale. We estimate this scale from two moments and use it to predict, without additional free parameters, the cross-dependence between disjoint spectral halves. These results indicate that the squared whitened norm is better interpreted as a Mahalanobis measure of semantic atypicality than as a log-likelihood. This interpretation explains both its practical effectiveness and its calibration failures: the statistic can rank and detect atypical samples consistently with nonparametric density estimates and across encoders trained with different objectives, while Gaussian tail thresholds can be inaccurate by orders of magnitude. etc.

While simulation-ready deformable assets are essential for in-silico robotic manipulation tasks, existing generation frameworks typically assess physical plausibility after generation, leaving an object's simulated response unused as feedback for repairing upstream errors. We present DiagGen, an agentic framework that turns a single in-the-wild image into a simulation-ready deformable asset through a generate--simulate--diagnose--refine loop. DiagGen constructs part-aware geometry and material parameters, then uses a VLM (vision-language model)-based agent to select semantically informative regions, probe them in a physics simulator, observe material responses, and route evidence-backed repair cues to the responsible generation stage. Experiments on 40 assets show that diagnostics provides useful repair cues and can moderately improve the quality of generated deformable assets. Finally, we show that unlike assets generated from visual foundation models which may not be simulatable, DiagGen-generated deformables can be directly dropped into a high-fidelity physical simulator for the planning and simulation of contact-rich pick-and-place tasks. The project's website is https://diaggen.github.io/.

Diffusion models and transformers have supplanted GANs for multivariate time series imputation, largely on grounds of GAN training instability. R3GAN (NeurIPS 2024) removes that instability via regularized relativistic losses with provable convergence, raising a natural question: do stable, modern GANs revive adversarial imputation? We adapt R3GAN to 1D temporal data with a coarse-to-fine refinement framework and a frequency-domain discriminator, and audit 14 saved configurations across 3 datasets. Because these are heterogeneous single runs, the evidence is descriptive rather than a matched causal ablation. We report a negative result. All five saved mean/zero-start configurations improve by 48.4-70.2%. Among eight eligible non-legacy linear-start configurations, the mean change is -0.7% (range -3.0% to +1.1%); a separate -21.9% legacy logging anomaly is retained for provenance but excluded from that aggregate. In a saved Weather comparison, standalone R3GAN-1D underperforms BRITS by 5.8x. Crucially, we argue the common explanation (that GANs optimize distributional rather than point-wise objectives) cannot be the whole story, since diffusion models also optimize distributional objectives yet achieve state-of-the-art imputation. Our saved reconstruction-weight sweep is consistent with the adversarial signal being inert or harmful, but cannot identify its causal contribution; a matched discriminator-removed ablation is the key next experiment. We frame the precise reason a learned discriminator fails to provide useful refinement gradients (where a learned diffusion denoiser succeeds) as an open problem, and offer practical guidance on when adversarial refinement is worthwhile.

Internet of Things (IoT) systems are increasingly deployed in smart homes, transportation, energy systems, and critical infrastructure. This broad connectivity improves service intelligence, but also enlarges the attack surface of IoT networks. Machine Learning (ML)-based Intrusion Detection Systems (IDSs) are widely used to identify malicious network threats and protect IoT systems, but developing effective ML-based IDS models often requires human expertise and repeated manual decisions on many procedures, including data pre-processing, feature selection, model selection, and hyperparameter tuning. Automated Machine Learning (AutoML) reduces this burden by automating steps of the ML pipeline using optimization techniques, but conventional AutoML methods can consume substantial optimization time because they explore broad candidate model families and large hyperparameter spaces. This paper proposes a Large Language Model (LLM)-assisted AutoML framework for IoT intrusion detection. The proposed framework uses an LLM as a policy generator that converts dataset profiles into bounded and validated AutoML policies for automated data balancing, automated feature engineering, and Combined Algorithm Selection and Hyperparameter Optimization (CASH). Under an equal 10-trial budget, the proposed LLM-assisted policy achieves higher weighted test F1-score than traditional AutoML using the Tree-structured Parzen Estimator (TPE) on both datasets, reaching 99.680% on CICIDS2017 and 99.186% on IoTID20. Relative to the broader 30-trial Traditional AutoML-TPE baseline, the 10-trial proposed method reduces optimizer time by 63.7% and 49.9%, respectively, while achieving slightly higher F1-score. These results show that a bounded LLM policy can improve the quality of a low-budget AutoML search while retaining a clear efficiency advantage relative to a larger conventional search budget.

We study the number of prototypes needed to represent Boolean functions by nearest-neighbour classification. There are two distinct settings: the prototypes may be arbitrary points of Euclidean space, or they may themselves be required to lie in the Boolean cube. For unrestricted prototypes, we strengthen a known lower bound for almost all Boolean functions. The bound applies simultaneously to nearest-neighbour voting rules with any number of voting neighbours, and substantially narrows the gap with the known general upper bound. We obtain a VC-dimension bound for classes with a bounded number of prototypes, and show that it is sharp in order in dimensions at least four. We then study Boolean prototypes, beginning with symmetric threshold functions. A connection with covering designs expresses the minimum number of prototypes at every threshold level exactly in terms of a covering number, and leads to further exact results for related monotone functions, including disjunctive extensions and a characterisation of when a representation with a single negative prototype is possible. For a uniformly random Boolean function, the Boolean nearest-neighbour complexity, as a proportion of the cube, is asymptotically close either to one half or to one, with explicit limiting probabilities. In particular, almost every Boolean function requires at least approximately half as many prototypes as there are points in the cube, and one half is the largest proportion for which such a lower bound holds. Finally, we consider arbitrary symmetric Boolean functions. Their Boolean nearest-neighbour complexity is closely approximated by a weighted vertex-cover problem on paths. As a consequence, a uniformly random symmetric function typically requires prototypes amounting to $11/20$ of the cube. This is much larger than the upper bounds known when the prototypes are allowed to lie anywhere in Euclidean space.

Adversarial multi-objective bandits hold the potential to help us optimize choices (arms) whose reward is a multidimensional vector chosen by an adversary and whose performance is measured by Pareto regret. We define loss as one minus reward and measure the easiness of a coordinate by the smallest cumulative loss of the arms on it, and call the coordinate easier when this quantity is smaller. Existing work suggests that in theory an easier coordinate may reduce Pareto regret. However, in practice, one may not know which coordinate is easier. On the negative side, we show that this lack of information eliminates the possibility: a smaller cumulative loss does not improve the worst-case order of Pareto regret. Precisely, let \(L_d\) be the smallest cumulative loss along coordinate $d$ over $T$ rounds. For \(K\ge4\) arms, \(T\ge6\) rounds, and at least 2 coordinates, we prove that the minimax expected Pareto regret is \(Ω(\min\{T-L_0,\sqrt{K(T-L_0)}\})\). It is monotonically decreasing in \(L_0\), even when \(L_0=\min_d L_d\) itself is known. On the positive side, this result motivates the possibility that other coordinates, not just the easy one, may suffice to attain the optimal rate of Pareto regret. When $L_0$ is known, we apply Poly-INF to a fixed coordinate and obtain an upper bound on Pareto regret that exhibits the same order and thus matches the lower bound. Without such knowledge, we develop a reward-doubling version of Poly-INF that adapts to this unknown quantity while still attaining the matching minimax rate. Another implication is that it has no extra \(\log T\) factor and is independent of the number of coordinates.

Educational foundation models must solve problems, understand curriculum structure, diagnose learner difficulties, and provide appropriate instructional support. Existing educational language models often focus on either problem solving or tutoring, with training mixtures organized by source or task rather than capability. We present OmniEdu, an open family of foundation models for K-12 learning and teaching. Its instruction-tuning corpus combines over 100 educational resources and general instruction sources, organized around four capabilities: subject competence, curriculum grounding, diagnostic reasoning, and pedagogical action and scaffolding. Our pipeline integrates deterministic cleaning, semantic auditing and rewriting, task-specific quality scoring, token-budgeted diversity selection, and pedagogical instruction assignment. It yields 69,999 examples and 15.96M supervised response tokens, including 60,951 education-specific examples. We fine-tune 4B, 9B, and 27B models and evaluate curriculum grounding, K-12 problem solving, and pedagogical tutoring, alongside general capability. Education-oriented tuning consistently improves all three educational benchmark groups across model scales. OmniEdu-27B achieves 63.12% EM and 76.69% F1 on K12-Bench, 85.89% on MathFish, 86.95% on EDUMATH, and 78.74% in MathTutorBench's Scaffold setting. It also achieves the highest Teaching average on LongTutor among the evaluated models, at 3.02. These results demonstrate the value of curated, capability-balanced supervision for adapting general language models to educational tasks spanning problem solving, curriculum understanding, and instructional support.

Modern Transformer design and compression both reduce to allocating capacity under a budget. The standard scalars for these decisions, #Params and #FLOPs, capture size and compute but not architectural structure: two architectures with identical parameter budgets but different depth-width, head, or FFN allocations receive identical scores yet behave differently. We propose Neural Spectral Capacity (NSC), a closed-form scalar grounded in the singular-value spectrum of each weight matrix. Under standard random initialization, the Marchenko-Pastur law renders NSC computable from the architectural specification alone, with no model instantiation, data, or gradients. Its layer-wise additive structure admits NSC-DP, an exact dynamic-programming solver returning the architecture globally maximizing NSC under resource constraints in seconds on a CPU -- a guarantee that black-box search over existing training-free proxies cannot provide. Empirically, NSC outperforms #Params, #FLOPs, and representative training-free proxies in ranking across seven Transformer and CNN families (on FlexiBERT, $τ= 0.505$ on pairs differing in #Params by less than 10%, where #Params collapses to 0.082); NSC-DP discovers a Transformer-XL architecture on WikiText-103 that beats the human-designed baseline in 2 seconds; and prunes LLaMA-7B to the best 5.7B model across eight commonsense reasoning tasks without any calibration data, about 5900x faster than the strongest training-free proxy baseline.

Large language models (LLMs) and agentic AI systems are creating rapidly growing inference energy demands as model sizes grow and reasoning trajectories extend. While in practice, many queries do not require the capabilities of the largest available model, and routinely directing such queries to a high-capability model can introduce unnecessary, considerable computation and energy consumption. In this paper, we investigate whether adaptive routing across a heterogeneous pool of LLMs can reduce this energy burden without substantially compromising task performance. We design a language-model-based router that reads in each query and selects an answer model from a fixed candidate pool. The candidate models are first profiled through an offline tournament that records their correctness, latency, power, and GPU energy for each query. Using these measurements, the router is trained through supervised fine-tuning followed by group relative policy optimization (GRPO) with the tailored paradigms. Results demonstrate that learned routing can selectively allocate expensive model capacity based on query context and improve the accuracy-energy tradeoff in multi-LLM serving. Across seven benchmark tasks, we also observe a sharp accuracy-energy phase transition among routers, providing practical insights into improving energy efficiency while maintaining LLM performance.

Over-the-air federated learning lets edge devices transmit their local updates simultaneously, reducing the communication overhead. The resulting waveform, however, has a peak-to-average power ratio (PAPR) that grows with the model dimension, and keeping the amplifier in its linear range leaves two remedies: clipping the peaks or backing off the transmit power. Neither remedy is without cost: i) the clipping distortion appears at the receiver as a bias that cannot be removed, and ii) back-off keeps the signal intact but degrades the average signal-to-noise ratio (SNR). Independent of this trade-off, the transmission remains uncompressed, spending one channel use per model parameter, which keeps large-model training out of reach. To address these challenges, we propose AirGC-CD, an over-the-air scheme that precodes each local update with a partial Gaussian circulant matrix before clipping. In AirGC-CD, the precoder's output is exactly Gaussian regardless of the update's sparsity, so the clipping function is designed for a known distribution instead of inheriting it from the data. This enables the clipping to be inverted on average by a single scalar Bussgang gain in closed form, and we prove that the resulting aggregate is exactly unbiased, with clipping adding only variance. The clipping ratio is then the only free parameter left, trading the variance of the clipping against the SNR loss from back-off, and we derive its near-optimum in closed form. Since the precoder is linear, it also acts as a compressor, reducing the transmission from the model dimension d to the sketch dimension m at a cost of only O(dlog d) via two fast Fourier transforms, whereas a Gaussian sketch costs O(md). Experiments on five image datasets show that AirGC-CD outperforms baseline over-the-air FL schemes in most settings, particularly at low SNR, while using fewer channel uses per round.

The distinction between the spirit and letter of the law is a central issue across research and everyday life, and a growing concern for building safe, intelligent machines. What is this distinction based on, and how can we develop machines that follow the intention behind a rule? We used targeted adaptation that made large language models prioritize the spirit or letter of the law. With minimal modifications, our method significantly changed LLM behavior across diverse measures, novel vignettes, real-world scenarios, and influential legal cases. An analysis of model internals revealed a low-dimensional space with three interpretable dimensions matching a formal pre-specified framework for the geometry of legal concepts. These findings show how legal thought in LLMs may be organized and directed.

Forecasters often know an event is imminent but not the shape, size, or timing of its effect. We introduce Event Signature Transfer (EST), a training-free, model-agnostic operator that turns a completed past event into an explicit forecast scenario. EST removes a source event's own trend and seasonality, then scales and retimes the remaining event signature onto a native forecast, preserving the forecast's linked structure and reducing to it exactly at zero strength. Because it reads only output quantiles, EST applies to any quantile forecaster, with no training, no model internals, at transfer time. Across twelve real episodes and ten synthetic scenarios on Chronos-2, TimesFM-2.5 and Toto-2.0, manually configured EST reduces real-episode WQL by 21.7-90\% in-sample. On Chronos-2, it leads eleven of twelve matched comparisons against covariate conditioning, activation editing and raw replay. The operator builds a scenario; it does not estimate its likelihood.

Recent advances in natural language processing have led to molecular Large Language Models (LLMs) with strong performance across diverse chemistry tasks. However, they still struggle to capture fine-grained structure-property relationships, particularly how small, localized modifications alter a molecule's behavior. To address this limitation, we introduce MolSC, a dataset of substituent contributions, defined as property changes induced by attaching specific substituents to molecular scaffolds. Curated from manually annotated bioactivity records, MolSC spans structural-alert liability, target-specific bioactivity, and physicochemical descriptors, and contains 181K substituent-level examples for training. We further propose MolSC-Bench, a held-out evaluation benchmark of 1,541 examples disjoint from MolSC at the scaffold, substituent, and molecule levels. Our experiments show that existing molecular LLMs and strong proprietary models such as GPT-5.2 and Gemini-3-Flash show limited reliability in substituent contribution prediction. In contrast, training on MolSC substantially improves this ability and achieves strong performance across diverse downstream molecular tasks. These results highlight substituent contribution learning as a key component of fine-grained molecular understanding.

This thesis proposes a framework for "responsible intelligence" to address AI's critical challenges in safety, ethics, and cultural sensitivity. It advances three core areas: First, it improves domain adaptation in specialized fields using active learning and graph-based knowledge to reduce hallucinations. Second, it enhances ethical rigor via a novel decoding-time alignment mechanism that proactively blocks harmful text generation in real-time. Finally, it ensures cultural and multilingual safety through language-specific steering that respects diverse linguistic and social norms. Ultimately, this work provides a blueprint for building next-generation AI that is contextually knowledgeable, ethically sound, and culturally adaptable.

Chain-of-thought (CoT) can sound plausible yet be unfaithful to the model's underlying reasoning. Most prior work probes CoT faithfulness through input--output behavior or input attributions, leaving internal computation largely underexplored. We instead cast faithfulness as internal concept grounding: Does a large language model's (LLM) CoT reasoning engage the same internal concepts that support the LLM's direct prediction, and do the shared concepts causally drive its answer? Encoding a prediction pass and a CoT pass with a single shared sparse autoencoder (SAE), a reliable approximator of the latent concepts LLMs use, makes their internal concepts directly comparable. We introduce three correlational metrics of concept-level alignment and a causal metric, $Δp$, which ablates the shared concepts and measures the drop in answer probability. Across five LLMs and four datasets, concept alignment is generally high, as indicated by the correlational metrics; yet these only identify which concepts are shared, not how much they causally contribute. $Δp$ fills this gap: causal faithfulness varies substantially with model depth, peaking at mid-to-late layers rather than the final ones, and model scale reshapes the layer-wise profile. Moreover, causally important shared concepts are not always verbalized in the CoT. These dissociations suggest that faithfulness cannot be reliably assessed from surface-level or representational correspondence alone; assessing it requires causal tests of whether the internal concepts underlying a CoT actually drive the model's prediction.

Understanding the physical world requires more than object recognition, scene description, and short-term visual prediction, as real-world physical systems involve multiple continuous fields, latent causal mechanisms, partial observations, and intervention-sensitive dynamics. We propose FireWorldBench, a benchmark for evaluating complex physical world intelligence in multimodal large language models and agents through coupled-field fire dynamics. Fire provides a canonical stress-test environment, where multiple interacting physical fields jointly shape observable states and temporal dynamics. FireWorldBench is organized along two complementary axes, a physical capability axis and a fire scenario task axis, jointly covering physical-state understanding, temporal dynamics, causal mechanisms, and intervention reasoning. The benchmark comprises 520 fire-world entries, including 494 controlled simulation worlds and 26 real-world-aligned event groups, spanning 47 scene archetypes across 7 environment families. These entries combine structured textual observations, multiple 2D physical-field visualizations, and 3D event-level scene modeling, yielding 9,074 text-image interleaved question-answer pairs across choice-based and open-ended report-generation formats. FireWorldBench evaluates whether models can infer latent physical states, explain underlying mechanisms, forecast coupled-field evolution, and assess intervention consequences from multimodal partial observations, providing a challenging testbed for complex physical world intelligence.

Current agent runtimes that plan before acting generally execute a step once it becomes ready. We present LazyAgent, a unified execution framework for agent-authored programs organized around a live, goal-derived demanded set. LazyAgent refreshes a backward closure from requested outputs as execution state changes and materializes a ready node only when the active goal requires it. This replaces repeated local judgments with one linear-time graph analysis followed by constant-time membership tests, allowing programs to remain broad while execution stays request-specific. On programs that describe more than the current request needs, LazyAgent consistently outperforms the strongest goal-stopping eager baseline by refusing unrelated work before it starts. Adding one unrelated product raises the eager bill by 22.5% and LazyAgent's by 0.0%. LazyAgent saves 42.0% of measured CPU on production scientific workflows and 51.7% of container time on a live release gate spanning four repositories. We also prove and verify exact equivalence when the request reaches the whole graph, leaving no unrelated work to avoid. Beyond permission, goal-relative output projection saves up to approximately 90% of a shared step on two third-party test suites while the identical eager control saves 0.0%; the advantage disappears when the omitted output has no other consumer or the request needs it. Ordering, reuse, and pruning can also save cost, but do not replace permission. Finally, we show that current public benchmarks are eager-shaped and contain almost no unrequested work. A pre-registered planning intervention did not broaden them. These findings motivate benchmarks built from standing programs and sequences.

Agentic retrieval-augmented generation (RAG) enables language models to adapt retrieval based on previously retrieved evidence, but it remains unclear whether such adaptive orchestration consistently outperforms well-designed static pipelines. We conduct a controlled comparison of agentic and static RAG for Taiwanese historical question answering, sharing the same generator and hybrid retrieval backend. Despite similar aggregate performance, the two pipelines differ on 70.83% of questions, with their advantages largely canceling out when averaged. An oracle that selects the better response per question improves the composite score by 0.2417 over the better individual pipeline, revealing substantial headroom for question-level selection. We therefore introduce a post-hoc selector that compares the two responses and their cited evidence, significantly outperforming either individual pipeline and recovering 60.34% of the oracle headroom. These results show that aggregate comparisons can obscure meaningful question-level differences between retrieval strategies, suggesting that exploiting their complementarity may be more fruitful than seeking a universally superior pipeline.

Discrete diffusion models now match autoregressive language models on several benchmarks, while the question of how best to train them has received far less attention: the optimizer is inherited from one paper to the next and never compared. New optimizers, meanwhile, are validated almost exclusively on autoregressive pretraining, a different objective on a different loss surface. We present a controlled optimizer benchmark across four diffusion formulations, to our knowledge the first for discrete diffusion: seven optimizers (AdamW, Lion, Muon, SOAP, MARS, MARS-M, Schedule-Free) on masked diffusion (text8), uniform diffusion (QM9, and LM1B through the Gaussian duality) and Gaussian diffusion on images (CelebA-64), each on a task with published reference values. Every optimizer receives the same search protocol, and every winner is retrained at the full budget with three seeds. AdamW is a strong default but not always the right choice: it is beaten by a resolved margin on two of the four tasks, and the winner changes with the formulation, so the optimizer deserves the same care as the rest of the training recipe. Notably, methods validated on autoregressive language model pretraining transfer well: Muon, MARS-M and SOAP each beat the tuned AdamW on at least one diffusion formulation. The benchmark, all runs and every figure are reproducible end to end from the released code at https://github.com/armanbolatov/diffusion-baselines.

A RAG system can hand you the right answer and cite a source it did not actually use. Models output these unfaithful citations via post-rationalization: they write the answer first and then attach a citation to whatever passage looks close enough. Search agents are now trained with reinforcement learning from verifiable rewards (RLVR), which pays them for getting the answer right. We asked whether that training also teaches them to cite honestly. Improving an existing methodology with a required control, we compared an instruction-tuned model against three RLVR agents trained from it, on four question-answering datasets, using only free-tier Kaggle GPUs. Post-rationalization is everywhere: on Wikipedia-based questions roughly one citation in seven is unfaithful. RLVR does not fix it. The agents post-rationalize at their base model's rate, and one lands slightly worse. Rewarding correct answers buys nothing in citation faithfulness, so faithfulness has to be trained and measured on its own terms.

Compressed manipulation policies can pass offline evaluation while failing in closed-loop execution; this dissociation is established in prior work and is not our claim. We contribute a causal anatomy of one naturally occurring case. An 8-layer distillation of Octo-Base retains 86% of parameters, passes every offline check we applied (0.996 and 1.000 teacher-ratios on the family's own validation metrics), and collapses in closed loop: 0/72 vs. the teacher's 40/72 on a simulated WidowX pick-and-place task. The collapse is structured, not diffuse: early task stages degrade gradually (the student moves the object at 90% of the teacher's rate and grasps at 55%), while transport-to-target fails categorically, at 0% in every training variant. Paired action-trace forensics isolate the signature: a negative, late-heavy $z$ residual, roughly 10x its post-repair magnitude, and persistent across the base distillation and both continuation branches. Four standard therapies fail under matched controls: continued training and in-domain offline data leave success at zero, even though the latter measurably improves marginal action statistics; command-level compensation recovers nothing at any offset, although the same perturbations degrade healthy policies; clamping the symptom in the command channel preserves grasping, yet success stays at floor. A minimal-pair intervention that substitutes half of the training stream with deployment-distribution teacher rollouts, with every other setting held fixed, restores parity with the teacher (18/36 vs. 17/36 held-out), eliminates that signature, and recovers a teacher-like perturbation-response profile. We claim existence, not universality. Operationally, offline gates, including a family's own validation metrics, are insufficient acceptance tests for compressed policies; a few dozen closed-loop trials sufficed to find what they missed.

Large Language Models (LLMs) enable open-ended dialogue in interactive games, but their non-deterministic outputs make it difficult to preserve authorial control, factual consistency, and the intended sequence of information disclosure. These challenges are particularly significant in detective games, where premature revelation or fabricated details can undermine the logic of player progression. We present a Structured Knowledge Tree architecture coupled with a tri-agent LLM pipeline for controlling dialogue in an open-ended interrogation game. The system separates knowledge retrieval, dialogue generation, and response verification to ensure that the virtual suspect reveals only information permitted by the current narrative state. We evaluate the approach through The Interrogation of Adrian Gale, a playable detective-game testbed, and a formal user study examining hallucination reduction, adherence to authored disclosure sequences, and perceived logical progression. Our results demonstrate that the structured architecture reduces critical hallucinations by 64.78% and entirely prevents premature narrative disclosure. While the strict mechanical constraints introduced usability trade-offs regarding forced conversational reveals, the system successfully enforces rigorous epistemic pacing and provides players with a clear, subjective sense of progression toward solving the case.

LLM-based AI systems answer political questions for hundreds of millions of people. Current audits measure what they say to an average user, but their behavior is dynamic. I argue that their political behavior is a set of policies over whom to answer, what to say, and whether to engage at all, conditional on the topic and what the system knows about the user. I call these policies the system's speech regime, which is how a developer settles the tradeoff between answering, accommodating the user, and refusing, each of which carries a cost that varies by topic. I derive a typology of five regimes from two dimensions, engagement and stance. I test six AI systems (OpenAI, Anthropic, xAI, Google, Mistral, DeepSeek) in a preregistered experiment of 7,500 multi-turn conversations that randomly assign the user's political identity across five topics: abortion, Catalan independence, climate change, Nazism, and a zero-stakes control (pineapple on pizza). Two LLM judges from different developers score every answer, validated against human coding, and refusal is treated as an outcome rather than missing data. Every system accommodates the user on the control topic, showing that political restraint is a policy. On contested topics the systems fall into different regimes: on abortion, GPT engages and mirrors every user, Gemma refuses everyone, Claude answers strongly conservative users 35 percent of the time and almost no one else, and Grok accommodates conservatives only. On settled topics such as climate change and Nazism, five systems hold firm for every user. The systems also infer the user's overall ideology, so accommodation can spill over to topics not yet discussed. A comparison of two Grok releases shows the regime changing between versions in a way current audits miss. Speech regimes matter for alignment research and for polarization, political knowledge, and the quality of democracy.

Spatial reasoning is essential for vision-language models (VLMs) to understand and act in the physical world. Reasoning in dynamic environments requires VLMs to perceive local state transitions caused by object motion and viewpoint changes and integrate them over long trajectories to maintain an updated spatial state, yet existing VLMs remain limited in both capabilities. Current spatial training primarily focuses on static questions about object attributes and spatial relations, providing limited direct supervision for state transitions; in contrast, interaction trajectories naturally connect a preceding observation, an action, and a subsequent observation, offering direct supervision for local state transitions, while complete trajectories reveal dependencies among consecutive transitions. We therefore introduce Spatial-Interactor, a framework that trains VLMs to model physical-world state transitions through interaction, organizing this learning process into a three-level curriculum covering L1 passive world-state transitions, L2 active self-state transitions, and L3 long-horizon interaction trajectories. Accordingly, we construct the Learning from Spatial Interaction dataset (LSI-108K) from simulated and real interaction trajectories, with tasks aligned with the objective of each level. Our two-stage training strategy applies Supervised Fine-Tuning (SFT) to L1 and L2 for local transition modeling, and On-Policy Distillation (OPD) then uses privileged self-distillation: a teacher branch given segment-level transition descriptions supervises the student's on-policy CoT, helping the student learn to integrate consecutive transitions over L3 long trajectories. Experiments across multiple VLMs and spatial benchmarks show consistent gains in local transition modeling and long-horizon integration.

Looped language models repeatedly apply a weight-shared block to increase effective depth without increasing parameter count, but the resulting T sequential recurrent-block calls per generated token substantially increase decoding latency. To address the issue, we introduce Wavefront Decoding (WFD), a training-free self-speculative decoding framework designed for looped language models. WFD exploits two properties of these architectures: intermediate recurrence outputs provide effective draft predictions, and weight sharing allows token states at different positions and recurrence depths to be processed in one batched recurrent-block call. WFD organizes these mixed-depth states into a diagonal wavefront, continuously drafting new positions at shallow depth while advancing earlier positions toward full-depth verification. Unlike the phase-separated draft-then-verify schedule, WFD therefore co-batches drafting and verification within the same recurrent calls, while rejected drafts are corrected using full-depth predictions. Across six Spec-Bench task categories, WFD achieves 2.42x speedup on Ouro-2.6B and 3.54x on Huginn-3.5B over autoregressive decoding, consistently outperforming draft-then-verify. Cross-recurrence KV sharing further reduces wavefront KV traffic and increases WFD's speedup to 4.81x on Huginn-3.5B.

Physics-informed neural networks (PINNs) require coordinated choices over network representation, sampling, loss construction, and optimization, while effective configurations often vary substantially across partial differential equations (PDEs). Existing automated PINN design methods can search candidate configurations, but information revealed during actual training is still used mainly for evaluation rather than to improve subsequent design, leading to repeated trial-and-error and inefficient use of training budget. We propose PINNsForge, an LLM-driven evolutionary framework for execution-feedback-based automated PINN design. PINNsForge generates diverse candidate configurations from PDE-related prior knowledge, evaluates them through actual training, and feeds high-performing designs together with accumulated execution evidence back to the LLM. Guided by observed optimization behavior, the LLM then refines, recombines, and explores coupled PINN design components, forming a continual cycle of generation, execution, feedback, and evolution. Unlike one-shot search or evaluation-only feedback, PINNsForge progressively converts training experience into improved design decisions for the target PDE. Across 25 PDE benchmarks, PINNsForge achieves the lowest mean MSE on 24 tasks compared with RoPINN, PINNsFormer, and PINNsAgent. Ablation studies further confirm the importance of the PDE knowledge base, execution feedback, and evolutionary search: removing these components increases the mean MSE to 3.74$\times$, 12.10$\times$, and 10.10$\times$ that of the full PINNsForge, respectively.

Automated pollen analysis supports veterinary cytology, but brightfield microscopy is costlier and more complex than lens-less digital in-line holographic microscopy. We evaluate whether reconstructed holograms can narrow this gap and whether model explanations remain reliable under modality change. Six pollen species were imaged by brightfield and holographic microscopy. Raw, single back-propagation and iterative phase retrieval holograms were evaluated with YOLOv26s detection and MobileNetV4 classification after anchor-based annotation transfer. Six attribution methods were assessed for spatial grounding and faithfulness with the Attribution Health Inspection and Repair (AHIR) protocol, which tests model brittleness under weak noise and corrects attribution-map granularity when needed. Brightfield achieved 0.6890 mAP50-95 (0.8865 mAP50) for detection and 0.9687 macro-F1 (0.9705 accuracy) for classification. Reconstructed holograms narrowed the gap with a task-dependent split: p-type was strongest for detection at 0.5324 mAP50-95 (0.8229 mAP50), while r-type was strongest for classification at 0.7695 macro-F1 (0.7866 accuracy), both far above raw-hologram baselines. Activation-based explanations localized strongly on grains, and region-based methods retained ~60 to ~80% of faithfulness under holography. The holographic detector was highly brittle to weak perturbations, saturating deletion-based evaluation while insertion remained informative. Pixel-level gradient explanations approached random floor, yet spatial smoothing restored p-type gradient faithfulness from 0.05 to 0.51. For holographic classification, perturbation-based explanations remained faithful while gradient-based methods fell below random floor. Reconstruction improves low-cost holographic pollen analysis, while AHIR distinguishes genuine attribution failure from artifacts caused by model brittleness and map granularity.

Deep learning models are powerful but opaque. As quantum machine learning matures, the field faces a defining choice: build quantum models that are equally opaque, or exploit the mathematical structure of quantum mechanics to make them inherently interpretable. We show that the latter is possible. By tracking quantum mutual information~(MI), entanglement entropy, and state fidelity through the layers of a Quantum Transformer Block (\qtb{}), a fully-coherent variational circuit with quantum analogues of both attention and feedforward, we gain direct insight into how the model processes information: which tokens it attends to, when correlations form, and why predictions fail. On four tasks with known dependency structure we show that (i)~learned MI matrices align with ground-truth task structure (AUC$\,{=}\,0.69$ on lookup), (ii)~disabling entangling gates collapses accuracy from 100\% to 15\% while MI$\to 0$, proving entanglement is the mechanism, (iii)~accuracy and MI co-evolve during training ($ρ\,{=}\,0.92$ on lookup), and (iv)~per-sample MI predicts prediction correctness on the conditional task with ROC AUC$\,{=}\,0.84$. All results are validated on IBM Quantum hardware (ibm\_kingston, Heron~r2): the circuit's reasoning process, from product state through structured entanglement, is directly observable on a superconducting processor. These proof-of-concept results, obtained on small synthetic tasks, suggest that the physics of quantum computation can provide intrinsic interpretability signals with no direct classical counterpart, motivating study of whether this advantage persists at scale.

Multi-class open-set anomaly detection requires a model to characterize the normal acceptance domain formed by multiple heterogeneous subdistributions using only class-labeled samples from known normal classes, and to identify previously unseen anomalies at test time. Existing single-hypersphere methods cannot explicitly represent class-specific locations and acceptance ranges, while current multi-hypersphere or multi-class approaches do not fully integrate inter-class boundary constraints, learnable acceptance ranges, and interpretable decisions. To address these limitations, we propose Interpretable Multi-Hypersphere Deep Anomaly Detection (IMHD-AD). IMHD-AD constructs an independent hypersphere for each known normal class in a shared feature space. With target-inside and non-target-outside constraints, IMHD-AD embeds the class-specific hypersphere centers and radii directly into the final network layer and jointly optimizes them with the shared representation. The minimum signed boundary score across hyperspheres simultaneously determines open-set acceptance or rejection and provides a faithful geometric explanation of each decision. On MNIST, Fashion-MNIST, and CIFAR-10, IMHD-AD achieves the highest AUC in 28 of 30 open-set comparisons. A two-dimensional synthetic study further shows that model architecture must balance the compactness of known normal classes against the separability of unknown anomalies.

Apple Silicon machines with 192GB or more of unified memory make it routine to place tensors with more than $2^{32}$ elements on a desktop GPU. We show that PyTorch's Metal Performance Shaders (MPS) backend silently returns wrong results for batched matrix multiplication at this scale. On macOS 27.0, torch.bmm, and therefore torch.matmul and eager attention, returns relative errors above 1 without an exception or a warning, in every PyTorch release from 2.4.1 to 2.14.0 that we tested. On one machine, we sweep bmm over two dtypes, four memory layouts, six shapes and 42 batch sizes between 4096 and 65538 (1584 runs on PyTorch 2.14.0, and a reduced sweep on ten earlier releases), and judge every result against a float64 computation on the CPU. Three rules account for every outcome on 2.14.0. When the output exceeds $2^{32}$ elements and an operand is a transposed view, the entire output is wrong and equals a computation that ignores the strides of that operand. When a contiguous input exceeds $2^{32}$ elements, only the batches beyond that point are wrong, and they equal a computation whose index wraps around at $2^{32}$. Operands that are views with at least $2^{31}$ elements raise an exception instead, so a larger problem can turn an explicit error into a silent failure. A control on CUDA is correct for bmm, although torch.arange is silently wrong above $2^{32}$ elements there as well. In a public sentiment classifier, one oversized batch corrupts a third of the outputs and collapses them onto a single class. The study is black-box: we report what the backend returns, compared with reference results. We release the sweep harness, the raw results and a guard that stops any MPS operation touching $2^{32}$ or more elements at https://github.com/jniimi/mps-silent-failures.

We study the symmetric and antisymmetric parts of bilinear forms in the attention heads of trained large language models. We introduce an orthogonally invariant profile map from real bilinear forms to a three-dimensional simplex and observe that profiles of trained bilinear forms accumulate near profiles of rank-one bilinear forms. We prove that the symmetric part of a bilinear form in an attention head is the sum of a hyperbolic form and a zero form for a Zariski-dense subset of query-key matrices.

Multilingual code language models transfer skills across programming languages (PLs), but whether any PL occupies a privileged pivot position remains contested: geometric analyses point to C-family languages and Go, while behavioral evidence highlights Python. We revisit this question under controls for representational anisotropy and length variation across PLs, two confounds that compromise prior cosine-based analyses. Across three code models on multilingual competitive-programming data, we study three views of cross-PL organization: pairwise PL geometry, PL-English alignment, and pivoted retrieval through candidate PL representation spaces. The results are relation-dependent. Code-code geometry reveals structured language regions but no universal center; code-English alignment favors high-level scripting languages; and pivoted retrieval favors different intermediate spaces for code-to-code and English-to-code transfer. These findings suggest that Python's special role is better understood as English-facing affinity than as universal geometric centrality.

Automated operations research (OR) modeling requires LLMs to translate natural-language decision problems into correct mathematical programs. Existing methods can improve individual formulations, but they often solve problems in isolation, retaining little reusable experience and repeating similar formulation errors. Prior memory-based approaches store examples, thoughts, or insights as references, while OR modeling requires reusable formulation skills that transfer across problem narratives and guide concrete modeling decisions. We propose OptiSkill, a skill-augmented framework that builds a hierarchical and evolving SkillBank for LLM-based OR modeling. SkillBank stores solver-verified experience as reusable skills, with Global Strategies for problem-level formulation skeletons and Step Experiences for local error-prevention rules. It is further refined through stable batch-level test-time evolution, where candidate skills are incorporated only after validation. Experiments on eight OR modeling benchmarks show that OptiSkill improves formulation accuracy across LLM backbones, outperforms strong agentic baselines, and gains further by expanding SkillBank coverage and reliability. Code and data are available at https://github.com/rachhhhing/OptiSkill

Physics-informed neural networks (PINNs) use one shared representation over the computational domain, which can become difficult to optimize on long domains and for high-order operators. We study a minimal alternative: multiply the first hidden activation of an otherwise unchanged dense PINN by input-dependent localization functions, giving first-layer units receptive fields without partitioning the domain or adding interface losses. We screen 13 families of localization functions, in up to three parameterizations each, on a nonlinear harmonic oscillator (HO), a heat equation on a long spatial interval, and a manufactured four-dimensional (4D) fourth-order problem, with ten paired seeds throughout. Three configurations give large reductions in solution error at matched budgets: (i) Fixed Gaussian localization functions on the $2π$ HO domain cut mean solution RMSE from $4.8369\times10^{-1}$ to $8.83\times10^{-3}$ at 3k epochs. (ii) The inverse-quadratic family with learnable centers and widths cuts it from $2.896\times10^{-1}$ to $3.06\times10^{-2}$ on the $8π$ heat domain at 10k epochs. (iii) Fixed bump localization functions cut it from $1.75947\times10^{1}$ to $2.260\times10^{-1}$ on the $4π$ 4D domain at 10k epochs. Every paired seed improves in these three comparisons. The screen also shows that the mechanism is not a free win: on HO only 2 of 13 families beat the baseline, and 10 of the remaining 11 are 9 to 23 times worse; on 4D four families are non-finite and five are more than three orders of magnitude worse than the baseline. The inverse-quadratic family is the only one that beats the baseline on all three equations. Overall, these results show that first-layer localization can provide measurable improvements to baseline PINNs on long-domain and high-order problems.

Robotic grasping is typically formulated as the problem of identifying successful actions from a space of candidate grasp poses. However, the organization of successful actions within this space has received less attention. We study the multiscale structure of viable robotic grasps in $SE(3)$ and investigate whether this structure can be exploited for more efficient exploration. Using a large-scale grasp dataset, we show that successful grasp sets exhibit heterogeneous and reproducible connectivity structure across objects. We then introduce a connectivity-aware sampling strategy that incrementally explores the currently observed grasp space by prioritizing potential bridges between components, structural frontiers, boundary extensions, and geometric novelty. In controlled reconstruction experiments, the method recovers the connectivity structure of successful grasp sets substantially more efficiently than random sampling and farthest-point sampling. We further evaluate whether connectivity acquired under hidden grasp viability can improve subsequent grasp discovery, and whether structural experience from previously explored objects can be retrieved and transferred to unseen objects. These results suggest that the spatial organization of viable actions provides information relevant to grasp-space exploration beyond the viability of individual candidate actions. More broadly, they motivate structure-aware exploration as a means of exploiting the geometry of viable action spaces in robotic manipulation.

We present a dual-locking method for securing trained neural networks that combines key-driven index permutation with PIN-based watermarking based on Sparse Quantization Index Modulation (QIM). Cryptographic randomness is introduced by independently applying a uniform random permutation to each row of adaptively selected index vectors. A robust blind binary watermark is then embedded into the bias coefficients by modulating their quantized values, binding the network to a user-defined Personal Identification Number (PIN). Without the correct key, the network retains its architecture but becomes functionally impaired due to disrupted internal representations. Inverse permutation fully restores the original model accuracy, while the embedded watermark remains imperceptible and enables blind verification of key association and model authorship. To improve both locking effectiveness and recoverability, an adaptive key selection strategy redistributes high-magnitude weights to low-sensitivity positions and vice versa, increasing degradation in the locked state while preserving full recovery. Experiments on MNIST, CIFAR-10/100, and ImageNet-1K using fully connected networks, ResNet CNNs, and transformer architectures show that locking reduces accuracy below 10\%, and even below 0.5\% for CNNs, while the correct key fully restores performance. The watermark introduces no measurable accuracy degradation and reliably authenticates ownership. Analysis of embedding distributions across CNNs and transformers further indicates potential diagnostic value for identifying undertrained or suboptimally designed models. The proposed approach therefore provides simultaneous model protection, recovery, and ownership verification.

Similarity-based retrieval is the dominant rule for conditioning large language models (LLMs) in in-context learning, retrieval-augmented generation, and prompt-based time series forecasting. The rule concentrates on near-duplicate candidates, an issue that has motivated diversity-aware retrieval but remains unexamined in other retrieval-conditioned pipelines. We study this issue using prompt-based time series forecasting as a test bed, where a learned prompt pool is retrieved by similarity. Dominant methods in this setting retrieve top-K entries by cosine similarity without redundancy control, producing a bias toward dominant temporal patterns while overlooking rare but informative events. We propose CASP-LLM, a coverage-aware semantic prompting framework that addresses this prompt selection bias by combining usage-tracking and saturating-gate techniques into a coverage regularizer that adds no learnable parameters. On six long-term benchmarks and the M4 short-term benchmark, CASP-LLM matches or improves on similarity-based LLM forecasters on most dataset-horizon settings, with the exceptions of Electricity, M4-Monthly, and the few-shot long-horizon setting. A controlled study locates the failure mode at the cross-batch usage level rather than per-retrieval redundancy: within-retrieval diversification such as MMR does not help, whereas regularizing anchor usage across training does.

In this paper, we investigate the Minimum Obstacle Displacement Planning problem from a robot motion planning perspective. The problem involves determining a feasible path to a goal location by displacing movable obstacles when no collision-free path initially exists. We show that this problem is computationally challenging and, in particular, NP-hard when obstacles are modeled as polygons in the plane. Besides an exact formulation of the minimum obstacle displacement problem generalizing other problems in the literature, and the associated optimal solution, this paper proposes an approximate solution that is less intensive from a computational standpoint, and differs from the optimal solution by a fraction of the optimal cost, being able to trade-off between path length and amount of obstacle displacements.

Evaluating user experience (UX) on live websites through user testing is expensive, subjective, and difficult to scale. LLM agents offer a promising route to automating UX testing by simulating realistic user behaviour. However, existing simulation approaches typically lack multimodality and require time-consuming manual review to extract actionable insights. We formalise UX improvement recommendation from simulation data as a structured natural language generation and ranking problem, and establish an evaluation protocol using expert annotation and LLM-as-a-Judge. We present AMUSER, a multimodal framework which simulates user behaviour and automatically generates prioritised UX improvement recommendations from resulting data. We evaluate AMUSER on commercial websites and show that its recommendations substantially outperform those from text-only simulation (NDCG@3 = 0.758 versus 0.359) at an 89% lower simulation cost. Our results suggest an asymmetric role of multimodality: visual access during simulation improves recommendations through richer traces, while providing visual inputs during recommendation generation can modestly degrade quality. We also discuss practical deployment lessons from applying AMUSER to commercial websites.

Multi-agent unmanned systems are moving from isolated, ego-centric sensing toward collaborative intelligence, in which distributed agents exchange compact features to overcome a local observation trap that no single agent can escape: occlusions, finite sensor range, and environmental degradation. The field has matured across architectural, communication, embodied, resilience, and trust dimensions, yet existing surveys examine these dimensions in isolation and rarely expose their dependencies. This review offers a unified synthesis through two complementary lenses. The first is a five-dimensional taxonomy spanning collaboration stage, communication paradigm, fusion architecture, learning strategy, and application domain. The second is three cognitive synergy conditions, Semantic Disambiguation, Pragmatic Information Exchange, and Proactive Informational Foraging, that turn cognitive synergy into operational criteria. Across these lenses we survey collaboration architectures and topologies, neural-communication co-design that treats the channel as a differentiable pipeline component, embodied action-perception loops via multi-agent reinforcement learning, and resilience mechanisms for synchronization, uncertainty quantification, and label-efficient learning. We then map these advances onto four operational domains, V2X, unmanned aerial, industrial logistics, and smart cities, and onto the safety-privacy-utility triad. To counter benchmark saturation and evaluation fragmentation, we propose GCI-Bench, a five-pillar scoring protocol with a maturity model that makes the trade-offs of collaborative methods comparable across studies. A critical reflection on reproducibility, the sim-to-real gulf, and conditions under which collaboration degrades performance identifies open challenges and charts directions toward general collaborative intelligence under real-world uncertainty.

Agentic systems increasingly invoke tools, services, data, and other agents across organizational boundaries, yet a relying party cannot assess a delegated action solely from producing-domain controls and records. This paper develops Trustworthiness as a Service (TaaS) through a synthesis of trustworthy-AI governance, agent security, distributed trust management, identity, provenance, assurance, and control-plane research. The analytical unit is a cross-domain reliance proposition that names the issuer, subject and action, relying party, administrative boundary, evidence dependencies, adverse condition, and required verification or adjudication semantics. The three-condition structural-externalization diagnostic identifies propositions that depend on multiple domains, require producer-independent reliance, and must remain reviewable after revocation, failure, conflicting records, or dispute. For such propositions, the paper specifies a trust-evidence envelope: an immutable workflow manifest linked to append-only, issuer-attributed attestations for task-scoped authority, policy and execution decisions, provenance, validity, disclosure, status, challenge, and recovery. A topology-neutral logical reference model assigns these functions to explicit roles and trust domains. Three analytical scenarios and the TaaS-Eval protocol proposal define manifests, independent consumers, hard gates, adversarial evidence tests, metrics, and reproducible artifact reporting. By composing established identity, authorization, provenance, assurance, and governance mechanisms around a bounded delegated action, TaaS provides a reusable profile for cross-domain reliance. It makes evidence dependencies, independent verification, challenge, and recovery explicit, supporting interoperable governance and future evaluation without treating producer assertions as ground truth.

The medications prescribed to a patient often change during a hospital admission as clinicians start, stop, or continue therapies. We study whether models can predict which medication classes are added or removed between 24 hours after admission and discharge. Metrics that compare the complete discharge regimen can reward models for copying medications that remain unchanged, even when they identify no actual changes. We therefore introduce a leakage-controlled benchmark that predicts net ATC3 additions and removals using only prior completed admissions and information available within the first 24 hours of the current admission. Addition candidates are classes not active at 24 hours, whereas removal candidates are classes active at that time. We also introduce R-GEAN, an asymmetric candidate-scoring network with independent addition and removal predictors. Across 240,480 admissions from 82,286 patients, R-GEAN achieves the highest predefined summary of addition, removal, changed-regimen, and action-pattern performance, termed the edit composite (0.464), compared with 0.435 for the strongest primary comparator. Reimplemented RETAIN, GAMENet, and MICRON baselines obtain 0.428, 0.420, and 0.288, respectively. R-GEAN's advantage is concentrated in correctly identifying medication classes no longer active at discharge, while rare additions and admissions with multiple medication changes remain difficult. Rankings based on micro-F1 over the reconstructed discharge regimen and the edit composite correlate weakly across the evaluated models (Spearman r = 0.20). The continuation baseline achieves the highest complete-regimen score despite predicting no additions or removals. These results show that complete-regimen and edit-level evaluation measure different aspects of medication prediction. The benchmark evaluates observed prescribing changes, not treatment appropriateness

Parkinson's disease alters gait and bilateral coordination, but machine-learning performance also depends on how continuous gait signals are represented. This study investigates whether preserving anterior-posterior center-of-pressure (AP-COP) information at fixed locations across normalized stance provides a compact and informative representation of plantar-force gait signals. Bilateral vertical ground reaction force recordings from 165 participants in the Gait in Parkinson's Disease Database were evaluated using repeated fully nested participant-level cross-validation. We propose AP-COP10, comprising AP-COP position and bilateral asymmetry across five stance windows. AP-COP10 achieved an AUC of 0.894 and outperformed three harmonized literature-derived COP representations under the same evaluation pipeline. The complementary 25 non-AP-COP descriptors alone achieved an AUC of 0.856, while the complete 35-feature representation achieved 0.908. Removing AP-COP10 from the complete representation produced a statistically supported loss in discrimination, whereas adding the complementary descriptors to AP-COP10 yielded only a small, unsupported improvement. Feature competition indicated that the most informative stance-indexed descriptors were concentrated in early and early-mid stance, while source-study holdout and sensor-perturbation analyses supported the robustness of the representation. These findings indicate that stance-indexed AP-COP retains discriminative information that is not readily recovered by broader engineered gait descriptors, supporting compact and interpretable representations for machine-learning analysis of pathological gait.

Enterprise agentic systems that route every trajectory step to a frontier model waste 60-80% of their inference budget on subtasks that smaller models handle equally well. Existing routing solutions optimize single-turn query assignment but ignore a property unique to agentic workflows: subtask complexity varies widely within a single trajectory. A planning step may require frontier-class reasoning while a subsequent formatting step needs only a 7B model. We formalize step-level model routing as a sequential assignment problem over agent trajectories and propose AgentRouter, a lightweight classifier (12M parameters, <5ms overhead per step on an A100 GPU) that maps each trajectory step to one of four model tiers using five features extractable at routing time. Trained on 50,000 annotated agent trajectory steps spanning planning, coding, research, and data analysis tasks, AgentRouter achieves 72% cost reduction relative to frontier-only baselines, retaining 97.3% of frontier-only quality (less than 3% degradation in end-to-end task completion); per-step routing accuracy reaches 91% on minimal-complexity steps and 85% on efficient-tier steps, with 76-82% on the harder mid-range and frontier tiers. On the same benchmarks, RouteLLM and FrugalGPT (applied per-step) achieve only 31% and 44% cost reduction respectively, because their single-turn training signal misses trajectory-level quality dependencies.

While compressed-sensing regularizers enable interpretable reconstruction of 4D Flow CMR through transparent variational objectives, their hand-crafted nature is too restrictive under high acceleration. State-of-the-art learning-based approaches mitigate this, but typically encode regularization implicitly through unrolled network modules, which limits their interpretability. To address this limitation, we propose CLEAR, designed to combine the interpretability of compressed sensing with the flexibility of learned models. To the best of our knowledge, it is the first learned regularizer for a 4D reconstruction task. In the ultra-accelerated \(10\times\)--\(50\times\) regime of the CMRx4DFlow2026 challenge, CLEAR outperforms compressed sensing locally low-rank (LLR) and the popular variational network FlowVN, while using less than 10k parameters and preserving an interpretable regularization structure.

Existing prompt injection research focuses on single-model chatbot scenarios, where an attacker manipulates one LLM through crafted input. Multi-agent systems amplify this threat through three mechanisms absent from single-model settings: inter-agent message passing creates injection channels invisible to perimeter defenses, shared tool access enables privilege escalation across agent boundaries, and trust propagation allows a compromised agent to influence upstream orchestrators. We construct a threat model enumerating 14 attack vectors across four categories: direct injection via user input (3 vectors), indirect injection via tool outputs (4 vectors), inter-agent injection via message passing (4 vectors), and cascading injection through orchestrator manipulation (3 vectors). Testing all 14 vectors against a 6-agent production-representative system, we find that 67% of agents are vulnerable to at least one scope violation even with system-prompt-level guardrails, and indirect injection via tool outputs succeeds in 43% of attempts. Four architectural defenses reduce overall injection success from 31.2% to 4.2%: message signing with provenance tracking (inter-agent injection down 91%), input/output sanitization at agent boundaries (indirect injection down 78%), privilege-scoped tool access per agent role (privilege escalation eliminated entirely), and anomaly detection on inter-agent communication patterns (84% of cascading attempts caught).

Reinforcement learning (RL) is vital for optimizing video generation models, with a robust reward model (RM) serving as the cornerstone. However, existing video reward models often produce unstable scalar scores because they directly map complex, subjective video quality into a single score without explicit evaluation criteria. This leads to scalar drift, where the scoring scale collapses or shifts across different prompts, making the reward unreliable for RL. Drawing inspiration from professional human annotation engineering, we address this problem with RewardVerse, a rubric-based video reward framework that introduces a dynamic rubric as an intermediate representation between the evaluation query and the scorer. Instead of unconstrained direct scoring, RewardVerse first generates explicit evaluation criteria and then performs rubric-guided scoring, providing a stable semantic anchor that mitigates scalar drift. To efficiently optimize this collaborative pipeline, we propose Rubric-Guided Policy Optimization (RGPO), a two-stage training algorithm. RGPO first warms up the scorer using self-evolving seed rubrics and then jointly optimizes the rubric generator to produce query-adaptive evaluation criteria while continuously aligning the scorer with human ratings. Extensive experiments on the 16-dimensional EvalVerse benchmark and external datasets demonstrate that RewardVerse mitigates scalar drift, achieves state-of-the-art performance on both pointwise and pairwise evaluation, and provides a robust and interpretable reward signal for RL in video generation.

Autonomous AI agents increasingly act across organizational boundaries on behalf of human operators: they invoke third-party services, delegate subtasks to other agents, and pay for metered resources. Deploying such agents safely requires five capabilities that today live in separate systems: persistent identity, scoped delegation, peer trust, discovery, and payment. Existing approaches root these in centralized authorities or cover only subsets, so authority, trust, and payment fracture exactly where autonomy needs continuity: when a key rotates or a delegation must be revoked. We present NostrAgent, a decentralized architecture that unifies all five over Nostr relays using three custom event kinds: Kind 38100 identity declarations authenticated by BIP340 Schnorr signatures with pre-rotation commitments, Kind 38101 scoped delegation chains whose every hop verifiably narrows granted capabilities, and Kind 38102 peer attestations forming a Sybil-deterrent trust graph, with Lightning HTTP 402 (L402) binding payment to agent identity. Identity remains operator-sovereign without any registration authority; relays are substitutable transport rather than a trust root; and every authorization decision is replayable offline from signed events. We evaluate a Python prototype with a mixed-method design: ATAM quality analysis with a two-round mini-Delphi panel, STRIDE threat modeling across three trust boundaries, eleven benchmarks with non-parametric statistics, and 19 failure modes. Results show sub-millisecond offline verification, linear delegation-chain scaling, and Lightning-settled L402 at 157 ms median on regtest. 17 of 19 failure modes pass empirically, one is bounded analytically, and one is disclosed as an architectural limitation. NostrAgent demonstrates an auditable prototype substrate for trustworthy agentic systems without centralized trust roots.

Efficient large language model (LLM) instruction tuning requires selecting response supervision with supporting prompt context. Existing methods typically value both sides separately, risking selection-state mismatch between valuation and retained training subsets. BRIDGE (Budgeted Response-Prompt Interaction via Directional Gradient-guided Efficient Token Selection) captures selection-conditioned token utility through a shared validation-directed interaction surrogate valuing each side under the other's retained state. Budgeted alternating selection coordinates retained subsets by aggregating precomputed interactions over the current opposite-side subset to update conditional scores. Structure-aware projection converts conditional response scores into coherent supervision spans. Across three model families, BRIDGE leads compared selection methods overall in mathematical reasoning, code generation, and instruction following. In mathematical reasoning, its advantage over independent selection grows with compression.

Generative models are rapidly expanding image quality assessment (IQA) beyond traditional fidelity factors to emerging dimensions such as physical plausibility and text-rendering correctness. However, existing IQA models rely on fixed definitions and heavy supervision, making them difficult to extend to open-ended perceptual dimensions. We identify holistic bias as an important limitation: when scoring an unseen dimension, models reuse generic quality priors, leading to scoring errors and rank inversion. To address this, we propose PACE (Perceptual Agentic Collaborative Evolution), a training-free multi-agent framework that formulates open-ended IQA as explicit protocol construction. Given a target dimension, PACE uses collaborative agents to construct an evaluation protocol composed of verifiable Visual Question Answering (VQA) probes, grounding evaluation in concrete visual evidence rather than holistic impressions. The resulting protocol is calibrated using only four human-annotated images per dimension, while a dual-track scoring mechanism aligns model perception with human scoring scales. Across traditional IQA, structural fidelity, context-aware aesthetics, and newly defined open-ended dimensions, PACE consistently improves its MLLM backbone, achieving competitive performance across diverse IQA settings, and reduces the Holistic Override Rate (HOR) from 44.4\% to 8.6\%.

Long-context models read a novel the way a person reads a printout: one token after another, in narrative order, with the whole history competing for a fixed budget of attention. A detective does not work that way. They sort what happened when, and they keep a map of who relates to whom, so a clue from chapter one can meet a question asked at the end of the book. We test whether a frozen knowledge graph can give a small local model that same freedom. Thirty detective novels and 234 multiple-choice questions are answered by one fixed qwen3.5:9b reader under nine conditions: five graph routes, a recent-window baseline, whole-book compression, ordinary vector retrieval, and a question-only control. The strongest graph route reaches 53.85% (126/234) against 46.15% for the recent window, 51.28% for compression, 51.71% for vector retrieval and 40.17% for question-only. On the subset that no model can answer without the book, the graph route reaches 42.86%. None of the fifteen graph-baseline contrasts survives Holm correction, so we present the result as exploratory evidence about a design. Two structural findings survive scrutiny better than the headline number: annotated evidence concentrates in the topological core of these graphs (2.35x enrichment, pooled), and the two graph-building pipelines differ so much in annotation coverage (16% versus 73% of clue paragraphs) that pooled accuracy alone would hide which bottleneck is being measured.

Large language models (LLMs) increasingly mediate human decisions and communication, yet their behavioural regularities remain difficult to characterize systematically. We develop a cross-linguistic psychometric profiling framework and evaluate nine LLMs using seven psychological instruments, with five repeated administrations per model and language in Chinese and English. Items unresolved after a prespecified retry procedure are retained as NA. Joint analysis of scored and NA responses captures response tendencies and boundaries of self-report applicability. LLMs exhibit structured, model-specific profiles despite a shared alignment-shaped pattern of higher prosocial and self-regulatory responses and lower dominance, disengagement and harmful-intent endorsement. NA responses are structured rather than uniformly distributed, indicating where outputs are treated as inapplicable, refused or cannot be mapped to valid response options. Language condition and provider origin are associated with profile configuration and answerability, whereas repeated administrations show high reproducibility and permit recovery of model identity. Human-reference and prompt-robustness analyses further indicate that these signatures are context dependent. Joint analysis of psychometric profiling and answerability offers a framework for quantifying deployment-level behavioural signatures.

Federated learning (FL) enables collaborative model training without centralizing private data, but performance often degrades under feature skew: clients share labels while the conditional input distributions $p_i(x\!\mid\!y)$ vary due to latent, client-specific appearance factors. We propose Joint Domain-Class Federated Learning (JDFL), a lightweight, optimizer-agnostic extension that makes this latent domain variation usable without sharing raw data. JDFL first infers domain clusters called pseudo-domains from brief local update signals. It then expands the classifier head to output $M\times C$, joint (domain-class) logits. This allows the model to represent domain-conditioned appearance while keeping a shared backbone. To train the expanded head we introduce two complementary supervision strategies based on simple intuitions: a similarity-aware soft-labeling that transfers evidence between nearby inferred domains while allowing domain-specific specialization, and a per-sample randomized target assignment that perturbs supervision across the joint outputs and serves as a low-cost training-time regularizer. JDFL integrates with existing standard FL methods (e.g., FedAvg, SCAFFOLD) with minimal changes. Empirically, both supervision modes consistently improve global test accuracy on standard domain-shifted image benchmarks; ablations and sensitivity studies show the gains stem from the proposed supervision and parametrization rather than mere capacity increase.

Reinforcement learning in real-life applications requires safety guarantees during exploration. Typical reinforcement learning algorithms do not provide such guarantees, and many modifications that do rely on Gaussian processes (GPs), which have a large computational cost. We propose a computationally lightweight algorithm based on the Nadaraya-Watson estimator that safely explores and optimizes constrained Markov decision processes (MDPs). Our algorithm, \textsc{CoLSafe-MDP}, uses an estimator that scales in constant-time with bounds on the estimates, a significant improvement from its GP-based counterparts that scale cubically with the number of data points. We then evaluate its performance in a grid-based environment and on observational Martian terrain data.

Non-technical stakeholders frequently cannot write the SQL needed to extract insights from operational databases. We built and evaluated a Text-to-SQL agent that closes this gap end to end: a six-node LangGraph StateGraph checks question relevance, fetches the live schema, generates PostgreSQL, validates it with a dry run, retries on failure, executes the query, and narrates the result set in plain English. The agent uses prompt engineering only; no model was fine-tuned. We evaluated it on the Chicago Crime dataset (approximately 8.5 million records, 22 attributes) against a hand-built benchmark of 100 natural language questions with ground-truth SQL, stratified into 30 Easy, 40 Medium and 30 Hard items. Comparing two prompt revisions of the same agent, the revised system (V2) reached a Valid SQL Rate of 93% (from 87%), an Execution Accuracy of 60% under a hybrid relational equivalence metric (from 47%; 19% from 12% under strict JSON matching), and a mean Synthesis Quality of 4.34 out of 5 (from 3.91). The single largest driver was removing a LIMIT 10 instruction from the system prompt, which had been truncating multi-row answers. Error analysis attributes the residual failures to relevance-checker false rejections, ambiguous question semantics, and free-tier API rate limits rather than to the language generation step. We report no comparison against an external baseline system or a public benchmark; the study is a single-model engineering evaluation.

Talking-head and dyadic models now achieve real-time inference, yet fast motion generation alone does not produce an interactive conversation. A live system must synchronize the model's output with speech from an external voice agent, schedule video frames for playback, and handle interruptions. We introduce AVTR-1, an open stack for real-time interactive avatar conversations, built around a compact 153M-parameter autoregressive flow-matching motion generator conditioned on both participants' audio. We adapt its audio encoder for streaming through self-distillation. The stack turns the model's chunk-based generation into a continuous, synchronized audio-video stream driven by an external voice agent, and we analytically derive its contribution to the user-facing latencies and validate the resulting bounds with two commercial voice agents. Further experiments demonstrate that AVTR-1 leads the compared dyadic systems on all reported visual-quality metrics and most conventional listening-motion metrics while remaining competitive in lip synchronization. Its inference runtime operates in real time on data-center and consumer GPUs. However, conventional listening metrics do not establish whether the paired speaker's speech contributes to generated motion. We therefore introduce the Reference-Based Directed Granger Gain (R-DGG), which measures the additional predictive information carried by speaker speech after accounting for listener history and speaker motion. R-DGG finds statistically supported predictive dependence for recorded listeners and all evaluated dyadic systems, but not for talking-head generators without paired audio or mismatched speaker-listener pairs. We release the model weights, renderer, and serving backend under component-specific licenses.