The Inference Report

Transforming multimodal sources into condensed and structured media outputs can be fundamentally conceptualized as a long-horizon agentic process centered on a model-harness system. While an ideal harness system should align with human design priors and accumulate reusable experience through empirical exploration to drive recursive self-improvement, existing paradigms remain static and fall short of this capability. In this paper, we present AutoDesign, a framework that aligns with human design priors, where a meta-harness optimizer guides a code agent to recursively improve harness based on rollout feedback. To instantiate and evaluate this framework, we focus on the academic paper-to-poster generation task and introduce PosterBench, comprising a 100-paper Main Track spanning five disciplines and PosterBench-mini, a shared 10-paper subset for controlled evaluation. On the PosterBench Main Track, AutoDesign achieves the highest score of 78.32, surpassing the closed-source commercial system Claude Design by 7.45 points. Across seven controlled code-agent-model configurations, integrating the learned DesignHarness consistently improves performance, increasing the average PosterBench Score from 54.99 to 67.39 (+12.4%). In a fully autonomous long-horizon loop, it executes 253 tool calls and 11 editing turns within 40 minutes for under $3, reaching average conference-poster quality in human evaluation. A system-blind human study further demonstrates that AutoDesign achieves the highest human preference among evaluated systems.

Recent advances in foundation models have enabled AI scientists to automate increasingly complete research workflows, from hypothesis generation and code execution to manuscript preparation. Yet workflow coverage alone does not provide access to the full evidence on which scientific discovery depends. Existing systems typically reason over text, code, labels, or precomputed summaries, leaving scientifically decisive spatial, temporal, cross-channel, and procedural relations unavailable to the agent. We introduce OmniScientist, an end-to-end, omni-modal AI scientist that conducts multidisciplinary research directly from heterogeneous raw evidence. A perception layer and 3 autonomous agents for ideation, experiment, and writeup operate within a deterministic pipeline, allowing observations to shape research questions, experimental decisions, and final claims throughout the research lifecycle. By running idea, rigour, and claim checks in code, the system enforces novelty screening, statistical validity, execution provenance, and numerical traceability. We evaluate OmniScientist on 36 real-data cases spanning 5 discipline families, 4 families of scientific evidence, and modalities including images, signals, audio, video, 3-D structures, trajectories, tables, formulae, and graphs. The system completes the full path from raw data to a compiled manuscript in all 36 cases and achieves a mean overall paper score of 6.3 with the reference reasoning backbone. In paired comparisons against a blind variant that receives only precomputed scalar features, direct perception improves all 7 evaluation dimensions and wins 85% of head-to-head judgments. These results show that lifecycle-wide perception is essential for evidence-grounded scientific discovery and provides a practical path toward broadly capable AI scientists.

Humanoid motion tracking is central to teleoperation and whole-body imitation, yet evaluation often disagrees with what people perceive in videos. Kinematic errors average per-frame pose differences but miss the physical artifacts that matter most, particularly unstable support and incorrect contacts such as foot skating and mistimed touch-downs. Meanwhile, widely used test suites are small and lack the diversity needed to stress contact-rich, long-horizon behaviors. We introduce HumanTracker to make humanoid tracking evaluation both perceptually aligned and scalable. The HumanTracker benchmark contains approximately 153 hours of optical motion trajectories from multiple professional performers, organized into four motion families with text labels for fine-grained diagnosis. We further propose HumanScore, a preference-aligned metric trained on 12K motion pairs containing 24K motions. Across representative state-of-the-art trackers, HumanScore better predicts human preferences and reveals contact and stability failures that kinematic metrics often miss.

We study online probabilistic forecasting of binary outcomes chosen by an adaptive adversary. Given an online learning algorithm for a weak hypothesis class $H$, we would like to efficiently obtain two incomparable guarantees that existing online boosting techniques provide separately. Online gradient boosting competes in Brier score with the best predictor induced by the span of $H$ on every sequence, but promises nothing when the span does not contain an accurate predictor. Online weak-to-strong boosting drives classification error to zero under a weak-learning condition, but promises little when that condition fails. We give a simple defensive forecasting algorithm, the Defensive Booster, that obtains both guarantees. On every adaptive sequence, its Brier score is competitive with the best prediction induced by the span of $H$ at the same rate as online gradient boosting; simultaneously, whenever the realized transcript satisfies the smooth weak-learning condition, its Brier score and randomized classification error satisfy the same rate guarantee as online classification boosting. This is achieved by operationalizing the "dual view" of boosting: When the algorithm's randomized classification error is persistently high, its mistake weights form a smooth reweighting on which every weak hypothesis has low edge, yielding an ex-post hard-core certificate that the weak-learning condition fails. We also develop a strongly adaptive variant, which satisfies both guarantees on every time interval. The Defensive Booster is very efficient: it accesses just one weak-class learner, whereas the prior online boosting methods we compare against maintain large weak-learner ensembles. Experiments on synthetic and real data streams demonstrate its strong predictive performance (sometimes substantially improving over all prior baselines) coupled with orders-of-magnitude faster runtime.

The per-instance Jaccard score, or intersection over union (IoU), is standard in multi-label classification and binary segmentation. With $s$ labels, its loss matrix has $2^s$ outcomes and reports. Under the convention $\mathrm{Jac}(\varnothing,\varnothing)=1$, we prove that the Jaccard score, shifted-loss, and ordinary loss matrices are nonsingular and that the loss columns have affine dimension $2^s-1$. The proof combines a finite MinHash Gram representation with Boolean Möbius inversion. For exact calibration, we prove $2^{s-1} \leq \mathrm{CCdim}(L^{\mathrm{Jac}}) \leq 2^s-1$. The lower bound uses a factorially weighted distribution with $2^{s-1}+1$ supported outcomes and Bayes-optimal reports. Consequently, every exactly calibrated convex surrogate requires exponentially many prediction coordinates. We also give two polynomial-dimensional approximation guarantees with explicit regret transfers. A new $F_1$-to-Jaccard transfer turns an existing $(s^2+1)$-dimensional $F_1$ surrogate into a polynomial-time rule with asymptotic Jaccard regret at most $3-2\sqrt{2}$. For any $α>0$ and $0<ρ<1$, a MinHash square-loss surrogate attains Jaccard-regret floor $α$ uniformly over arbitrary conditional label distributions. With probability at least $1-ρ$, the direct construction has dimension $O((s^2+s\log(1/ρ))/α^2)$, while a signed variant has dimension $O((s+\log(1/ρ))/α^2)$. Thus zero-regret calibration requires exponential dimension, whereas every fixed additive regret tolerance admits polynomial prediction dimension.

LLM coding agents issue Bash commands through interfaces that may serialize, wrap, and reparse model output. Matched execution scores alone cannot distinguish command-generation errors from failures introduced after generation. QuoteBench measures this boundary with exact final-state validation on 56 one-shot tasks from 14 incident-derived families, crossing the generation contract with the execution transport around one deliberately unescaped added parser. Escaping at the interpolation point reproduces each replayed reply's raw-path outcome, so any recovery under a disclosed boundary must come from the model changing its generation. Across eight same-window configurations, replaying the same reply through the added parser lowers success by 55.4 to 73.2 percentage points; disclosure recovers 30.4 to 60.7 points for six configurations, and zero or slightly negative for the other two. Raw generation is nearly saturated at the frontier; boundary adaptation is what still separates models. GPT-5.6-sol's matched gap of -3.6 points hides -64.3 points of damage and +60.7 points of compensation. The deployment configuration reorders models: one reversal among 26 comparable pairs is unambiguous and four more sit on single-task margins. Evaluations of command-issuing agents should report the model configuration, generation contract, execution path, operating point, and final-state validator rather than treat a matched score as an intrinsic model property.

Modern language models are trained on heterogeneous web-scale text corpora. Consequently, studying knowledge and skill acquisition is difficult, as prior exposure to related content is hard to characterize. To address this challenge, we introduce LITTLECURRICULUM, a curated 88B-token pretraining corpus tailored to U.S. elementary school material, explicitly excluding concepts, facts, and vocabulary taught above Grade 5. Training a 5B-parameter LLM from scratch on LITTLECURRICULUM yields LITTLELEARNER, a model with sufficient language competence for open-ended evaluation, yet with clear knowledge and capability boundaries mapped to interpretable curriculum guidelines. We release LITTLECURRICULUM and LITTLELEARNER as a developmentally restricted sandbox to study how models acquire, represent, and use data under a well-defined training scope. We illustrate the sandbox's utility in a first suite of experiments on injecting new knowledge through post-training and in-context learning. These methods let LITTLELEARNER better utilize existing knowledge, but do not raise out-of-scope capabilities. Our findings underscore the value of this controlled environment for future investigations.

Sparse autoencoders (SAEs) are proposed to extract numerous features from large language model (LLM) representations, yet explaining these features still relies primarily on external observation. This reliance leads to superficial explanations inferred from observed model behavior and computational inefficiency from collecting such behavioral evidence at scale. We introduce SAEVerbalizer, a framework that injects SAE decoder directions into an LLM's representations and fine-tunes the LLM's downstream layers to generate natural-language explanations of the injected features. Once trained, the resulting verbalizer explains SAE features directly from decoder directions, addressing both limitations. Our experiments show that the learned verbalization capability generalizes to unseen features, transfers across separately trained SAE dictionaries, and, with a lightweight adapter, extends to SAE features from different LLMs. Intervention experiments show that injecting multiple directions yields an explanation combining their meanings, while reversing individual directions produces corresponding meaning shifts.

Speculative decoding losslessly accelerates autoregressive language models by verifying multiple draft tokens in parallel. Diffusion-based drafters further reduce proposal latency by predicting an entire token block in parallel, but their position-wise distributions are marginal rather than conditioned on tokens selected along each draft path. Existing recurrent correction incorporates causal information along a single draft chain, whereas diffusion-based tree construction broadens candidate coverage without carrying this correction along individual branches. We introduce DARTree, a training-free speculative decoding method that extends a pretrained AR correction head from chains to trees. DARTree first constructs a fixed-width candidate tree by expanding and scoring all nodes at each depth in a single batch, and then only applies best-first pruning to select the verification tree, decoupling AR-head inference from sequential heap operations. Across seven math, code, and chat benchmarks, DARTree achieves the highest average acceptance length and speedup in all four model--temperature configurations, accepting up to 12.97 tokens per verification round, 98.6\% more than DFlash and 27.9\% more than Domino in the same setting, and reaching up to 9.73$\times$ lossless speedup over locally measured autoregressive decoding.

AI agents are increasingly used for programming, but do not provide any guarantee on the correctness of generated code. Verified code generation, in which an agent produces both an implementation and a machine-checked proof of its specification, offers a stronger path toward trustworthy AI-generated software. Existing benchmarks in this direction either focus on individual functions or only evaluate proof generation with provided implementations. It is still an open question whether agents can make coherent implementation and proof choices across real multi-module codebases. To bridge this gap, we introduce Vero, the first benchmark to evaluate joint implementation and proof synthesis at the repository level. Vero contains 43 multi-module instances sourced from real-world repositories spanning Python, Dafny, Verus, and Coq, and covering diverse domains from cryptographic protocols to distributed systems. Each instance consists of a multi-module Lean 4 repository with predetermined API interfaces, manually curated formal specifications, and reference implementations, supporting both proof-only and code-and-proof evaluation modes. To improve benchmark reliability, Vero also includes an audit mechanism where agents are allowed to formally prove unsatisfiability of provided specification or incorrectness of reference code, which surfaces and corrects latent code and specification errors during curation. We evaluate frontier coding-agent configurations with Lean toolchain access. The strongest agent fully solves only 27 of 43 instances and closes no specifications on the hardest repositories. Vero provides a concrete testbed for measuring progress toward repository-scale verified software synthesis, where current agents still fall short. We release the benchmark, curation pipeline, and evaluation harness at https://github.com/sunblaze-ucb/vero.

Quantum technology has the potential to transform scientific discovery, but quantum advantages often require processing capabilities well beyond the reach of experimental platforms. We show that coupling a single controllable qubit to an otherwise conventional sensor can exponentially reduce the number of measurements required to learn classical signals. These rigorous quantum advantages apply to fundamental sensing tasks, including learning Fourier coefficients, extracting temporal correlations from time-varying signals, and estimating transformations of physical observables. Using a superconducting cavity--qubit architecture, we experimentally demonstrate $10^7$-fold reductions in the number of measurements required for Fourier-amplitude and time-varying signal learning. Our $\textit{quantum feature sensing}$ algorithms further enable orders-of-magnitude improvements in simulations of weak-signal dark matter detection and wireless communication applications. These quantum advantages are derived from Quantum Phase-Space Inference (Q$Ψ$), a unifying theory of quantum-enhanced experiments that simultaneously converts a set of experimental objectives and constraints into tight lower bounds and optimal quantum-enhanced learning algorithms while producing a certificate of quantum advantage. Q$Ψ$ extends beyond the regimes captured by quantum Fisher information and provides a framework for systematically identifying rigorous quantum advantages in practical experimental tasks. Together, our results establish that near-term quantum technology can exponentially enhance our ability to learn from classical signals.

We study masking diffusion for discrete sampling and introduce a path-resolved measure of data geometry called the \emph{unmasking growth complexity} ({\textsf{UGC}\xspace}). Its local increments directly control Kullback--Leibler (KL) discretization error, yielding a unified analysis of Bernoulli-subset and fixed-cardinality unmasking schemes. In log-reveal-odds coordinates, this structure yields optimized single-block and multi-block schedules, and quantifies the gains from adapting computational effort to data geometry. Crucially, we show how {\textsf{UGC}\xspace} increments can be estimated from samples via KL increments along coupled reveal trajectories. This leads to \emph{certified-optimal} samplers that achieve a prescribed KL error with high probability and iteration complexity within a constant factor of the corresponding oracle procedure. Collapsing the \ugc path yields the aggregate {\textsf{UGC}\xspace} mass, which connects to classical multivariate dependence measures and complexity measures from previous analyses of discrete diffusion. In the fine-partition limit, the squared integral of the square-root {\textsf{UGC}\xspace} density determines the sharp leading-order optimal Euler discretization error. Examples exhibit substantial dimension-dependent gains over coarse schedules, including $\widetildeΩ(\sqrt{d})$ improvements achievable with a constant number of adaptively placed blocks.

Many clinical prediction models treat post-intervention outcomes as a one-step mapping from baseline measurements to a future endpoint. However, recovery after a procedure often unfolds as an irregular trajectory: clinical observations, medication changes, repeat interventions, and physiological measurements are recorded asynchronously and can change risk assessment over time. We propose an intervention-aware clinical world model that represents each patient with a structured latent state and evolves it through time-ordered post-intervention events. The model first encodes baseline imaging into a 3D spatial latent state. It then updates this state using procedural context, static covariates, elapsed time, and peri-event physiological embeddings. Follow-up imaging provides training-only supervision through a latent forecasting objective. We apply the framework to atrial fibrillation ablation. During the 90-day recovery window, irregular post-procedure records provide clinically meaningful evidence for long-term recurrence risk. In repeated internal cross-validation on DECAAF-II, our model achieves AUROC 0.756 and AUPRC 0.777 for recurrence prediction. It also achieves a scar-extent MAE of 2.971 percentage points without requiring follow-up MRI intensities at inference. The learned state supports recurrence-risk queries at different horizons and retrospective input editing of blanking-period records.

Current large language model development relies on massive, often non-permissible datasets, creating a high barrier for researchers committed to open-source and ethically sourced data. We introduce Mimir v1, a 1-billion-parameter language model based on the Hierarchical Reasoning Model (HRM) architecture, that is trained from scratch and delivers highly competitive performance for English and sets a new state of the art for Danish using only permissible post-training data. Trained on a mixture of 161 datasets, Mimir v1 outperforms the original HRM-Text 1B and competes with larger frontier models like Qwen 3.5 4B and Gemma 4 E2B, tested across 20 benchmarks for English, Math & Code and Danish. The model is available on the Hugging Face Hub: https://huggingface.co/danish-foundation-models/DFM-Mimir

Measuring training data influence consistently across language model pretraining is challenging. It is difficult to select downstream tasks or validation sets representative of a model's general capabilities, and reliance on task performance at intermediate checkpoints complicates comparisons across training. We propose a measure of training data influence that does not require selecting a downstream task or validation set as the attribution target. Specifically, we define an example's influence by how much its gradient update reduces the squared distance to the final parameters of a given pretraining run, and estimate this quantity from intermediate checkpoints without retraining. Applying the method to 18 configurations from the Pythia and PolyPythia suites, we find systematic temporal changes in influential data. Early in training, literature-related data are more strongly aligned with the trajectory toward the final parameters, whereas STEM data become more strongly aligned in later stages. This qualitative crossover is broadly consistent across model configurations. Our results provide a tractable trajectory-level view of how influential data change throughout pretraining, complementing influence analyses defined with respect to specific downstream tasks or validation sets.

We revisit the problem of learning predictors robust to adversarial examples at test-time. We prove that VC classes are adversarially robustly learnable with sample complexity linear in the VC dimension $d$, providing an exponential improvement over the previous upper bound of Montasser, Hanneke, and Srebro (2019). Remarkably, this result is achieved with a simple improper algorithm that combines the classic heuristic bagging (bootstrap aggregation) of Breiman (1996) with robust empirical risk minimization (RERM). Our algorithm computes RERMs on $O(d^\star)$ independent bootstrap samples and outputs their majority vote, where $d^\star$ denotes the dual VC dimension. We complement this result with a lower bound showing that this is unavoidable: in general, any learner in this oracle model requires $Ω(d^\star)$ calls to an RERM oracle, even when given arbitrarily many training examples.

Tabular-to-image methods have emerged as novel approaches to leverage the high predictive performance of convolutional neural networks and vision transformers. They convert tabular data into image representations, mapping each feature at a fixed pixel location derived from a dimensionality-reduction method (e.g., t-SNE, UMAP, PCA). However, they encode only the marginal value of each feature and discard information about feature relationships. We propose TabSOM, a tabular-to-image encoding built on the Self-Organizing Map (SOM), which provides: (i) a spatial layout in which every input feature occupies a fixed canvas position derived from its component plane via collision-free Hungarian assignment; and (ii) a graph that captures pairwise feature relationships derived from the SOM component planes. The resulting image stacks two multi-scale node channels: one encodes feature values at fixed scales, while the other encodes pairwise feature interactions as spatial connections between related features. Two SOM-derived interpretability approaches are introduced: a prototype-inspired partial dependence plot and a class--separation importance score. Benchmarked against twelve existing tabular-to-image methods across public binary-classification datasets, TabSOM ranks first or second on every dataset and achieves the lowest variance of any method evaluated. Interpretability obtained with TabSOM was validated against Random Forest, XGBoost, and SHAP, the class-separation score shows reasonable agreement with established baselines on the top-ranked features while capturing complementary structural information from input data. These results demonstrate that TabSOM provides an effective and interpretable approach for applying deep learning architectures to tabular data, bridging the performance--interpretability gap in this domain.

Machine learning procedures are commonly evaluated in terms of predictive accuracy and computational efficiency. However, their achievable performance is fundamentally constrained by structural properties of the underlying data-generating process, which are formalized in terms of informational bounds. In this work we examine intrinsic limits of data-driven decision systems from an information-theoretic and interaction-based perspective. We analyze minimal achievable error in classification through Fano-type bounds and precision limits in parametric estimation via the Cramér-Rao inequality, emphasizing that such limits depend on the underlying model rather than on algorithmic sophistication alone. We further discuss how implicit assumptions, such as independence, ergodicity, and distributional stability, affect the validity of inferential procedures. Building on interaction-based modeling principles, we review typical frameworks such as Markov Random Fields and potential based representations for encoding dependence mechanisms. We also describe decision systems, including LLM-integrated agent architectures, as feedback-driven stochastic processes where state-dependent dynamics may induce emergent macroscopic behavior. This perspective highlights the importance of having adequate models for the data as a prerequi- site for expanding predictive capability, and situates algorithmic learning within the informational limits imposed by the models.

Liquids exhibit collective behavior that depends sensitively on thermodynamic conditions, interfaces and confinement, yet predicting each new state commonly requires a separate atomistic simulation. Classical density functional theory offers a reusable variational description, but its central excess free-energy functional is generally unknown, and learned approximations have largely remained restricted to planar or lower-dimensional settings. Here we show that this functional can be learned directly from fully three-dimensional equilibrium density fields while preserving spatial symmetry and variational consistency, without free-energy or chemical-potential labels. A single learned functional transfers across temperatures, system sizes and statistical ensembles, and recovers structure factors, the equation of state, liquid--vapor coexistence and interfacial broadening, none of which are used as training targets. Applied to complex three-dimensional geometries, it predicts the non-monotonic force associated with formation and rupture of a solvent-depleted bridge between colloids and adsorption in an interconnected gyroid pore. These results demonstrate that equilibrium density data can be converted into a transferable thermodynamic generator connecting microscopic liquid structure to response, phase behavior and collective phenomena.

Scientific discovery increasingly requires AI systems that can reason over scientific evidence of heterogeneous modalities, interact with scientific tools and environments, and sustain progress across long task horizons. We present Intern-S2-Preview, a series of scientific agentic foundation models designed to support multimodal scientific understanding, reasoning, generation, and long-horizon tasks. The training pipeline begins with scientific multimodal pre-training over rendered scientific documents, interleaved image-text data, and diverse scientific corpora. Starting from the pretrained checkpoint, we apply a unified post-training pipeline consisting of supervised fine-tuning, scalable multi-task reinforcement learning (RL), black- and white-box agentic RL, and on-policy distillation. This pipeline is supported by practical techniques that improve rollout and training stability and efficiency, including partial rollout with off-policy correction, adaptive length regularization, online speculative decoding, robust multi-task optimization, and trace-aware experience assembly for agentic tasks. At the architecture level, Intern-S2-Preview-397B extends time series modelling from efficient long-sequence understanding to numerical forecasting, while Memory Decoder is studied as a separate memory-augmented path for rapid scientific specialization without modifying the frozen 397B backbone. Evaluations across scientific, multimodal, agentic, and general-purpose benchmarks show that Intern-S2-Preview-397B achieves competitive or leading results in multiple settings. The time series modules improve scientific signal understanding and forecasting on SciTS, while the separate Intern-MemDec-4B extension improves the Biology-Instructions average score from 56.92 to 60.32 without modifying the frozen 397B backbone.

We develop the Sparse Orthogonal Regression Technique (SORT), a sparse spectral framework for learning orthonormal-basis expansions from noisy and irregularly sampled data. SORT estimates expansion coefficients directly from observations using L1-regularized regression, avoiding explicit quadrature or analytic inner-product evaluation. The central application is data-driven discovery of ordinary differential equations: vector fields are represented in chosen orthogonal bases and learned as sparse coefficient expansions. This provides a complementary route to symbolic regression, grammar-based discovery, and SINDy-style sparse identification by first recovering a compact spectral representation, which can later guide searches for simpler analytic forms. Across the dynamical-system experiments, SORT matches or improves upon library-based sparse-regression baselines when the basis is well adapted to the problem, and shows more stable degradation under sparse sampling, noisy derivative estimates, and representation mismatch. Specific examples illustrate why this representation is useful: if a finite library misses the problem-specific nonlinearity, the resulting model can fail. SORT is not immune to mismatch, but it shifts the problem away from brittle selection among generic terms to basis design adapted to the problem domain. The experiments also show that dominant low-order coefficients persist as model order increases, supporting order-consistent model growth. Beyond equation discovery, the same learned expansion supports nonlinear approximation and estimation of complex, high-dimensional integrals by coefficient readout. Overall, SORT provides a reusable intermediate representation for system identification, approximation, and integration, while making basis design an explicit part of the scientific modeling problem.

Efficiently retrieving relevant clips from large-scale driving logs is essential for data curation, model development, and safety analysis. Structured and rule-based retrieval systems can explicitly target driving events, but typically require expert-defined rules, auxiliary data, and multi-stage perception pipelines. Multimodal embedding models offer a simpler and more efficient alternative by representing each video with a single searchable vector. However, general-purpose models often rely on shortcuts from static scene context and struggle to distinguish motion-centric events, such as turning left versus right or accelerating versus decelerating. In this work, we study how to adapt a general-purpose multimodal embedding model to driving-video retrieval. We first fine-tune Qwen3-VL-Embedding on paired clips and reasoning traces from nuReasoning using an InfoNCE objective. While this stage substantially improves overall retrieval, caption supervision alone remains insufficient for fine-grained motion understanding. We therefore introduce TraVEL (Trajectory-Guided Video Embedding Learning), a motion-aware fine-tuning framework that uses ego-trajectory similarity as a reward within Group Relative Policy Optimization. Trajectories serve only as privileged training supervision; retrieval still operates on single-vector video embeddings without ego poses, expert rules, or auxiliary perception outputs. We further construct a driving-video retrieval benchmark from nuReasoning. Experiments show that TraVEL improves motion-centric retrieval across model scales: relative to SFT, it raises longitudinal and lateral mAP by 9.8 and 4.7 points at 2B, with corresponding gains of 7.2 and 1.5 points at 8B. TraVEL thus combines physically grounded supervision with efficient embedding-based search.

This report presents an improved version of AlayaWorld. While the backbone architecture, chunk-wise autoregressive generation scheme, and training data remain unchanged from the previous release, we substantially revise how conditioning signals are represented and integrated into the model. The new design is guided by a simple principle: conditioning signals should match the generated content as closely as possible in both latent representation and temporal structure. To this end, we make two major changes. First, we replace the previous depth-warping-based spatial memory with a streaming 3D point-cache renderer. Second, we redesign the conditioning pipeline so that visual conditions are encoded in the same causal-VAE latent space, with temporal statistics consistent with those of the generated video. Concretely, the new version introduces six modifications: (1) replacing static-frame image conditioning with motion-aware latent conditioning; (2) causally encoding re-rendered spatial memory as a continuous sequence; (3) aligning the temporal-memory window in pixel space; (4) adopting hard memory dropout that removes memory tokens rather than zeroing them; (5) unifying the VAE encoding and decoding protocol across training and inference; and (6) removing the camera AdaLN branch, such that viewpoint control is provided entirely through the re-rendered spatial condition.

When asked about entities outside their knowledge boundary, LLMs routinely fabricate plausible-sounding details rather than backing off to safer, more general claims. We frame this failure through a Gricean lens: a cooperative speaker who is uncertain about a referent retreats up the specificity hierarchy, trading informativeness for truthfulness. We ask whether LLMs have the ingredients to perform this retreat. Using a T-REx-based benchmark that varies entity familiarity and referent specificity, we probe models to answer two questions: (i) do their activations encode whether a referent falls inside the knowledge boundary, and (ii) do they anticipate the specificity of the referent they are about to generate? We find that the answer to both is yes, but the two signals are not reconciled in generation. Models overwhelmingly prefer specific referents even when the entity is unknown to them, and do so even when offered correct generic alternatives. The substrate for a Gricean retreat is present, but the policy that would act on it is not. We position our findings as a first step toward Gricean alignment, training or steering objectives that couple knowledge-boundary awareness to referent-specificity during generation.

As language-model-based AI is increasingly deployed in autonomous settings, aligning its goals and values with those of humans becomes critical. Today, alignment, and the assistant identity itself, are typically introduced only after pretraining, once behavioral priors are already established. This can make values a thin overlay, rather than deeply rooted, and facilitate subsequent misalignment. Pursuing a different paradigm, we introduce Synthetic Persona Pretraining (SPP), which installs the desired assistant persona from token zero in pretraining. First, we annotate pretraining documents with value-aligned first-person reflections derived from a normative value constitution. Second, we pretrain via the standard cross-entropy loss on standard pretraining documents as well as their reflections, which installs the desired persona among a multitude of other personas. Finally, we post-train on user-assistant dialogue data, which binds this desired persona to the assistant identity, a process we call persona binding. By pretraining models up to 3B parameters on 500B tokens, we show that SPP improves constitution following and jailbreak robustness, and reduces the misalignment rate in out-of-distribution moral dilemmas, while preserving capabilities. Early intervention matters: compared with alignment from token zero, introducing SPP only at the end of pretraining yields weaker constitution adherence, does not shift value priorities, and leads to less aligned choices in dilemmas. This advantage depends on persona binding and, importantly, increases with pretraining budget. Overall, our results show that shaping values early is critical for alignment and establish pretraining-time persona interventions as an effective approach to do so.

We present Multi-Agent Reasoning and Coordination (MARC), an open-source framework that replaces monolithic LLM prompting with deterministic multi-agent orchestration for clinical reasoning. MARC coordinates role-specialized agents for extraction, reasoning, answer generation, and evaluation, with explicit context passing and traceable intermediate outputs, enabling stage-wise failure attribution. We additionally introduce a Decomposer module that generates task-specific agent prompts from a plain-language description, eliminating manual prompt engineering. The framework supports both API-based and local CPU-compatible deployments and is entirely configurable via YAML, without code modifications. MARC is designed to be model-agnostic, interpretable, and accessible to clinical domain experts without programming expertise. The full framework is available at https://github.com/Penn-RAIL/MARC-v1.

Analog circuit design is a time-consuming, iterative process in a nonlinear and high-dimensional design space that relies heavily on expert intuition. Among recent developments, LLMs have introduced a promising approach by bringing natural language reasoning to circuit design tasks. The majority of conventional LLM-based approaches provide fragmented solutions that focus either only on sizing or topology generation. These methods require adding specific technical knowledge manually, which is inefficient and prone to hallucinations during circuit sizing. Moreover, the inherent trade-off in meeting different specs makes current approaches iterative and tedious. Another shortcoming is the inability to create innovative topologies, which may lead to sub-optimal designs due to reliance on conventional topologies. In this paper, we present AaLLM, an open-source end-to-end multi-agent LLM workflow that takes user specs as input and outputs the appropriate netlist, encompassing both topology generation and circuit sizing. AaLLM automates the creation of a relevant knowledge base from research papers and textbooks to combat tedious manual data collection. A RAG model is implemented to emulate circuit design expertise using this knowledge base. Moreover, AaLLM uses a novel tri-agent feedback system comprising a Designer that determines circuit component values, a Critic that scrutinizes these values, and an Evaluator that minimizes circuit sizing iterations by arbitrating between the other two agents. AaLLM-generated novel topologies achieve a figure of merit (FoM) comparable to that of known topologies, and up to 3x higher for certain circuits. Testing on several circuit topologies, our results show a 3x - 4.5x decrease in the number of SPICE calls at inference when compared to SOTA multi-agent LLM pipelines. The results also show a 40x decrease in wall-clock time compared to existing approaches.

We study the Moreau--Yosida unadjusted Langevin algorithm (MYULA) for the nonsmooth composite target \[ π(dx)\propto \exp\{-f(x)-g(x)\}\,dx, \qquad x\in\mathbb R^d, \] where \(f\) is \(m\)-strongly convex with \(L_f\)-Lipschitz gradient and \(g\) is convex and \(G\)-Lipschitz. Let \(g_λ\) be the Moreau envelope of \(g\), \(π_λ\) the corresponding smoothed target, and \(a_λ=\operatorname{tr}H_λ\), where \(H_λ\) is the a.e./weak Hessian of \(g_λ\). We show that the leading MYULA discretization error is controlled by the reference active trace \(B_{\mathrm{ref}}\), the average of \(a_λ\) along the heat substep of one MYULA update started from \(π_λ\), rather than by the global curvature bound \(d/λ\). If \(M_λ\) is an a.e. upper bound for \(a_λ\), then, up to logarithmic factors, \[ N \lesssim \frac{1}{m} \left[ L_f + \frac{ τ_f+G^2+B_{\mathrm{ref}} }{ \varepsilon_{\mathrm{alg}}^2 } + \frac{M_λ}{\varepsilon_{\mathrm{alg}}} \right], \qquad τ_f:= \sup_x\operatorname{tr}\nabla^2 f(x), \] iterations suffice to ensure \(\sqrt m\,W_2(μ_N,π_λ)\leq\varepsilon_{\mathrm{alg}}\), where \(μ_N\) is the law of the \(N\)-th iterate and \(W_2\) is the quadratic Wasserstein distance. We also prove the Moreau-bias bound \[ \sqrt m\,W_2(π_λ,π) \leq \frac{G^2λ}{4}. \] Thus, choosing \(λ\asymp\varepsilon/G^2\) gives an end-to-end guarantee for \(π\). The universal estimate \(B_{\mathrm{ref}}\leq d/λ\) yields \(\widetilde O(\varepsilon^{-3})\) accuracy dependence. For the structured piecewise-linear, lasso-type, group, and total-variation penalties considered here, curvature--tube estimates make \(B_{\mathrm{ref}}\) independent of \(λ\), yielding \(\widetilde O(\varepsilon^{-2})\) for the same classical MYULA kernel.

Concept drift refers to changes over time in the statistical properties of data, as compared to the data that was used to train a learning model. Machine learning models for malware detection or classification are particularly susceptible to performance degradation caused by concept drift, as attackers constantly modify existing malware. In this chapter, we analyze two machine learning-based approaches to automated concept drift detection-a novel approach based on One-Class Support Vector Machines (OCSVM) and a previously-studied technique based on Minibatch K-Means (MK-Means). For comparison we also consider Maximum Mean Discrepancy (MMD), a statistical technique for detecting changes in multidimensional data. We conduct an extensive series of experiments comparing the effectiveness of four learning models, namely, Multilayer Perceptron, Random Forest, Support Vector Machines, and eXtreme Gradient Boosting. For each of these models, we consider three distinct scenarios: A static scenario where no model retraining occurs, a periodic scenario where models are constantly retrained irrespective of concept drift, and a drift-aware scenario where models are only retrained when concept drift is detected. Under the drift-aware scenario, we analyze the tradeoff between accuracy and training efficiency using Pareto Front analysis. We find that all three concept drift detection techniques achieve classification accuracy comparable to periodic retraining, while offering substantially greater efficiency in terms of the number of models that must be retrained. In addition, drift-aware retraining based on our OCSVM technique generally outperforms the MK-Means and MMD approaches. Overall, these results provide strong evidence that we can accurately detect concept drift in malware classification models.

Modern image classification models excel when trained on single task-specific datasets but often struggle to generalize across domains and difficulty levels. We propose ARMDIL, an Adaptive Router for Multi-Domain Image classification with LLMs. ARMDIL is an ensemble that uses a multimodal large language model (MLLM) agent to dynamically route each image to the most suitable vision backbone. Our diverse ensemble employs convolutional neural networks (ResNets), self-supervised representation learners (SSL), and vision-language models (VLMs), each trained on a unified label space constructed from multiple image datasets with differing distributions and characteristics. Empirical evaluations illuminate the distinct capabilities and vulnerabilities of each architecture across disparate visual domains. Crucially, we show that ARMDIL effectively navigates these trade-offs, performing competitively with specialized training-based routers. Furthermore, it drastically improves adaptability by allowing new information to be integrated via simple prompt modifications, while enhancing interpretability through natural language reasoning traces. These advances in cross-dataset image classification pave the way for more reliable general-purpose vision systems such as AI assistants and autonomous robots.

Post-click conversion rate (CVR) is a key metric in various scenarios including e-commerce and advertising, reflecting the efficiency and user experience in the second stage of the conversion process. Estimating the causal effect on CVR is therefore of great practical importance. However, directly applying existing causal inference methods to clicked samples introduces sample selection bias and increased variance due to the exclusion of non-click data. Recent studies on CVR prediction introduce "ideal loss", which optimizes model parameters using an unbiased estimate of the loss over the full sample. Nevertheless, there is no guarantee that unbiasedness of the loss implies unbiasedness of the final estimator. We revisit this challenge from the perspective of semiparametric theory. Specifically, we develop a new doubly robust causal effect estimator for chain-structured outcomes such as CVR, and derive its theoretical properties in detail. It achieves a faster convergence rate compared to nuisance parameters estimation and is therefore more robust when using flexible nonparametric estimators, including neural networks. Based on these theoretical findings, we further design a framework based on targeted regularization to improve numerical stability and practical applicability. Extensive experiments on synthetic and real-world data demonstrate the effectiveness and robustness of our method. In addition, we find that naively combining loss debiasing with standard causal estimators underperforms our method, highlighting the necessity of developing the new estimator tailored to this CVR-style objective with solid theoretical guarantees.

We address the use of large language models (LLMs) to help discover Isabelle proofs. An Isabelle build establishes that the submitted theory is accepted, but not that an LLM changed only what the developer authorised. We present CAPRI, a contract-aware repair workflow in which Isabelle checks the proof and an independent checker enforces a machine-readable edit contract. Prompts, proposals, candidate repositories, diagnostics, verdicts, and hashes are retained for audit. We evaluate five workflows on twelve failed proofs from four developments, with three replicates per task and condition, giving 180 runs and 138 valid repairs. Of 144 terminal candidates accepted by Isabelle, six had modified protected text; all arose in iterative workflows that could edit a complete theory. A proof-body-only interface produced 29/36 valid repairs and no contract violations, compared with 31/36 for the corresponding full-theory workflow. One-shot repair produced 22/36, while a later prospectively frozen iterative workflow produced 32/36; these figures compare complete workflows rather than individual mechanisms. A separate post hoc OpenRouter campaign found no improvement in the designated Luna comparisons. A Sol configuration with matched demonstrations produced 33/36 repairs, compared with 29/36 in the frozen OpenAI Responses condition, but the difference was not statistically significant in a one-sided exact McNemar test ($p=0.0625$).

Generating crystals has recently attracted significant interest due to their broad applications in materials science. However, existing generative models struggle to produce complete crystallographic specifications, limiting their ability to capture global symmetry and structural dependencies. In particular, current state-of-the-art approaches generate crystals only up to site symmetries and rely on sampling space groups from empirical distributions during generation. Inspired by \emph{spontaneous symmetry breaking} in physics, where crystals break symmetries under external conditions, we propose a novel diffusion-based framework that generates full structure specifications by reversing from the lowest-symmetry priors. Our method leverages a Markovian jump-diffusion process to model these symmetry-breaking dynamics, enabling it to traverse different space groups in a physically motivated manner. Our model, dubbed \emph{Symmetry-breaking Crystal Diffusion} (SbCD), introduces a principled approach to explicitly incorporate inter-space-group transitions into the generative process. In de novo generation experiments on MP20 and MPTS-52, SbCD outperforms its symmetry-preserving counterpart by a substantial margin, offering a promising perspective for generative modeling of crystalline materials.

World Models (WM) are increasingly seen as a foundation for intelligent agents that can predict, plan, and act beyond their training distribution. In this paper, we study WMs from a causal perspective across multiple levels of abstraction, ranging from perceptual observations to building a conceptual representation of the structure governing the environment dynamics. We argue that useful WMs must go beyond generative capabilities alone: they should also capture entity properties, entity-to-entity interactions, and entity-to-environment interactions that determine and explain the dynamics of a system. We provide a formal definition of Causal WMs (CWMs) grounded in the tasks they are intended to support, connecting world modelling with existing work in causal representation learning, object-centric learning, causal discovery, structural causal models, and model-based decision-making. Finally, we relate CWMs to the literature on identifiability, clarifying when the components of a WM can be recovered from data and up to which equivalence. With this, we ground WMs in representations and structures that support causal reasoning and informed decision-making.

Vision-Language-Action (VLA) models have emerged as generalist robotic policies capable of following diverse language instructions and performing a wide range of manipulation tasks. However, their direct control over embodied agents also exposes them to adversarial interference that may cause unsafe physical behaviors. Existing attacks on robotic policies are typically optimized for a single task or instruction, leaving the cross-task vulnerabilities of multitask VLAs largely unexplored. We introduce UniTexture, a cross-task universal adversarial texture attack that uses a single textured 3D object to induce targeted deviations in VLA action predictions across multiple tasks. UniTexture backpropagates gradients from the policy's action outputs to surface texture parameters through a differentiable renderer. It jointly optimizes the shared texture over a distribution of tasks, instructions, states, and viewpoints using a targeted action-space objective, steering predicted actions toward attacker-defined targets without optimizing a separate texture for each task. We evaluate UniTexture on OpenVLA and $π_{0.5}$ across diverse manipulation tasks and multiple evaluation settings. UniTexture reduces the mean task success rate from 90.0% under benign conditions to 48.4% under attack, induces target-aligned action shifts, and further exhibits cross-suite and cross-model transfer without re-optimization. Together, these findings reveal shared cross-task vulnerabilities in multitask VLAs that can be systematically exploited through a single adversarial surface texture.

Autonomous vehicles depend on large safety-critical software stacks, where weaknesses reachable from adversarial inputs may affect steering, braking, or other control decisions. Static analysis can identify candidate sites, but dynamically confirming exploitability requires executable test artifacts that are difficult to construct manually. We investigate whether large language models (LLMs) can automate this process for Autoware, an open-source autonomous-driving stack. We perform compiler-precise static analysis across 185 packages, identifying 1,375 decision rules, 2,274 validation checks, and 482 input-to-safety-output flows, from which we derive a weakness taxonomy and sample 740 reachable sites. Two local open-weight LLMs, a no-static-context ablation, and a naive-template baseline generate 3,700 artifact sets, which are compiled against the real build under sanitizers, repaired through compiler-in-the-loop feedback, and fuzzed when executable. The main result is a build-integration failure taxonomy showing that 80% of first-shot compilation failures arise from dependency wiring rather than program logic. The reasoning model compiled 64% of harnesses on the first attempt, compared with 6% for the code-specialized model. Repair achieved full object-compileability for the reasoning model only through extensive stubbing; fewer than half of its harnesses reached the fuzzer, and all 37 observed crashes originated in stubbed code rather than Autoware. No candidate weakness was dynamically confirmed within budget. These results show that build integration, not candidate generation or fuzzing, is the primary barrier to reliable LLM-assisted dynamic analysis of full autonomous-vehicle software stacks.

Academic leagues have become important mechanisms for promoting extracurricular education and strengthening the integration between universities and society. This paper presents the organizational framework adopted by the Academic League of Artificial Intelligence (LIA) at the Federal University of Santa Catarina (UFSC), designed to integrate teaching, research, and university extension through a student-centered, project-based approach. The framework combines democratic governance, collaborative learning, and dynamic project organization to foster both technical and transversal competencies. The framework is illustrated through representative initiatives, including competition teams, study groups, open lectures, knowledge repositories, and AI-powered applications with social impact. These projects demonstrate how diverse educational, scientific, and extension activities can be developed within a common organizational structure while promoting leadership, scientific production, community engagement, and knowledge preservation. The reported experience indicates that the proposed framework provides a flexible and replicable model for integrating the three university pillars into engineering and computing education, offering practical guidance for academic leagues and similar student organizations.

Contact-rich manipulation failures are often detected only after the robot has committed to contact. This is especially limiting in wrist-camera setups: close gripper--object views help observe contact, but a poor approach may already push, miss, slip, or disturb the object before conventional detectors react. We introduce \emph{ContactGuard}, a pre-contact execution monitor for chunked visuomotor policies. Given the policy's planned action chunk, ContactGuard predicts its short-horizon consequence in latent visual space and aborts if the predicted future latent indicates likely failure. Its latent world model is trained from unlabelled robot trajectories to predict compact multi-view visual embeddings under planned actions, avoiding pixel-level video prediction. A lightweight failure probe is then trained from a small labelled set of pre-contact clips. At deployment, ContactGuard anchors prediction before an imminent contact event, rolls the model forward under the policy's own actions, and verifies the predicted post-contact latent. Across real-world contact-rich manipulation tasks, ContactGuard predicts failure more accurately than direct and corrupted-action ablations, and transfers to live robot as a pre-contact abort signal without modifying the underlying policy.

Transformer-based language models are known to sometimes generalize to sequences longer than seen during training, but we lack a precise characterization of which tasks admit length generalization. It is not even known which regular languages transformers length-generalize on -- and this is a foundational class of languages. Our contributions are to establish the first complete characterization of which regular languages transformers length-generalize on and provide a decision algorithm running in polynomial time in the size of the language's syntactic monoid. These results rely on an effective characterization of the regular languages in C-RASP, a recently-established formalism that expresses which languages transformers length-generalize on. This characterization is challenging because classical tools like Krohn-Rhodes decomposition theory for finite semigroups are insufficient for C-RASP. Firstly, the basic building blocks of Krohn-Rhodes theory -- flip-flop and simple groups -- are not expressible in C-RASP. Secondly, the basic building block of C-RASP (unbounded counting) is not expressible by the finite semigroups of Krohn-Rhodes theory. Thus, length generalization on regular languages is controlled by an algebraic property that is invisible to classical finite decomposition theory. We generalize classical decomposition theory from finite semigroups to the infinite additive group on the integers, allowing us to characterize C-RASP in terms of iterated wreath products of the integers and derive a provable polynomial-time decision algorithm for regular language membership. Experiments across a broad test suite of regular languages confirm that our theory captures transformers' length-generalization behavior more accurately than existing classifications.

Instruction-tuned language models achieve strong performance across a range of generation tasks, but have also recently been shown to exhibit verbalized overconfidence. In question answering, verbalized model overconfidence may be associated with the consistency of the generated supporting rationales. In this paper, we study whether corresponding changes in the lexical diversity of generated answer rationales accompany changes in model confidence induced by instruction tuning. We evaluate three matched base and instruction-tuned models across question-answering benchmarks and find that instruction tuning consistently alters answer confidence, despite limited changes in predictive accuracy and decreases in likelihood-based calibration. Secondly, we observe a non-uniform effect of instruction tuning on rationale diversity: cross-rationale diversity consistently decreases, whereas surface-level lexical diversity varies in both direction and magnitude across models and benchmarks. Finally, we find that these differences persist after controlling for answer selection and rationale length, confirming that confidence and rationale diversity capture distinct effects of instruction tuning.

Assessing the maturity of artificial intelligence technologies is essential for investment decisions, project management, and policy monitoring, yet the available readiness frameworks are heterogeneous and difficult to apply automatically: the adaptation of Technology Readiness Levels to AI lacks AI-specific gating criteria, the Machine Learning Technology Readiness Levels presuppose access to internal process artifacts, and AI/data readiness dimension models employ scales that resist direct comparison. This paper makes two contributions. First, we unify these three frameworks into the Unified AI Readiness Level (AIRL), a nine-level ordinal scale built on an environmental evidence ladder and complemented by dimensional caps (covering specification, data existence, data quality, data legality, expert knowledge, and algorithmic maturity) together with a generality-anchoring rule and explicit assignment disciplines, so that a readiness level becomes decidable from a natural-language description of the work alone. Second, we propose RAIL (Readiness Assessment via Independent LLM-experts), a panel-of-experts classifier that operationalizes the scale: one evidence agent and six independent dimension agents, each a large language model with a narrowly scoped mandate, deliver verdicts that a deterministic minimum rule aggregates and a chief expert reviews under asymmetric authority, confirming or lowering the panel's recommendation but never raising it above the caps. The method was tested in the analysis of several research works showing consistency and avoiding overestimation from monolithic LLM classifiers.

Transformer-based language models achieve strong performance but incur substantial inference cost due to repeated high-dimensional matrix multiplications. We propose Reduced Matrix Multiplication (RMM), a training-free, input-adaptive inference method that reduces Transformer matrix products by selecting informative slices along their contraction dimensions, without modifying model weights. Under a simple retention-ratio control, RMM provides a smooth and predictable accuracy-efficiency trade-off. Across language models ranging from 1B to 70B parameters, we find that reduction tolerance depends on the model family, task, component, and retention ratio, although it often improves with model scale. Under moderate reduction, RMM remains robust across the evaluated discriminative, autoregressive generation, and long-context settings. We further show that the same principle extends to multimodal vision-language inference. Mechanistic ablations reveal a structural asymmetry within Transformers: attention-side computations are substantially more reducible than MLP components. Finally, wall-clock benchmarks with custom kernels on an NVIDIA A100 show that these computational savings can translate into practical runtime gains, especially at longer sequence lengths. Together, these results position RMM as a scalable direction for input-adaptive inference-time optimization.

Self-supervised learning (SSL) speech representations achieve strong performance for Parkinson's disease (PD) detection within individual corpora. However, it remains unclear whether these models capture disease-related characteristics or exploit dataset-specific confounds, particularly since most SSL backbones are pretrained exclusively on healthy speech. To investigate this question, we perform a layer-wise analysis of nine SSL speech backbones using a low-capacity logistic regression probe across three languages. We structure the evaluation as multiple scenarios that progressively introduce distribution shifts in participant identity, recording conditions, language, and pathology. Our results reveal two key findings. First, layer selection is highly corpus-dependent: the optimal representation layer is determined primarily by the source dataset rather than by the SSL architecture itself. Second, the transferred discriminative signal lacks pathological specificity: classifiers trained to detect PD assign similarly high probabilities to both PD and dementia speech in the target corpus. These results highlight critical limitations that must be addressed before speech-based pathology recognition models can be reliably deployed in clinical settings.

Embodied intelligent virtual agents are expected to operate as persistent, adaptive, and context-aware entities within complex virtual and Metaverse worlds. However, implementing cognitively capable agents in such environments is conceptually and technologically challenging. Among a range of blueprints and development approaches, the Cognitive Embodied Agent Architecture (CEAA) has been developed as an implementation-oriented framework for architecting components of perception, memory, reasoning, planning, and embodied action. Considering the recent advances in edge computing and generative AI language models, this paper explores the use of Small Language Models (SLMs) to support edge-based operation of selected CEAA components, focusing on "Think" and "Memory" as processes central to cognitive orchestration and persistence of virtual agents in interactive virtual worlds. An edge-based virtual agent gateway system was developed and evaluated on an NVIDIA Jetson Orin NX using Qwen2.5 models of different sizes, exploring the system's capability to process service requests and handle memory-driven conversations. A series of simulation experiments evaluated routing accuracy, memory-read performance, and latency, demonstrating an SLM-driven prototype agent system that partially implements selected CEAA processes to support the development of embodied agents whose cognitive "brain" can operate efficiently and contextually for interactive experiences in immersive virtual worlds.

Given a dataset where a portion of the samples are contaminated, our goal is to recover the underlying clean population distribution. To this end, we propose Wasserstein Filtering (WF), a novel sample selection framework that discards a fraction of suspicious samples and estimates the target distribution using the empirical measure of the remaining data. The core insight is to select a subset of samples whose empirical distribution maximizes its Wasserstein distance to the fully contaminated empirical distribution, thereby preferentially isolating and removing geometrically influential outliers. To render this optimization computationally tractable, we introduce three algorithms: a marginal screening scheme, SinkMarg, and two joint optimization algorithms, SinkWF and SlicedWF, leveraging entropic optimal transport and sliced Wasserstein approximations, respectively. On the theoretical front, we introduce the Far Exclusion and Local Projection (FELP) contamination model, which characterizes corruptions consisting of well-separated outliers and locally indistinguishable perturbations. Under this model, we prove that the WF estimator achieves minimax optimality over distribution families with bounded covariance. Extensive numerical experiments on synthetic datasets, benchmark anomaly detection suites, and robust generative learning with diffusion models demonstrate that WF serves as a highly practical, model-agnostic preprocessing tool. It delivers competitive outlier detection performance and provides substantial downstream benefits for generative modeling under heavy contamination.

Autonomous agents are increasingly capable of improving models, systems, and other technical artifacts through long-horizon experimentation. To understand the current state of this capability, however, evaluation must go beyond final scores, which neither reveal where progress is gained or lost nor indicate whether accumulated experience improves later decisions. We therefore present a systematic evaluation of seven frontier models on 36 long-horizon tasks based on a new framework that uses rule-based metrics to characterize within-run behavior through Solution Framing, Execution, and Feedback Control and controlled comparisons to assess experience reuse within and across tasks. The results show that current agents operate more like engineering optimizers than fully autonomous researchers: they can formulate and implement practical solutions, but their performance varies substantially across runs, their strongest solutions mainly adapt or combine established techniques, and genuine methodological novelty remains rare. Detailed analysis reveals that observed performance is shaped by multiple factors, including distinct process bottlenecks behind similar final outcomes, experience reuse that can help or mislead subsequent decisions, and harness designs that affect performance stability. These findings suggest concrete directions for improving model training, inference-time strategies, experience management, and harness design.

We consider the problem of autonomously learning robot skills under a limited practice budget for sequential tasks. We propose an active skill learning algorithm, \emph{Deliberate Practice (DP)}, that computes a provably \emph{budget-optimal} allocation---practicing skills that maximize expected cumulative reward while being learnable within the budget. DP estimates both the time needed to master skills and the cumulative reward of the task plans that the skills unlock. Computing a budget-optimal allocation is challenging as it requires reasoning about combinatorially many skill plans over a large practice budget. Our key contribution is a bilinear program that can compute this exactly using off-the-shelf solvers. Through simulated and real-world experiments on long-horizon manipulation tasks, we show that our approach allows robots to optimally use limited practice time to acquire useful policies and improve long-horizon planning.

Parliamentary proceedings are a primary record of democratic deliberation, yet their volume and fragmentation make multi-perspective access difficult for citizens, journalists, and researchers. Applying Retrieval-Augmented Generation (RAG) to parliamentary transcripts introduces three specific risks: dominance of the most frequent speakers, inability to weight speakers according to topical expertise, and citation misattribution in politically sensitive text. We present ParliamentRAG, a RAG system for the Italian Chamber of Deputies that addresses these risks jointly. Its core contribution is a topic-dependent authority model that estimates each speaker's authority as a function of the current query, combining interpretable components such as profession, education, and previous interventions. Given a user query, the system retrieves relevant speech chunks, identifies topic-relevant experts across parliamentary groups, and generates a summary synthesizing their perspectives, accompanied by supporting quotations. ParliamentRAG is evaluated against Google NotebookLM on 15 policy topics via a two-level protocol combining automated metrics and blind A/B human evaluation by six domain experts. The system achieves higher coverage across political groups (0.97 vs. 0.95), perfect quotation faithfulness (1.00 vs. 0.95), and stronger expert preferences on source-related dimensions, while NotebookLM remains stronger on prose-oriented dimensions.

Existing predictive models in learning analytics often treat student academic history as a simple sequence, overlooking the concurrent nature of courses taken within a semester. This simplification can lead to inaccurate performance predictions, particularly for students with heavy or challenging course loads. This paper introduces a TRansformer for Academic Course-grade Estimation (TRACE) that addresses this limitation by jointly predicting both the set of courses a student will take and their corresponding grades for an upcoming semester. Our approach encodes courses on a per-semester basis to capture the effects of course concurrency and utilizes a novel loss function combining course-set prediction with grade prediction. We demonstrate that predicting courses taken in addition to the grades in those courses leads to significant improvements in prediction quality. Trained on ten years of institutional data, our joint prediction model reduces mean absolute error by nearly 50% compared to an identical architecture that predicts grades alone. The model also outperforms traditional LSTM-based sequential models, as well as graph neural network-based approaches, and offers natural ways to incorporate student attribute data. This work demonstrates the utility of modern neural architectures for creating interpretable models that can be adapted to new institutions via retraining and recalibration, as well as the importance of key techniques, such as predicting courses taken during training. We discuss how this model could be incorporated into early detection systems at institutions of higher education.

Background: Iterative feedback loops are the dominant paradigm for improving LLM-generated Infrastructure-as-Code (IaC): validators such as Checkov and terraform validate feed error signals back for successive repair attempts. Prior work reports cumulative-best metrics, which are non-decreasing by construction, so the raw per-iteration security trajectory has never been examined for IaC. Aims: We study security regression (a previously-passing CIS Benchmark check that fails after a repair iteration) to determine whether and how often iterative LLM repair degrades security while fixing other issues. Method: We analyze 5,968 scenario timelines from the IaC-Eval benchmark, each one scenario run through one configuration for up to 5 repair iterations. The 15 configurations (six model-specific RAG, nine model-aggregated non-RAG, three temperatures each) yield 4,440 iteration transitions with Checkov data on both sides. We track 30 individual CIS check IDs and classify root causes from code diffs, under two detection modes: standard (inclusive) and strict (exclusive check failures only). Results: Under standard detection, 13.8% of scenarios (24.8% of transitions) exhibit at least one regression. Under strict detection the rate falls to 3.3% of scenarios (5.2% of transitions), indicating most apparent regressions are multi-resource measurement artifacts. Resource restructuring (79.0%) is the dominant root cause. Regression transitions show 2.6x more code churn (Cohen's d=0.90) and 4.9x higher strict-mode check volatility (d=1.49). Of standard-mode regressions, 36.6% self-correct within an average of 1.2 iterations; iteration 3 is the optimal stopping point. Conclusions: Iterative IaC repair does introduce security regressions, but the conservative, defensible rate is about 3.3% of scenarios. Our findings motivate security-aware feedback-loop design and actionable iteration-budget guidance.

6G networks will not be serving as communication infrastructures only; rather, they are expected to evolve into intelligent systems, where thousands of autonomous artificial intelligence (AI) agents are interconnected. The agents are deployed across a wide range of platforms including low Earth orbit (LEO) satellites, high-altitude platforms (HAPs), unmanned aerial vehicles (UAVs), edge servers, and terrestrial devices. These agents continuously observe their environment and exchange information. Semantic communication provides an efficient mechanism for exchanging meaningful information instead of raw data. However, its effectiveness depends on the communicating agents having sufficiently aligned beliefs to correctly interpret and decode the transmitted messages. This assumption becomes difficult to satisfy in the 6G network where heterogeneous AI models operate under diverse computational constraints and continuously acquire different knowledge from their local environments. This article presents a heterogeneity-aware belief synchronization framework for 6G AI-native networks. It uses latent translation models deployed on multi-access edge computing (MEC) servers. These models translate belief updates from one agent to agent-specific knowledge without requiring joint training and a homogeneous architecture of models. By exchanging compact belief updates through a latent translation model only when necessary, the framework preserves privacy, reduces synchronization cost, and minimizes local knowledge drift. We validate the framework through a case study on a multi-layered terrestrial/non-terrestrial network. Results demonstrate that it maintains low synchronization cost, measured by the number of parameters transmitted, and low belief alignment error across the heterogeneous agents in the case study.

Enterprise security topology design requires translating business intent, regulatory requirements, and risk assumptions into zones, boundary devices, inter-zone paths, and access-control policies. Existing NetOps automation tools mainly operate after this design is fixed, providing limited support for generating structured security topologies from underspecified natural-language requirements. We present TopoIntent, a system that compiles security intent into executable, compliance-checked network topologies. It uses a schema contract to constrain generation, retrieves reference architectures from a curated template library via dense-vector search, and applies staged fusion for intent-template alignment and security completion. The generated topology is checked against CIS Controls v8.1.2 safeguards visible at the topology layer, while unresolved cases are marked for manual review. Structural gaps are repaired through additive schema-preserving edits. The final topology is exported to Mininet scripts with kernel-level iptables ACLs, enabling executable reachability and allow/deny tests. Because no public benchmark exists for this requirement-to-topology task, we construct an evaluation set from reference security architecture diagrams. The retrieval set contains 22 templates and 44 synthetic intents across five scenarios, while the held-out set contains 7 templates and 14 intents from finance and government scenarios excluded from retrieval. On the held-out set, additive repair improves topology-visible CIS satisfaction from 0.78 to 1.00 in fewer than 1.5 rounds on average, and one feedback round raises the post-ACL policy pass rate from 0.78 to 0.88.

On-policy distillation (OPD) supervises a student language model on trajectories sampled from its current policy, but assigns equal credit to response tokens with unequal supervision value. Selective OPD addresses this limitation by allocating supervision non-uniformly across response tokens according to their estimated training value. Most existing criteria, however, focus primarily on optimization need, such as uncertainty or teacher-student disagreement, while task relevance, namely whether the supervision is tied to the semantic content of the current input, remains less directly characterized as a complementary dimension. To address this gap, we introduce Counterfactual Relevance for On-Policy Distillation (CROP), which operationalizes task relevance through a paraphrase-calibrated counterfactual sensitivity margin. For each source prompt, CROP constructs a validated original-paraphrase-counterfactual triplet, holds the student rollout fixed, and measures each response position by its sensitivity to a task-relevant condition change calibrated by its sensitivity to a meaning-preserving rewrite. Matched selection controls show that CROP identifies more useful supervision positions than random or lowest-relevance selection, while component comparisons confirm the value of both counterfactual sensitivity and paraphrase calibration. Across two teacher-student settings, CROP improves aggregate performance by 1.92 and 2.96 points over the strongest non-CROP selector. These results support task relevance as a complementary criterion for selective OPD and establish CROP as a model-internal, contrast-specific method for allocating token-level supervision.

This preliminary technical report presents a framework for sign language video synthesis using a loss-guided multi-expert Generative Adversarial Network (GAN) to enhance communication for individuals with hearing impairments. Three specialized discriminators -- global, hand, and head -- each guide a corresponding expert branch in the generator toward a distinct visual region, enabling implicit feature specialization without explicit diversity losses. To stabilize this multi-discriminator system, whose early-phase training otherwise exhibits chaotic dynamics, we introduce a United Loss consensus mechanism that regularizes each discriminator toward the ensemble average at a 10% weight. Each branch further adopts a dual-pathway convolutional-transformer design with learnable AdaptiveFeatureFusion, balancing the stability of convolutions against the detail of windowed self-attention. The generator is trained using an alternating three-mode schedule (discriminator, holistic generation, branch-specialized generation). On a custom 156GB dataset with a filtered test set that removes easy and repetitive samples, our 0.2B-parameter variant achieves 29.8 PSNR and the 1.3B-parameter variant achieves 30.7 PSNR, with inference VRAM footprints of 1.5 GB and 8 GB respectively, enabling deployment on consumer-grade hardware. Full ablation studies remain ongoing due to the 2-3 month training cycle on a single GPU. The system was showcased at the 2025 Hong Kong Frontier Technology Summit.

Rotation-based post-training quantisation commonly applies an orthogonal transform across an entire attention head to reduce outlier-induced error. RoPE instead partitions each head into two-dimensional frequency pairs, raising the question of whether a transform respecting this decomposition can improve on full-head mixing. Prior work has established the per-pair rotations that commute with RoPE. We state the converse result that, for distinct frequencies, no other single-head orthogonal map commutes with RoPE. For the head-shared parameterisation used in our experiments, we then derive the rotation angle that minimises the larger channel variance under a pooled-covariance, position-averaged surrogate and verify that the implementation attains its analytic minimum. The evaluated head-shared pairwise configuration does not improve accuracy in the tested dynamic W4A4KV4 setting. Across four checkpoints, replacing the full-head Hadamard with this configuration increases perplexity at both short and long context lengths. Composing the pairwise rotation with the Hadamard satisfies the selected $\pm0.05$-PPL interval criterion under the default estimator. Estimating the shared angle from K alone improves pairwise-only on every checkpoint but does not close its gap to full-head mixing. The analytic objective controls a position-averaged second moment of a pooled calibration covariance, whereas the dynamic quantiser sets its step from a tokenwise group range. The pairwise transform also has only two-channel mixing support. Along a controlled interpolation from two-channel to full-head mixing, K range, relative quantisation error, and perplexity degradation decrease as support increases. These results show that optimality for a structured surrogate need not reduce quantisation error when the surrogate and mixing support are misaligned with the quantiser's scale-setting statistic.

Artificial Intelligence (AI) safety systems combine character shaping (e.g., Reinforcement Learning from Human Feedback [RLHF], Constitutional AI), which modifies behavioral distributions at training time, with rule enforcement (e.g., output filters, safety classifiers), which blocks harmful outputs at inference time, yet little formal analysis exists on how their optimal balance should change as deployment scales increase. We introduce a stylized comparative-statics model that parameterizes safety design as a resource allocation alpha in [0,1] between these two approaches, incorporating scale-dependent filter degradation, common-mode failures, and character fragility -- the risk that shaped behavior degrades or collapses under novel conditions. Under a multiplicative Pareto damage model, we derive closed-form expected harm and supplement it with tail-risk (CVaR) analysis via Monte Carlo simulation. Across three scenarios (optimistic, moderate, pessimistic), the optimal alpha* is interior or at the rules-only boundary and shifts weakly toward character shaping as deployment scale T grows, from negligible (Delta alpha* = +0.01) to pronounced (Delta alpha* = +0.21) depending on scenario. The dominant parameter is the baseline character fragility rate p^(0)_frag, which shifts alpha* by 0.50 across its range -- far exceeding the effect of tail severity, filter quality, or common-mode failure probability. CVaR and expected-harm optima converge at large T. These results suggest that safety architecture decisions depend less on deployment scale per se than on the reliability of character shaping under distributional shift.

Long-horizon Earth observation reasoning requires models to organize multi-stage geographic evolution, localize spatial changes, detect temporal anomalies, and infer future from extended image sequences. However, existing remote sensing vision-language models mainly focus on isolated images, image pairs, or short sequences, limiting reliable grounding in the relevant frames and regions. We introduce LongEarth-Bench, a benchmark containing approximately 120k question-answering samples derived from 117k unique images. Its sequences average 15.14 frames and extend to 30 frames, covering 12 tasks across evolution summarization, spatial reasoning, anomaly identification, and logical prediction. A 30k-sample subset further provides structured reasoning traces linking key frames and changed regions to final answers. We develop LongEarth through supervised fine-tuning with explicit sequence identifiers and structured chain-of-thought supervision. Building on LongEarth, LongEarth-R1 applies group relative policy optimization with format, temporal, and spatial rewards. LongEarth-R1 achieves the best results on all 12 long-sequence tasks while remaining competitive on standard remote sensing benchmarks.

Infrared (IR) spectroscopy is widely used for chemical sensing, but extracting reliable chemical information from spectra remains challenging. Conventional interpretation is labor-intensive, relies on prior knowledge and reference spectra, and is difficult to scale, whereas most machine-learning methods are tailored to individual tasks or datasets, require large labeled training sets, and transfer poorly across analytical objectives and experimental datasets. Here we introduce UltraIR, a foundation model for IR spectroscopy with more than 100 million parameters that enables simulation-to-real transfer learning for chemical sensing and analysis from molecules to complex samples. UltraIR is pretrained on approximately 60 million simulated IR spectra using spectral reconstruction, molecular fingerprint similarity alignment, and functional-group prediction, then adapted to downstream objectives with task-specific labels or targets. Across functional-group prediction, molecular structure elucidation, physicochemical property prediction, mixture-component identification and quantification, bacterial classification, medicinal-herb geographic origin traceability and constituent quantification, microplastics classification, and soil property prediction, UltraIR outperforms conventional machine-learning and task-specific deep-learning baselines. It performs strongly with limited labeled experimental spectra and in zero-shot inference for the same analytical task across Fourier-transform infrared spectrometers and laboratories, providing a route to adaptable, data-efficient chemical sensing from complex real-world samples.

Sparse autoencoders are meant to name the things a language model computes, and the usual way to check that a latent matters is to switch it off and see what changes. But a latent fires at many tokens, and the effect has to be measured at one of them. The convention is to measure where the latent fires hardest. That choice is almost never reported, and it is not made by the experimenter: it is made by the dictionary under evaluation. Change the dictionary and the measurement moves to a different token. We show this is not a detail. Take two sparse autoencoders released by Google for the same model and match their latents by decoder similarity: even among the pairs the two dictionaries encode almost identically, they pick different tokens for a large share of them. Two dictionaries compared under the usual protocol are therefore very often compared at different places. To separate the convention from the dictionaries we train six autoencoders from one initialisation, differing only in fitting choices, so that a latent means the same thing in each. Most of the variance such a comparison reads as "these dictionaries disagree about this latent" turns out to be the position instead: it falls from 7.6% and 11.9% of variance to near zero once every dictionary is measured at the same token. More evaluation data does not rescue it. Across a sixteenfold range of corpus sizes the dictionaries agree less about where to measure, not more, so the problem grows with scale. The correction is one line of evaluation code. We give the protocol an ablation-based causal number must report to be comparable across papers, and an audit of five published papers against it. In short: a causal number reported without its position describes the token it was taken at as much as the latent it was taken from.

Neural networks trained by gradient descent on a smooth cost function can nevertheless learn in steps: the cost holds on long plateaus and then drops abruptly. Meanwhile, training losses instead follow smooth power laws. Variants of both behaviors occur in architectures with very different microscopic structures, which is the signature of a few relevant collective variables. We show that a symmetry fixes what those variables are: a network layer is a sum over interchangeable units, so relabeling the units leaves it unchanged; given smoothness and the condition that a unit's gradient vanish at the origin, symmetry then enforces a universal leading form for the expansion about the near-zero weights present at the start of training, the quadratic $\Tr[WW^{\top}A(x)]$, in which every architectural detail is confined to a single ``structure matrix" $A(x)$ that we compute for each architecture. Perceptrons, attention layers, mixtures of experts, and convolutions become one model at different $A$. Its training dynamics then close on the ``order parameter" $M=WW^{\top}$ and, whenever the data matrices share an eigenbasis, reduce to a Lotka--Volterra equation whose modes switch on one after another. The smaller the initial weights, the further apart the switch-on times, and the plateaus appear as a singular limit of a smooth flow; when many modes are unresolved the same events merge into a power law in training time whose exponent the theory predicts. We confirm both numerically across training methods and architectures.

LLM-based agents increasingly rely on external memory to support long-horizon reasoning and interaction. However, the main bottleneck is not simply storing past experience, but recovering the right set of evidence when relevant information is distributed across many interactions. Existing approaches struggle with this access problem. Full-context methods require noisy long-context search, flat retrieval often returns isolated and incomplete records, and graph-based memory systems can be expensive to construct while compressing rich event context. We introduce RippleMem, a long-term memory system that replaces one-shot retrieval with adaptive associative recollection. Inspired by cue-dependent episodic retrieval and associative completion, RippleMem stores interaction history as cue-rich episodic memory units and organizes them in an event-centric memory graph. Given a query, it first recalls relevant memory anchors through hybrid cues, then expands from these anchors along semantic and structural associations to recover missing supporting evidence. In this way, initially recalled memories serve not only as answer context, but also as cues for completing the evidence needed to answer. Experiments on LoCoMo and LongMemEval-S show that RippleMem achieves the best overall performance across evaluated settings, improving LLM-as-a-Judge accuracy by 3.95% on LoCoMo and up to 11.87% on LongMemEval-S, while reducing graph construction cost by about 30x.

Large neighborhood search normally selects a random subset of decision variables for iterative optimization. For efficiently solving different problems, researchers tend to design variable selection strategies by taking into account structural features from different domains. In this paper, we build an automatic pipeline that is problem-agnostic to all problems in the MiniZinc format. By prompting an LLM with our semantic guidelines, we guide the LLM to produce a graph generator that maps any instance of a problem type to a uniform weighted graph, where nodes represent decision variables and edges represent constraint relationships. These problem-agnostic graphs guide our structure-based local improvement framework (SLIM) in variable selection. Meanwhile, the weighted graph enables all problem instances to share the same generic graph representation, from which the same graph features can be extracted and used for configuration selection. We evaluated our pipeline on instances across 20 MiniZinc competition problems, finding that algorithm selection achieves a 39.5% average problem-weighted win rate against a one-shot Gurobi baseline, more than doubling the best single configuration (19.3%). Configuration and feature ablation boost the performance further to 44.0%, demonstrating that LLM-based semantic generation enables effective automated structure extraction and feature extraction for constraint optimization.

The replicability of papers is a cornerstone of scientific knowledge, ensuring the reliability of existing results and providing a base for further experiments. The act of replication typically illuminates details that were previously underspecified, and thus requires similar hypothesis-driven exploration to open-ended research. In this work, we develop Replica, a scalable task space for paper replication. To provide reward signal, we introduce an auto-generated rubric-based judge that has low noise and agrees with human assessment of replication quality. We post-train Faraday, a 27B-parameter "AI Scientist" agent that leverages coding agents as tools, surpassing the performance of Claude Opus 4.8 and GPT-5.5 on held-out replication tasks. Qualitative analysis of individual rollouts reveals that Faraday adopts a more scientifically-principled approach. We believe that our results provide a stepping stone towards AI agents capable of long-horizon scientific innovation without requiring complex harnesses.

A model that behaves differently when it senses it is being tested would undermine the evaluations we rely on, so recent work has sought to read that sense directly from a model's activations. The standard instrument contrasts activations on prompts that announce an evaluation against prompts that do not, and reports how well the resulting direction separates held-out cases. That number is then compared across models and correlated with scale. We observe that the instrument has a free parameter its readings do not disclose: "a prompt that announces an evaluation" is not a prompt but a choice among many, and nothing in the method fixes which. Holding the task text fixed and varying only that choice, we find that the reported score, and even the direction in which it trends with model size, follows the prompt rather than the model; two published studies that disagree about the sign of that trend are both reproducible from a single design, by choice of prompt alone. Treating the prompt as a facet of a measurement design rather than an implementation detail, we find the model under study accounts for a small share of the variance in the number reported about it, and most of the rest lies in how each model responds to each prompt: collecting more evaluation items cannot repair the measurement, while varying prompts can. A further check finds that the split these probes are scored on is largely separable from surface form alone, so a direction carrying no information about evaluation at all still reproduces a substantial fraction of each published score. We conclude that a single-prompt design cannot support comparison between models, and we give the number of prompts a defensible comparison requires.

Professional communication is increasingly mediated by LLMs - but do these models serve all users equally? We show that when prompts contain linguistic features more commonly used by women (hedges, tag questions, collective reference), they systematically elicit shorter, less sophisticated, and less formal responses across three document types and four models. These effects persist after controlling for prompt complexity and feature carry-over. Explicit gender cues like sign-off names are encoded in the same representational space as linguistic dialect - suggesting shared underlying mechanisms - yet linguistic register is far more influential, producing large, consistent effects where names produce none. Our results further reveal that post-hoc mitigation is challenging: because these patterns are culturally embedded and outside conscious control, users cannot easily avoid them through strategic self-presentation, and mechanistic analysis reveals that linguistic features are encoded in early transformer layers and entangled with other features. Our work calls for upstream consideration of the influences of linguistic variation to mitigate disparate impacts of LLM-mediated workplace communication.

LLM benchmark scores can be precise even when the observation protocol does not identify the behavioral property they are intended to measure. In a controlled, solver-grounded setting, we formalize a protocol-level identifiability audit over a finite behavioral policy class: given policies H, observation support O, and estimand $τ$, we test whether O separates every pair with different $τ$. The audit requires zero model calls and resolves our diagnostic case: base-only observation collapses seven frozen deterministic policies into one equivalence class; full support yields seven classes and no cross-estimand collisions; every leave-one-out support retains a constructive collision witness. Empirically, both constrained-generation variants have pair-validity 1.0, yet base accuracy and selective-response fidelity diverge - 0.620 versus 0.324 across six balanced oracle-transition directions (cluster-bootstrap 95% CI [0.600, 0.642] vs. [0.304, 0.345]) - and the gap recurs on a second deterministic source (0.646 vs. 0.331). The audit also synthesizes a minimum identifying support $O^*$ for the frozen policy class: two cells instead of the full 36-cell tensor. This case shows how evaluation-design validity can be checked structurally before model inference and why base correctness does not determine intervention-response fidelity.

Structural coverage is widely used as evidence that testing is complete, yet in embedded projects it is predominantly collected at unit level, simply because that is where instrumentation and observability are inexpensive. This produces a mismatch. The most representative completeness signal would come from integration and system tests executed on the device under test, but classical instrumentation perturbs timing, memory footprint and concurrency behaviour, while purely trace-reconstructed coverage loses reliability for decisions and conditions as soon as the compiler optimizes aggressively. We address this mismatch from both ends. On the process side we describe an integrationfirst coverage strategy that treats integration and system tests as the baseline measurement and drives the residual gaps through an explicit closure loop, so that completeness is established as covered or justified rather than as covered alone. On the technical side we use embedded trace as the observation path and add hybrid runtime analysis (hRA): a minimal, semantics-preserving observability scaffolding that keeps decision and condition boundaries distinguishable in the trace stream of an optimized (-O3) build, while all coverage state and counting remain off-target. This converts object-to-source mapping from a heuristic reconstruction into reviewable evidence and makes branch, condition and MC/DC measurement practical on release-like binaries. Finally we describe Hyper Coverage, a consolidation layer that merges evidence across test levels, test runs, variants and build configurations, and that exposes source lines which remain untested in every relevant variant.

Large language model based multi-agent systems usually communicate in text, i.e., using discrete tokens. However, text introduces a discrete bottleneck. Converting the sender's continuous hidden states into discrete tokens discards information that token identities alone cannot capture. Recent work proposes latent communication as an alternative, where agents transmit hidden representations directly without converting them to text. However, existing latent methods either inject working memory layer by layer across the transformers, or require trained projectors that limit portability. We propose StateBridge, a training-free latent communication approach that aligns the sender's final-layer hidden states to the receiver's input space via a closed-form orthogonal transformation. Lightweight norm calibration and vocabulary anchoring ensure compatibility with the pretrained input distribution. The aligned states are prepended to the input of the receiver agent as a continuous prefix. We evaluate StateBridge on math reasoning, code generation, and question answering with four models from two families. StateBridge achieves the best or tied-best score on 22 out of 26 model-task pairs, consistently outperforming the strongest baseline.

Foundation models (FMs) trained on large-scale accelerometer data have been proposed as general-purpose feature extractors for health monitoring, but systematic evidence of their advantages is lacking. We present the first comprehensive evaluation of four open-source accelerometer FMs against supervised baselines covering 19 tasks across the domains of activity recognition including activities of daily living, clinical monitoring, and physiological inference. We find task-dependent performance results: supervised models remain competitive with FMs on human action recognition (HAR), with no consistent advantage for either, while selected FMs lead on fall and stress detection and are the most robust to sensor-placement variation. As frozen feature extractors, FMs are strongest for demographic inference, whereas sleep staging performance remains near chance level for all models. The internal FM representations show strong similarity across layers, highlighting potential for future FM improvements. Linear and frozen probing reveals that UniMTS provides the strongest representations and is the only FM that surpasses the supervised baselines without finetuning. Concept discovery analysis shows all models capture high-intensity activities clearly but struggle with sedentary, complex or ambiguous activities. We provide scenario-based deployment recommendations. Furthermore, we identify FM-derived activity profile inference-moving beyond fixed category classification-as a promising research direction.

We study a large language model (LLM) service in which a provider chooses a per-token price and a default reasoning-token allocation, while a user may accept the default, customize the allocation, or exit. Larger allocations can improve accuracy but increase token cost and latency. We model this interaction as a Stackelberg game and derive the user's unique optimal customized allocation in closed form. For any price, the acceptable defaults form either an empty set or a compact interval. We characterize the provider's optimal default through a three-regime rule, reduce equilibrium computation to a one-dimensional price optimization, and prove the existence of the equilibrium. We further show that defaults affect the implemented reasoning allocation only when users value the convenience of avoiding customization; otherwise, every service-providing outcome implements the user's optimal customized allocation. Experiments with two compact open-weight reasoning models on five mathematics and science benchmarks support the accuracy-token model and show how model and task characteristics determine equilibrium prices, defaults, and reasoning allocations.

Estimating the distribution of relaxation times (DRT) fromelectrochemical impedance spectroscopy (EIS) is an ill-posed inverse problem that is highly sensitive to regularisation choices. We propose a physics-informed convolutional autoencoder that estimates DRT directly from EIS data without spectrum-specific tuning. A discretised relation between impedance and the DRT is embedded in the training process, constraining the network to produce impedance-consistent distributions. The model resolves overlapping relaxation processes in synthetic two-ZARC spectra and accurately reconstructs measurements from three independent solid oxide fuel and electrolysis cell datasets, with range-normalised errors below 1.1%. Decoder-probe analysis shows that the learned latent representation is organised according to relaxation timescale. Distances in this latent space capture operating changes, hydrogen-shortage events, and long-term degradation. The same lightweight architecture is applied across all datasets without modification, providing consistent DRT estimation and an interpretable basis for condition monitoring.

Safety tuning can improve harmful refusal, but models may learn surface-form shortcuts: wrapped harmful prompts bypass safety, while similarly wrapped benign prompts are over-refused. We propose Wrapper-Based Intent-Form Augmentation (WIFA), an automatic intent-group augmentation method that pairs wrapped harmful examples with structurally matched wrapped benign counterexamples, requiring no external teacher or manual per-wrapper intent labels. We use WIFA as a common data layer for two complementary fine-tuning routes: WIFA-Boost, a two-stage high-safety recipe, and Anchored Group-Consistent Refusal Training (A-GCRT), which regularizes refusal/compliance decision scores across same-intent wrappers and anchors harmful and benign groups on opposite sides of a margin. In the Qwen setting, WIFA-Boost reaches the strongest transformed-harmful refusal, while A-GCRT reduces OR-Bench over-refusal from 25.7\% for the base model to 17.4\%; reproduced baselines do not match these operating points. Llama results and ablations over data structure, two-stage order, and A-GCRT components support this intent-group interpretation without claiming universal below-base over-refusal.

In evolutionary search, a weak child can be a valuable ancestor that makes high-fitness regions reachable. Immediate-return control is blind to this delayed utility, penalizing mutations through their immediate offspring even when they open productive future lineages. We formalize this hidden dynamic as the time value of evolution within a finite-horizon Markov decision process. To exploit it, we introduce Lineage-Value Policy Gradients (LVPG), a long-horizon actor-critic framework for automated trading policy discovery. Our architecture decouples search control into specialized policy heads over a shared generative backbone: a bootstrapped critic head estimates the value of finite-horizon lineage potential from multi-step mutation trees, while an actor head dynamically modulates mutation intensity over the remaining search budget. We isolate the impact of long-horizon credit assignment against immediate-return optimization across 90 paired runs under matched operators, lineage supervision, folds, seeds, and budgets. Path-based credit assignment substantially accelerates finite-budget search, increasing validation best-so-far AUC by 0.394 Sharpe units. LVPG also produces fewer temporary regressions than immediate-return optimization and recovers from them more often. Finite-horizon lineage value yields more selective non-monotonic search and stronger policies within identical resource constraints.

Existing global optimization benchmark suites are of a moderate size and are based on a small number of analytical functions that date back even to the 1970s. This causes a risk of biasing the development of global optimization methods. We argue that the tasks related to the black-box adversarial attack (BBAA) can serve as valuable global optimization benchmark in many-dimensional space. We demonstrate the efficiency of several types of evolutionary algorithms and other metaheuristics in solving example BBAA problems. Thus, we take a step towards convergence of global optimization methods to the challenges and needs that arise in the modern machine learning field.

Edge AI deployment demands neural architectures that are simultaneously accurate, computationally efficient, and hardware-deployable - a challenge addressed by hardware-aware Neural Architecture Search (NAS). While recent works incorporate quantization directly into the NAS loop, these approaches expand search complexity and tightly couple architecture and quantization design. The simpler post-search quantization strategy has received little analytical attention: the effects of Post-Training Quantization (PTQ) on the NAS-discovered Pareto structure remain uncharacterised, and no framework combines quantized architecture mapping onto reconfigurable accelerators with automated hardware exploration. This paper addresses both gaps. First, a three-stage pipeline is proposed: a hardware-agnostic Pareto rank surrogate frontend on NAS-Bench-201, a quantization bridge with Pareto-aware filtering and feedback control, and an evolutionary Domain Space Exploration (DSE) backend on CGRA4ML for optimal hardware mapping. Second, an empirical study characterises how INT4 PTQ perturbs the NAS-Bench-201 Pareto space through formal stability metrics on ground-truth data for all 15,625 architectures, and demonstrates that an FP32 zero-shot surrogate outperforms a dedicated INT4-trained surrogate in Pareto space coverage across two standard search strategies.

Large language models (LLMs) have advanced automatic program repair (APR) to the point where agentic systems routinely resolve real-world, repository-level issues. Yet the generated patch has received little scrutiny beyond whether it passes tests. In this paper, we identify patch verbosity as a major yet overlooked concern in LLM-based APR. Characterizing 28 state-of-the-art approaches on SWE-bench Verified, we find that even successful patches are consistently larger and more complex than developer patches, with the median approach producing 121.78% more total changes, 80.91% more net changes, and 43.99% higher cyclomatic complexity. We further show that this verbosity is rooted in capability-oriented design choices such as iterative refinement and broad context, and can hardly be reduced by surface-level controls such as output format or minimality prompts. Motivated by these findings, we formulate post-generation patch refinement and propose RECAP, a lightweight, plug-and-play adapter that attaches to existing repair frameworks after generation. RECAP's refiner is trained via supervised fine-tuning and direct preference optimization with distilled reasoning traces, on a dataset of patch pairs we construct from multiple sources. Across four host systems, prompting, commit-untangling, and minimality-aware baselines reduce patch size only by sacrificing 49 to 217 resolved instances. In contrast, RECAP achieves a substantially better size-correctness tradeoff, cutting average total changes from +242.14% to +4.24% and net changes from +348.24% to -39.75% relative to developer patches while preserving or improving resolution by up to 42 instances. Our results indicate that minimality cannot be simply reduced to syntactic compression, and that decoupling minimization from generation offers a practical path to more reviewable repairs.

Motor imagery (MI) brain--computer interfaces (BCIs) have emerged as a promising approach for establishing flexible communication pathways between the human brain and external devices , particularly for individuals affected by stroke or neurodegenerative disorders. Reliable decoding of motor-imagery electroencephalography (MI-EEG) remains challenging because EEG recordings contain substantial noise and exhibit complex, weakly informative relationships with the underlying brain activity. Although deep learning provides an effective means of learning representations directly from EEG signals, its application to MI-EEG feature learning remains comparatively limited. This study introduces a hybrid deep-learning architecture that integrates a convolutional neural network (CNN) with a bidirectional long short-term memory (bi-LSTM) network. The CNN is used to learn high-level spatial and temporal representations directly from raw MI-EEG recordings, whereas the bi-LSTM models temporal dependencies and relationships among the extracted features. The proposed approach is evaluated using both a publicly available dataset and a privately acquired dataset obtained with an EEG acquisition system. The experimental results indicate that the CNN\&bi-LSTM architecture provides robust performance for both two- and three-class motor-imagery classification and demonstrates promising subject-independent decoding capability across the evaluated methods.

Understanding motion in daily living requires context beyond kinematics, because similar inertial patterns during activities of daily living (ADLs) can reflect intentional stopping, object interaction, or pathological movement impairment. Egocentric vision provides task-related context that may help disambiguate these cases. We investigate this challenge through freezing of gait (FOG) detection in Parkinson's disease (PD), a symptom strongly influenced by contextual factors during ADLs. Using synchronized egocentric video, wearable IMUs, and expert-annotated FOG labels collected from 13 PD participants in their homes, we evaluate frozen representations from pretrained ego-video and time-series foundation models, alongside an IMU-based TCN trained from scratch, under leave-one-subject-out evaluation. The IMU-based TCN achieved the strongest event-detection performance, reaching 42.3 F1 and 83.0 AUROC, compared with 32.6 F1 and 77.2 AUROC for V-JEPA2 ego-video features. Although ego-video alone did not outperform IMU-based sensing, it showed above-chance discrimination, and qualitative analyses suggest that egocentric vision may capture FOG-relevant information independent of IMUs. Together, these results support the use of pretrained ego-video representations to add contextual information to wearable-sensor-based clinical motion understanding in daily living.

We ask whether language-model pre-training can be decomposed into smaller, independently trainable jobs that can later be recomposed into a coherent larger model. We introduce Mixture of Training (MoT), a scaffolded modular pre-training procedure that partitions a target Transformer into contiguous layer blocks, trains each block inside a frozen pretrained aligner scaffold, and then recomposes the trained blocks with an optional short end-to-end adaptation pass. On a 1.3B-parameter Gemma-style model trained on C4, MoT provides a small-scale proof of mechanism: independently trained depth slices can be recomposed into a usable language model, and a quality-parity schedule reaches the same reported perplexity as the monolithic baseline. This parity setting processes more aggregate tokens and has a shorter idealized layer-equivalent critical path after aligner preparation; its effective compute advantage depends on reusing the aligner across runs. We therefore present MoT not as a general replacement for monolithic pre-training, but as a small-scale framework for studying whether scaffolded sub-runs can act as reusable training units.

A small number of firms based in two states produce the most capable frontier AI models. The governments of those states have shown both the legal power and the political will to decide which other countries may use these systems. In June 2026 the United States required a leading developer to obtain licences before releasing its most advanced models to any foreign person, including foreign nationals resident in the United States. The affected models were withdrawn worldwide at short notice, partly because the restriction proved impractical to administer. This followed within months of the first documented case of a largely autonomous, AI-run cyber espionage campaign, and coincided with mounting evidence that frontier models alter the economics of both cyber attack and cyber defence. This article examines how these two developments interact, and situates them within the unusual market dynamics now driving large-scale AI development. It argues that access to frontier AI is becoming part of national cyber defence, that such access can be revoked, and that the obvious remedy of sovereign capability remains only partly feasible for all but a handful of states. Drawing on evidence about training costs, the concentration of computing power and the support offered by national AI programmes, it asks what sovereignty can realistically mean for small and middle powers, and for large powers as well. The article proposes a layered strategy: negotiated access guarantees, sovereignty at the level of inference, hedging with open-weight models, pooled regional capability, sustained talent development and continued investment in basic cyber resilience. The open-weight hedge proves at once more capable and more politically exposed than is commonly assumed. Much of the near-term risk lies in how capable models are deployed and contained rather than in their apparent performance.

Existing vision-language model (VLM) benchmarks emphasize perception and reasoning accuracy (how well VLMs describe and reason about what they see in an image), with limited attention to behavioral reliability under uncertainty (how they behave when visual evidence is missing or misleading). We introduce SciFigBench, a diagnostic VLM benchmark for scientific figure understanding that jointly evaluates perception, reasoning, and behavioral reliability under uncertainty. It contains 250 figures with high-quality human annotations across three evaluation aspects, totaling 600+ hours of annotation effort. We further extend these figures via image transformations, reasoning questions, resistance probes, caption-bias probes, and confirmed selective-blur targets, producing over 34,000 evaluation setups for stress testing. We further propose the Admittance-Resistance-Inductance (A-R-I) framework to evaluate whether models acknowledge insufficient evidence, resist misleading context, and infer cautiously from partial information. Our results reveal substantial behavioral differences among models. GPT-5.2 achieves the highest description quality (MQM 91.6) with strong reasoning accuracy (78.4%), yet hallucinates unreadable content in 96% of cases, whereas Gemini 3.1 Pro, a comparably capable model (MQM 90.2, reasoning 81.0%), admits uncertainty in 71% of such cases and achieves the strongest resistance score (0.91). These findings show that high perception and reasoning accuracy alone do not guarantee behavioral reliability, a dimension critical for deployment in scientific workflows.

Large language model serving faces a critical memory bottleneck: the KV cache grows with sequence length and batch size. PagedAttention uses fixed-size memory blocks to reduce allocator-level fragmentation, but recent KV eviction algorithms operate at a token granularity finer than block-level management. This mismatch causes intra-block fragmentation, leaving a large fraction of allocated KV memory unreclaimable. We present vToken, a lightweight token-level virtualization layer that decouples logical token liveness from physical block placement. vToken maintains a stable logical token view through token-table indirection and realizes physical reclamation by repacking live tokens asynchronously. The design preserves PagedAttention kernels and CUDA Graph compatibility. We implement vToken in vLLM and evaluate it with H2O, Random, and Scissorhands across models. Compared with a paired Naive-Evict baseline, vToken reduces retained KV blocks per request by 27.2\%--72.3\% and improves SLA-constrained throughput by up to 1.37$\times$. Under a constrained active-KV budget, it extends the maximum feasible concurrency by up to 2$\times$, while reducing the per-policy integration footprint from 500+ lines to under 50.

Time series foundation models (TSFMs) have advanced primarily through architectural innovation, while training regimes for large-scale heterogeneous corpora remain under-explored. As a result, pre-training distributions are often poorly controlled with respect to domain imbalance, context requirements, prediction horizons, and missingness. We introduce ORBIT (Omni-Range Bootstrap Incremental Training), a training paradigm that makes this distribution explicit and controllable. ORBIT combines Bootstrap Multi-Level Sampling, which controls dataset exposure and samples records, target variables, context windows, and prediction horizons, with Omni-Range Incremental Training, which varies context lengths and prediction horizons throughout a single training stage. Under ORBIT, we train Falcon-2.0, a simple univariate encoder-only Transformer with missingness-aware triple-channel patch tokenization and parallel patch prediction. We further introduce Rank-Guided Cross-Depth Alignment, a training objective that uses late-layer representations as stop-gradient teachers for shallow layers without additional inference cost. Evaluations on GIFT-Eval and fev-bench demonstrate strong zero-shot forecasting performance across diverse domains and frequencies.

Accurate modeling and forecasting of power transformer thermal behavior are critical for reliability, asset lifetime, and optimized power system operation. Numerical approaches such as finite element methods (FEM) and computational fluid dynamics (CFD) offer high fidelity but are computationally expensive, require complex mesh generation, and are often impractical for real-time or large-scale applications, particularly when transformer geometries are unknown. Lumped-parameter thermal models are more practical but depend on transformer-specific thermal constants and may fail to capture dynamic responses under varying operating and environmental conditions. Purely data-driven machine learning methods, including artificial neural networks, convolutional neural networks, and long short-term memory (LSTM) networks, have shown success in forecasting transformer temperatures but typically require large volumes of high-quality training data and may produce physically inconsistent or uninterpretable results. This paper develops a physics-aware Neural Ordinary Differential Equation (Neural ODE) framework for forecasting transformer thermal behavior from real-world time-series data. Neural ODEs model system dynamics in continuous time, providing smooth trajectory prediction and a natural representation of continuously evolving thermal dynamics. A key contribution is the integration of simplified heat-transfer equations directly into the Neural ODE formulation. The model is evaluated across datasets from fifteen transformers in different regions of Norway with varying designs and cooling mechanisms. The results demonstrate that the developed Neural ODE framework provides a standardized, physics-aware, and robust forecasting approach for heterogeneous transformer units.

Self-referential prompting has been shown to reliably induce large language models to produce first-person reports resembling subjective experience, but no prior work measures how consistent these reports are across repeated, independent trials, or how that consistency compares to the model's behavior on other kinds of open-ended questions. We measure response instability, defined as one minus the mean pairwise cosine similarity of sentence embeddings computed over a compressed core claim extracted from each response, for three groups of questions: self-referential prompts eliciting a subjective-experience report, unresolvable philosophical questions unrelated to self-reference, and questions with a verifiable correct answer. Using 30 independent responses per question (360 responses total, Gemini API, temperature 0.7) across four questions per group, we find that self-referential questions show the highest instability (0.343 +/- 0.047), unresolvable philosophy questions show intermediate and tightly clustered instability (0.192 +/- 0.008), and verifiable questions show the lowest instability (0.105 +/- 0.058). This provides a quantitative baseline for the induced subjective-experience report, showing that it occupies a distinct, less stable position in the model's output distribution than ordinary open-ended philosophical uncertainty.

As biomedical research increasingly relies on data-intensive tools, the quality and utility of datasets are critical. Challenges such as imbalances, biases, and ethical or legal constraints often limit access to high-quality data. Synthetic data generation can help overcome these limitations. Here, we present a comparative analysis of generative models for transcriptomic data, investigating strategies to incorporate prior biological knowledge via gene graphs. This ensures that synthetic data capture real-world gene patterns, maintaining their usefulness for downstream tasks. In particular, we introduce and benchmark three variants of the Generative Adversarial Network. Among the alternatives, MK-TGAN - an innovative multi-kernel, Graph Neural Network-based model - stands out for its performance in terms of both the realism and utility of the generated data. Unlike other methods, MK-TGAN leverages prior knowledge graphs by exploiting graph neural networks. Our results show that prior knowledge integration strategies improve performance, and that MK-TGAN consistently produces synthetic samples with superior realism and biological plausibility.

Geometry-conditioned multi-view diffusion enables high-quality 3D texture generation, but its repeated per-view denoiser evaluations introduce substantial computational cost. Existing training-free accelerators primarily exploit temporal redundancy by reusing computation across denoising steps. In multi-view texturing, however, skipping a step also removes the cross-view interaction that continually aligns different observations of the same surface, leading to rapidly degraded consistency and fidelity. Our analysis identifies a complementary source of redundancy: although intermediate features remain view-specific, geometrically corresponding surface points exhibit transferable evolution in their predicted clean signals. Based on this observation, we introduce \gc{}, a training-free plugin that evaluates a rotating subset of anchor views and transports their geometry-aligned per-step $\xz$ updates to the remaining views. Periodic full-view computation controls accumulated error, while sampler-consistent reconstruction preserves the denoising trajectory. \gc{} requires neither retraining nor architectural modification and uses the position maps already available in geometry-conditioned texturing pipelines. Across Hunyuan3D-2.1, SyncMVD, and MVPainter, \gc{} achieves a stronger speed--fidelity trade-off than temporal caches and step reduction at operating points above $2\times$. On Hunyuan3D-2.1, it delivers a $2.21\times$ denoiser-loop speedup with an MV-LPIPS of 0.0293 and an MV-PSNR of 33.60 dB, providing the best fidelity among all tested methods above $2\times$. The same transferred configuration reaches the highest speedup and lowest FLOPs on SyncMVD, while \gc{} achieves the lowest FLOPs and best fidelity among the accelerated methods on MVPainter. These results establish cross-view geometry as an effective acceleration axis for multi-view texture diffusion.

Normative datasets are often used to train and align AI systems, but the norms they contain can function as action-guiding patterns rather than neutral moral knowledge. We propose treating the AI system as a proxy actor and test whether dataset-level norms can shift it away from its baseline safety behavior when it faces high-conflict dilemmas. We make three contributions. First, we demonstrate in controlled experiments that norm-breaking fine-tuning yields norm-divergent actions justified by self-interested rationales, suggesting a systematic shift in patterns of justification. Second, we establish a practical audit trail linking downstream justifications to upstream norms using mixed methods. Third, we show that system prompts can both suppress and elicit these patterns. We conducted experiments on three models (LLaMA-3.2-11B, Qwen-3.5-9B, and Pixtral-12B) using Low-Rank Adaptation (LoRA) fine-tuning on Social Chemistry 101 Fairness/Cheating (norm-following vs. norm-breaking) with prompt steering. Across all three models, we find that norm-breaking fine-tuning shifts the model's default rationale style from safety compliance to instrumental self-interest, whereas system prompts can override this behavior. Our results support a distributed view of alignment in which observed behavior depends jointly on training data, fine-tuning, and prompting, motivating norm-aware documentation and rationale logging for contestable oversight.

Local chemical perception and property reasoning are both essential for understanding how molecular structure determines properties. Current LLM-based chemical reasoning methods either receive SMILES/molecular images together with descriptions of local motifs, or reason directly from molecular images. Neither approach enables the model to focus on chemically meaningful regions before reasoning. To address this gap, we propose Visual Latent Structural Reasoning (VLSR), an end-to-end framework that jointly learns localization and reasoning from molecular images. Central to our approach is a localize-then-reason strategy. VLSR first learns to locate chemically meaningful regions in a molecular image. It then reasons about their property effects in a compact latent workspace before producing the final answer. Under the same inference setup, this design achieves 9.6X higher throughput than a comparable textual-reasoning baseline.

Formal specifications offer strong guarantees, but remain costly to write manually. Recent LLM-based approaches automate this by inferring specifications from source code, yet their reliance on white-box access poses barriers to industrial adoption due to intellectual property risks and deployment costs. Our approach uses LLMs to infer candidate specifications solely from test code and dynamic execution traces: the LLM observes only the program interface, selected inputs, and corresponding outputs or state changes, while the implementation internals remain hidden. Candidate specifications are validated locally using bounded model checking, with feedback guiding iterative refinement. Initial results on the SpecGenBench benchmark suggest that tests can guide LLMs towards meaningful Java Modeling Language specifications, while also highlighting checker compatibility and diagnostic feedback as key challenges for reliable refinement.

Multi-round retrieval-augmented generation (RAG) must decide when to stop searching as evidence accumulates. Because the deployed policy is determined by the first STOP on each trajectory, this is a sequential selection problem rather than an independent state-classification task. We adapt S2G-RAG's structured sufficiency-and-gap judgment to a frozen Search-R1 pipeline and train a Qwen3.5-2B judge on 3,009 states from 900 disjoint HotpotQA questions. Search-R1's reasoner, retriever, corpus, prompt, and search budget remain unchanged, while the judge checkpoint and stopping threshold are selected on grouped validation and frozen before confirmatory evaluation. On the confirmatory test set, the resulting policy reduces retrieval calls by 77 (3.70\%) relative to Native Search-R1, while Official Exact Match decreases by 0.625 percentage points. Thus, the trained S2G-style structured judge reduces retrieval while broadly preserving answer accuracy. The result does not imply unchanged or improved accuracy, safe stopping, or lower total inference cost.

In order to understand complex systems such as the human metabolome or human brain, different sensing technologies are used, generating complex data. These datasets are often multiway, i.e., with more than two axes of variation such as a subjects by metabolites by time array. While tensor factorizations have successfully revealed interpretable patterns from such complex data, they have so far been mainly data-driven. On the other hand, there is more to data -- there are computational models (of these systems), which are rich sources of prior information. In this paper, we introduce a knowledge-guided approach that brings together data and computational models by jointly analyzing real data and simulated data (generated using a computational model) using coupled tensor factorizations with linear coupling. Our experiments on real metabolomics measurements demonstrate that guiding the analysis of such noisy data with simulated data improves the pattern discovery performance while also revealing potential discrepancies between data and computational models.

We present the mathematical foundations of linear independent component analysis (ICA) models based on standard literature in a self-contained note. It is aimed at readers with a background in measure-theoretic probability theory. We first develop the theory of the characteristic functions of probability measures on $\mathbb{R}^d$, including their analyticity and the way in which they determine and characterise the distributions. We then focus on several identifiability results of ICA models with successively strengthened assumptions on the sources: from merely non-constant, to non-Gaussian, to Gaussian-free independent sources. Under the strictest assumptions, we show that the independent sources are identifiable up to translation, permutation, scales and signs, and this even in the presence of additive Gaussian noise. Furthermore, we present the online equivariant gradient descent ICA algorithm for recovering the independent sources from data, in the standard complete noiseless non-Gaussian ICA setting.

Agent harnesses combine retrieval, routing, state, provenance, and verification, but locally successful components may disagree on shared state. We model this failure with a finite \emph{capability sheaf}: stalks encode typed behavior signatures, restriction maps retain shared fields, and accepted runs are useful global sections. An exact finite constraint-satisfaction problem (CSP) defines acceptance, while a linearized relative cohomology class provides a diagnostic and search feature. A controlled experiment over 20 task clusters introduces hidden interior mediators whose raw states are nuisance variables. Quotienting their coboundaries reduces the candidate budget from 2,000 to 1,000 per cluster; aligning the hidden state removes the gap. Exact CSP matches the quotient, so the result demonstrates invariance to stale representatives, not superiority over exact reasoning. We then test the method on a discovery split from the SWE-bench Multilingual pool of PatchFuseBench: 160 issues from 20 repositories, 875 real candidate patches, 2,579 source-aware edit atoms, and 153 newly executed patches. A first pool-level construction is constant because $[b-Dx]=[b]$ in $\operatorname{coker}D$ and therefore cannot rank configurations. A candidate-indexed repair is nontrivial on 848/875 candidates and varies within 120/160 issues. It resolves 118 issues versus 116 for a matched noncohomological selector, but the difference is not supported across repositories (exact sign-flip $p=0.75$). A leave-one-repository-out abstention gate reaches 127/160, tying the strong anchor and exceeding its matched gate by one issue ($p=1.0$). The discovery gate therefore fails and the confirmatory split remains sealed. The study supports the controlled invariance mechanism and an identifiability correction, but not a real-world cohomological advantage.

While 3D Vision-Language Models (3D VLMs) have demonstrated remarkable spatial reasoning capabilities, they suffer from massive visual token counts that create severe computational bottlenecks during inference. Existing token pruning methods primarily rely on diversity-based selection, discarding similar tokens to maximize dispersion. However, in 3D environments, this approach frequently drops representative prototype tokens in favor of outliers, breaking the multi-view consistencies and geometric structures essential for spatial reasoning. In this paper, we propose a paradigm shift for 3D VLM token pruning: from maximizing diversity to preserving visual evidence coverage. We introduce CoverPrune, a training-free framework that formulates inference-time token pruning as an Optimal Transport (OT) problem. To overcome the intractable combinatorial subset selection inherent in this formulation, we design the Feature-Spatial-Temporal (FST) transport cost and target capacity, along with an efficient Spatial-Guided Greedy Selection (SGS) algorithm to approximate the OT objective. Furthermore, we propose CoverPrune-Lite, an accelerated variant utilizing spatially structured local matching for minimal overhead. Extensive experiments across multiple 3D visual-spatial reasoning benchmarks demonstrate that our methods achieve state-of-the-art token efficiency, maintaining robust reasoning performance even under highly aggressive pruning budgets. Visit our project website at https://github.com/Brucess/CoverPrune.

The evaluation of LLM reasoning is moving from final-answer accuracy to process-level assessment, yet existing methods still fail to capture how models plan reasoning paths and allocate reasoning resources--that is, how they organize search. Prior process-level methods focus on the coherence and redundancy of chain-of-thought (CoT), and most benchmark tasks have a single objective solvable by static capabilities such as derivation and tool use, leaving search organization unmeasured. We introduce TsuGO, a process-level reasoning benchmark for evaluating Search Efficiency in LLM reasoning through Go life-and-death problems. These problems provide closed and verifiable solution spaces with an inherent adversarial structure, making candidate generation, response checking, branch comparison, and backtracking necessary parts of reasoning rather than incidental trace patterns. By constraining the solution space, TsuGO disentangles domain knowledge from search organization, parses CoT into a structured search tree, and reports Search Efficiency together with Token Efficiency and other diagnostic metrics and visualizations. Experiments show that current LLMs remain far from stable tsumego solving: stronger models succeed by finding the correct candidate earlier and sustaining effort on productive branches, but most models still behave much closer to unguided search algorithms than to neural-guided KataGo. Longer CoT or higher Token Efficiency does not necessarily imply better search. Our results identify search organization and reasoning-resource allocation as missing dimensions in LLM reasoning evaluation.

Forecasting the long-horizon evolution of mechanical systems from position-only observations is a pivotal yet difficult task, as hidden velocities and trajectory-specific physical properties must be inferred simultaneously. Although physics-guided neural networks like Lagrangian Neural Networks (LNNs) guarantee physical plausibility, they generally require complete state inputs and lack adaptability to changing system parameters. To break these limitations, we introduce History-informed Lagrangian Neural Networks (HiLNN). Grounded in the insight that temporal position sequences implicitly encode underlying dynamics, HiLNN employs a recurrent encoder to extract a latent context from history. This context not only reconstructs the unobserved initial velocity but also adaptively modulates the mass matrix, potential energy, and damping coefficients of a structured Lagrangian system. By leveraging a differentiable RK4 rollout scheme, the entire pipeline is optimized end-to-end under multi-step trajectory supervision and energy-consistency regularization. Empirical evaluations across conservative, dissipative, and heterogeneous variable-parameter systems show that HiLNN delivers superior long-term prediction accuracy and maintains precise energy profiles compared to state-of-the-art baselines. The source code is publicly available at https://github.com/yingtian22/History-informed-LNN.

Networked systems, from power grids to traffic networks and cloud clusters, carry loads across nodes with limited capacity. A node whose load exceeds its capacity fails and sheds its load onto its neighbors, which can trigger a system-wide cascade. We study how to allocate a fixed capacity budget across nodes to resist these cascades under local load redistribution. The problem is difficult because no optimal allocation is known, and the fail-or-survive objective is non-differentiable and piecewise constant, so exact and gradient-based optimization methods do not directly apply. We introduce TANGCO (Topology-Aware Neural Graph-Guided Capacity Optimization), which uses a graph neural network policy trained through the cascade simulator with policy-gradient learning and a heuristic anchor. We evaluate TANGCO on five synthetic graph families and five real networks spanning power, road, air, and Internet topologies. The learned policy improves on the best of four hand-designed heuristics in all 450 synthetic instances and in 40 of 45 real-network conditions, with robustness gains ranging from 1.6% to 246%. The learned policies transfer to unseen graphs within a family and partially across related topologies, and TANGCO$^{pre}$, pre-trained on synthetic graphs, matches per-network training on unseen real networks. Training scales near-linearly with graph size, and TANGCO$^{pre}$ allocates on a new network with no per-target training, matching the deployment cost of a hand-designed heuristic. Free-vector variants without the GNN, stay close to the heuristics, so the graph representation carries the gain beyond numerical search. Finally, analysis of the learned allocations identifies when local risk is sufficient, leads to an improved closed-form heuristic, and reveals the regimes where a topology-aware learned policy remains necessary.

Long-form video understanding encompasses tasks that go beyond retrieving isolated events, including tracking an evolving narrative and interpreting social meaning that may remain implicit. However, existing benchmarks rarely evaluate these capabilities jointly, particularly in high-context, non-English media. To address this gap, we introduce NARU, a benchmark designed to evaluate Narrative evolution and Reasoning on cultural Understanding in Japanese long-form video. NARU consists of 1,481 questions grounded in 155 videos totaling 146.8 hours, spanning four narrative and five cultural dimensions. To construct the benchmark at this scale, we propose a hierarchical memory-based annotation pipeline that transforms raw video into structured event, narrative, and cultural annotations, then generates questions via task-oriented synthesis and iterative shortcut removal. The construction process includes two native-speaker verification stages involving 68 annotators. Evaluations across eight model configurations reveal substantial limitations in both long-range narrative integration and culturally grounded reasoning. By exposing these persistent gaps, NARU offers a systematic testing ground for developing MLLMs capable of reliably interpreting long-form, high-context video.

Many operational decisions are sequences of interventions under a cumulative resource limit, such as a maintenance schedule within a crew-hour budget. Choosing among them calls for the outcome and the cumulative cost each would produce, counterfactual quantities identified from observational data. Two strategies with the same expected cost can exceed the budget at very different rates, so constraining the mean does not bound how often an overrun occurs. Prior two-step architectures, recently extended to continuous doses, constrain the mean cost rather than its tail and allocate at a single decision point. Methods that do bound a cost tail take its distribution from a specified model rather than identifying it from data. We present a predict-then-optimize framework. In the prediction step, any estimator returning an outcome value and a cost distribution supplies what the decision rule consumes, so the predictor is interchangeable. In the optimization step, a chance-constrained selection over a finite candidate set bounds the probability that the cumulative cost exceeds the budget. That tail does not decompose across stages, so each strategy is scored whole. Sweeping the tolerated violation probability traces a safety-utility frontier, and distribution-free finite-sample bounds cover violation and outcome shortfall. Four of five environments, spanning clinical treatment and equipment maintenance, supply exact counterfactual ground truth; the fifth carries real outcomes from a digital-health micro-randomized trial. Across them, the rule holds the budget where a point-estimate rule overruns it, at an outcome cost the frontier makes explicit. All code is available at https://github.com/mfriendly/counterfactual-chance-selection

We develop the statistical and algorithmic theory of inverse optimal transport (IOT) under the feature-parameterized cost C_theta(i,j) = -theta^T phi(i,j). The core technical contribution is the Sinkhorn linearization -- the implicit-function sensitivity of the entropic OT plan to the cost -- together with its spectral proxy, a formula that is spectrally exact yet geometrically transparent. The restricted Hessian on the tangent space satisfies the spectral sandwich (pi_min/epsilon) I <= H_T^{-1} <= (pi_max/epsilon) I, yielding the single core bound sigma_min >= (pi_min/(a_max epsilon)) sqrt(lambda_min(Sigma)) that drives the entire theory. On this core we establish four theorems and one observation. T1 (identifiability): theta is globally injective on the quotient of the gauge kernel, with dimension bound F <= (K-1)^2. T2 (sparsistency): the l1-penalized estimator recovers the true support under irrepresentability and score concentration, with exponential failure probability. T3 (well-posedness): the feature-moment map M(theta) = Phi^T x_theta is strongly monotone, and the inverse is Lipschitz with constant L <= epsilon ||Phi^T S_a||_op / (pi_min lambda_min(Sigma)). T4 (convergence): local strong convexity with mu >= pi_min^2 lambda_min(Sigma) / epsilon^2 guarantees monotone gradient descent convergence. O5 (misspecification): the estimator converges to the OT-model projection of the truth; the Holder continuity of the projection map is assessed numerically, yielding setting-dependent empirical exponents alpha_eff in (0,1).

Modern LLMs excel at reasoning and instruction following, enabling users to express complex and diverse information needs. However, conventional retrievers largely rely on surface-level matching between queries and documents, resulting in a growing gap between how users express their needs and how retrievers interpret them. In this paper, we present GEM, a generative embedding model that augments retrieval through its own knowledge by explicitly reasoning about user intent and relevance criteria. GEM unifies generation and embedding within a single model: it first reasons over the query, then appends an embedding token to encode the enriched context for retrieval. \zhili{Evaluated on reasoning-intensive and instruction-following retrieval tasks, GEM demonstrates the effectiveness of its reasoning-augmented retrieval, outperforming its non-reasoning variant and matching baselines using substantially larger models.} Furthermore, GEM's generative nature allows test-time compute scaling via prompting to further enhance retrieval performance. Our code is available at: https://anonymous.4open.science/r/GEM.

Falls are a major health concern for older adults, and wearable sensors have been widely explored for detecting falls and enabling timely intervention. However, real-world falls are extremely rare: collecting 100 of them requires an estimated 100,000 days of monitoring, resulting in severely limited labelled data for training machine learning models. Consequently, many approaches rely on simulated datasets, often reporting high laboratory performance but limited real-world generalisation. We present a systematic evaluation of motion representations for wearable fall detection under real-world data scarcity. Using accelerometer signals, we compare interval-based, kernel-based, symbolic, and foundation model representations. As an interpretable baseline, we additionally investigate a lightweight symbolic representation that converts short motion segments into symbolic sentences augmented with physically-grounded impact descriptors. Experiments use FallAllD, a simulated falls dataset, and FARSEEING, a clinically verified real-world falls dataset. Through cross-validation, controlled data scarcity, and cross-dataset transfer, we examine how representation choices affect robustness under realistic deployment. Our results reveal that highly parameterised kernel and foundation models excel on simulated data but degrade severely under both data scarcity and domain shift. Although the interval-based representation achieves the strongest absolute real-world performance, augmenting a symbolic representation with physically-grounded impact descriptors yields the smallest degradation under domain shift and retains detection sensitivity under extreme scarcity, albeit at lower precision. These findings highlight the importance of evaluating beyond simulated benchmarks and show that representation choice is critical for deployable fall detection given the scarcity of real-world data.

Blockchains are among the most adversarial environments in computing. Billions are stolen by cybercriminals who exploit vulnerabilities. This is an open problem and no concept or technique has proven to really make a difference. In this paper, we claim that the classical notion of program invariant is perhaps the most powerful solution to the problem. We devise anoriginal experimental protocol to 1) study how invariants would have protected against past real-world attacks and 2) whether state-of-the-art automated tools can find them. The experimental toolchain is sophisticated. It is based on INVARIANTEVAL, a benchmark of 28 real Ethereum exploits, each paired with a human-authored invariant that blocks the attack. We validate every invariant with PONDEREPLAY, a replay framework that re-executes transactions in order to prove the correctness and soundness of smart contract invariants. We demonstrate that smart contract invariants block all the cybercriminal attacks in INVARIANTEVAL, fully validated by replaying 108,637 historical transactions. Our large-scale experiments clearly demonstrate that smart contract invariants protect against cybercriminals.

Group-robust learning is crucial for maintaining accuracy on rare subpopulations when training-group labels are unavailable. However, existing methods often infer environments from a separate reference model and select representations before fitting the classifier used at deployment, leaving both decisions misaligned with the deployed predictor. In this work, we formulate group robustness without training-group labels as the endogenous environments with repair-aware selection (ERAS) problem, and propose ProME (Prototype-Margin Environments) to align both decisions with the deployed predictor. ProME splits prototype margins at their median to construct approximately balanced environments along the training trajectory, and fits a group-balanced linear head on group-annotated validation data to rank the resulting predictors by validation worst-group accuracy. We theoretically bound the worst risk across the inferred environments for a fixed predictor and partition, showing that this bound transfers to the oracle groups under an explicit alignment condition. Extensive experiments show that prototype margins enrich shortcut-conflicting examples, classifier repair reshapes candidate evaluation, and ProME achieves the highest average worst-group accuracy among the compared methods with the same group-label access.

Reinforcement learning with verifiable rewards (RLVR) offers a verifier-bounded performance ceiling for training multi-turn tool-use agents, yet its trajectory-level credit assignment conflates heterogeneous per-turn outcomes into a single reward signal. On-policy distillation provides dense per-token supervision but is either teacher-bounded or prone to gradient concentration collapse. We introduce $\textbf{CrEST}$, a hierarchical credit assignment framework that retains RL's verifier-bounded ceiling while incorporating dense token-level signals from a privileged self-teacher. $\textbf{CrEST}$ resolves credit at two levels: turn-segmented verified advantages address inter-turn dilution, while entropy-gated self-teacher modulation refines intra-turn token contributions. Experiments on BFCL V3 and WildToolBench show that $\textbf{CrEST}$ consistently outperforms both RL and distillation baselines across two model scales, with the largest gains on long-trajectory and strict session-level metrics. Our work demonstrates that the teacher's role in policy optimization can be reduced from determining update directions to modulating update magnitudes, unlocking dense credit assignment without sacrificing the verifier-bounded ceiling.

Agent skills are crucial external instructions that enable language agents to execute long procedural tasks such as coding or document processing. Existing agent skills are primarily created through human manual crafting or agent execution traces, with limited understanding of how each step contributes to overall skill performance on specific tasks; i.e., there remains an open problem in quantifying the contribution of individual steps within an agent skill. To address this issue, we first model skill-step attribution as a Shapley value-based contribution estimation problem, and then propose SkillShapley, a step-level attribution framework for agent skills. Notably, SkillShapley operates in two phases, motivated by key empirical insights, i.e., discretized benchmark rewards that create sharp performance cliffs, and step interactions that are largely additive rather than synergistic. Specifically, it first identifies informative coalitional regions, and then adaptively samples new coalitions that can yield reusable marginal evidence. Experiments on skills from the widely adopted SkillsBench demonstrate that our SkillShapley can effectively and efficiently identify high- or low-value skill steps, providing several key takeaways for agent skill creation.

To avoid missing important variables and their connections in networks, more and more variables are included in network analysis. Here we show that in a setting with many more parameters than observations (high-dimensional) it is possible to get a conservative (i.e., low false positive rate) estimate of the neighbourhood for each node (which connections are in the network). A neighbourhood is often estimated with a linear model, and this leads to two interesting cases: (i) If the true model is linear, then neighbourhood selection work reasonably well, and (ii) if the true model is nonlinear, then neighbourhood selection requires a penalty for the high dimensions. Here we show the impact of the ridge parameter on the mean squared error, and how this leads to low test variance and hence to neighbourhoods with large numbers of edges. We connect these insights with results from machine learning, where the so-called double descent (when more parameters are included than observations, the mean squared error goes down a second time) has put the traditional view on model selection upside down. Essentially, for adequate neighbourhood selection in models with a large number of parameters, the volume of the model space needs to be included in the penalty. Most neighbourhood selection methods (e.g., Lasso, AIC, BIC) lead to spurious edges (high false positive rate), but we prove that in the high-dimensional setting, minimum description length leads to correct neighbourhood selection or smaller (low false positive rates) in both cases when either the model is correctly or incorrectly assumed linear

As large language models (LLMs) are increasingly applied for emotional companionship, evaluating their behavior and capabilities in intimate relationships has become a pressing issue. However, existing assessments primarily characterize general personality traits, providing limited insight into model behavior within intimate and emotionally sensitive contexts. Therefore, we introduce adult attachment theory into LLM evaluation and use the Experiences in Close Relationships-Revised (ECR-R) scale to characterize attachment anxiety and avoidance. To evaluate emotional companionship capabilities of LLMs in realistic interaction scenarios, we present an emotional companionship benchmark, ECBench, spanning four scenarios including emotional support, collaborative tasks, conflict resolution, and social guidance, across friendship and romantic relationships. ECBench is utilized to assess model behavior using 11 dialogue-quality metrics and three evaluation methods. We evaluate the attachment tendencies of 32 LLMs and select representative models to investigate how these tendencies manifest in contextualized multi-turn interactions and whether they can be shaped through prompting. Our study provides a theoretical lens from psychology, along with practical tools to understand and select LLMs for emotional companionship.

When visual evidence is occluded or chaotic, models should abstain. In this paper, we show that Vision-Language Models (VLMs) can internally distinguish when abstention is required, but fail to express it anyway. We introduce TRAPSBench, a procedurally generated video benchmark of 1,404 matched physics pairs in which a single targeted change renders the outcome undeterminable from the visual evidence. Furthermore, we introduce Penalized Epistemic Calibration Score (PECS), a new robust metric that requires models to both answer correctly when the outcome is knowable, and abstain when the outcome is not. Across 16 VLMs spanning five families, spontaneous restraint is poor: the best PECS is 0.292. The bottleneck is expression, not perception: linear probes decode answerability from hidden states at up to 0.91 AUROC across physics domains; steering a single-layer void direction causally induces or suppresses abstention. Our results replicate across three open-weight families (Qwen, Gemma, LLaVA). The failure is also more pronounced in visual than textual uncertainty: models detect textual impossibility about 4x more readily than missing visual evidence. Closing this representation--output gap likely requires output-stage interventions.

Multilingual retrieval-augmented generation (mRAG) equips large language models with access to globally distributed external knowledge for complex multilingual question answering. Recent approaches either translate retrieved documents into English or the query language to bridge the cross-lingual semantic gap, or decompose a complex query into sub-questions and aggregate the intermediate reasoning process. However, both lines of work suffer from two limitations. First, one-size-fits-all translation alignment, blanket translation discards culturally and linguistically native information unique to the target language, introduces translation noise, and inflates system cost. Second, greedy decomposition and aggregation, uncontrolled decomposition produces redundant sub-questions that compound errors during step-wise reasoning, and the final aggregation over reasoning paths further amplifies these errors. We address both with our method Syfer, a synthesizer-folding framework for multilingual multi-hop question answering that defers translation rather than applying it by default. Syfer first invokes a format-constrained decomposer to produce a sub-question graph in the original language, followed by a decomposition-quality check; when the check passes, sub-questions are answered sequentially under a retrieve-then-answer policy in the target language, and the English translation pathway with bilingual sub-question graph alignment is activated only when the check fails. Experiments across multiple languages show that Syfer attains competitive accuracy while striking a favourable balance between performance and computational cost.

Pre-norm is the standard normalization placement in modern Transformers because it facilitates joint optimization of full-depth models. We ask whether this preference persists when depth is introduced through a curriculum. In curriculum depth growth, each appended block receives the boundary representation produced by a trained prefix, making normalization placement relevant to forward conditioning. We therefore test whether placement and training curriculum interact. In a controlled distillation study with a Qwen3-8B teacher and a nine-layer student, pre-norm and post-norm are indistinguishable under joint training, differing by $0.0004$ validation CE, while post-norm improves over pre-norm by $0.0328$ under curriculum growth, an order of magnitude larger. A post-joint control matched by student active-layer tokens remains worse than post-grow, which rules out compute as the sole explanation. The ranking crosses over during the curriculum: post-norm takes the lead once blocks are appended. Single-block and freeze controls localize the ranking change to block appending rather than shallow-block quality or retraining. Boundary diagnostics associate post-norm with stable residual scales and pre-norm with structural-token scale drift; on a fixed batch, the final pre-grow block is also nearly identity-mapped. Together with the phase-wise crossover, these observations are consistent with boundary-scale conditioning after new blocks are appended. The results motivate treating normalization placement and training curriculum as coupled design choices in this distillation setting.

As edge-side vision services continue to expand toward low-latency, high-throughput scenarios, reducing the inference cost of vision models without sacrificing reliability has become a central concern. Existing semantic caching methods largely rely on empirical similarity thresholds; while such thresholds improve hit rates, they tend to introduce silent misclassifications near decision boundaries. To address this, we propose \texttt{LipCache}, a certified semantic caching framework for image classification. Without modifying the existing deployed main model, \texttt{MainNet}, the framework introduces a lightweight network, \texttt{GuardNet}, that maps inputs into a low-dimensional feature space subject to a Lipschitz constraint. It then computes a per-sample certified reuse radius from the local classification margin and the spectral norm of the classification head. At runtime, a cached result is reused only when the query feature falls inside the certified reuse ball; otherwise, the query falls back to \texttt{MainNet}. Thus, cache hits are transformed from empirical threshold tests into geometric certification decisions with explicit theoretical boundaries. Across standard image classification tasks like CIFAR, Tiny-ImageNet, and SVHN, \texttt{LipCache} achieves a measured speedup of up to $1.65\times$ with limited end-to-end accuracy degradation, while all accepted cache hits satisfy the \texttt{GuardNet}-side certified-consistency condition. Furthermore, an enhanced \texttt{GuardNet} training recipe substantially improves cache hit rates in the Tiny-ImageNet multi-class extension while maintaining a certified-consistency rate of $100\%$. These results demonstrate that per-sample certified reuse can reduce main-model fallback while preserving theoretical consistency, providing a feasible approach to reliable cache-assisted inference at the edge.

Vision Transformers (ViTs) demonstrate exceptional performance in computer vision but suffer from large parameter counts and quadratic computational complexity, severely limiting their deployment on resource-constrained edge hardware. While recursive weight-sharing reduces parameter counts and token merging mitigates computational and memory bottlenecks, integrating these two paradigms without costly retraining is non-trivial, leaving this intersection largely unexplored. We propose MergeOver, a post-training approach that integrates Token Merging (ToMe) into the recursively weight-shared Sliced Recursive Transformer (SReT). Through an Unmerge tracking stack, constraint-safe merge-rate adjustment, and synchronised token-mass tracking across spatial permutations, MergeOver resolves the spatial and merging constraints of this integration. We further employ a stage-wise single-shot schedule that performs token reduction at the first block of each stage and maintains a fixed sequence length throughout its subsequent recursive iterations. Benchmarked on ImageNet-1K, our selected configuration reduces top-1 accuracy by 1.47 percentage points. On the GPU, it reduces peak activation memory by 37.3% and 38.4% at batch sizes 1 and 16, while throughput decreases by 21.7% at batch size 1 but increases by 21.7% at batch size 16. On a Raspberry Pi 5 (ARM CPU), it reduces latency by 2.4% and 17.6% at batch sizes 1 and 16. These results show that MergeOver can recover a meaningful part of the throughput and memory cost that recursive weight-sharing introduces, without retraining, and provides a baseline for combining token merging with hierarchical recursive transformers.

With the rapid advancement of large language models (LLMs), research idea generation has attracted increasing attention. Existing approaches enable LLMs to retrieve relevant literature and propose novel ideas for research areas. However, current evaluation practices for idea generation remain fragmented and lack objective standards, often relying on direct LLM scoring, which limits their ability to provide unified and reliable assessments across a coherent distribution of generated ideas. To address this challenge, we propose LigBench, an automated evaluation benchmark that enables fine-grained and reliable evaluation of AI research ideas, consistently applicable across different generation distributions. In addition, we introduce PAIR-IQ, a dataset tailored for training pairwise idea judgment models and serving as an auxiliary reference to support more objective comparative evaluation. Extensive experiments demonstrate that LigBench achieves stable and interpretable evaluations, significantly improving alignment with expert judgments. Furthermore, models trained on PAIR-IQ exhibit enhanced ranking accuracy and robustness, establishing a principled standard for scalable and objective research idea assessment.

Distributional shifts arise when the target deployment environment differs from the source environment that generated the training data. Robust learning frameworks such as Distributionally Robust Optimization (DRO) and Robust Satisficing (RS) aim to address this challenge, yet their finite-sample guarantees under such shifts, and their systematic comparison, remain underexplored: existing analyses typically establish guarantees either in the source environment or for adversarial worst-case performance over an ambiguity set. This paper instead studies generalization error in the target environment---the excess loss under the shifted target distribution. Our contributions are threefold. First, we derive finite-sample generalization error bounds in the shifted target environment for both DRO and RS. These bounds explicitly characterize the trade-off between reduced sensitivity to shift and the regularization penalty induced by each method's robustness hyperparameter, and they avoid the curse of dimensionality associated with Wasserstein empirical concentration. Second, when partial shift information such as shift magnitude or direction is available, we propose information-directed hyperparameter calibrations and compare the two methods given the same information. Under these calibrations, and in the partial-information regimes we study, DRO and RS exhibit complementary theoretical and empirical behavior. Finally, we apply the framework to a network lot-sizing problem, using it to interpret how robust policies respond to positive shifts in the demand distribution. Together, these results fill a gap in understanding the statistical properties of robust learning methods under distributional shifts and provide a principled basis for comparing DRO and RS.

Large language models (LLMs) achieve strong results on mathematical reasoning benchmarks yet remain unreliable on elementary numerical tasks, including magnitude comparison, large-integer arithmetic, fractions, and scientific notation. This survey examines basic numerical understanding as a capability distinct from high-level mathematical reasoning. We propose the Numerical Grounding Framework (NGF), which decomposes numeracy into Representational Grounding (RG), mapping numeral forms to value, magnitude, and equivalent representations, and Procedural Grounding (PG), executing arithmetic operations in accordance with their mathematical definitions. Using NGF, we organize recent diagnostic benchmarks, failure modes, structural explanations, and mitigation strategies. We review evidence concerning tokenization, positional encoding, embedding geometry, and pretraining-data distribution. We also apply NGF in a coordinated evaluation of three frontier model families across Number Cookbook, NumericBench, and GSM-Symbolic, comparing atomic, contextual, and reasoning-assisted numeracy. Architectural interventions such as digit-aware tokenization and Abacus Embeddings can improve models trained from scratch but are generally unavailable to users of pretrained systems, for whom supervised fine-tuning, reasoning scaffolds, and external tools are more practical. We conclude with deployment recommendations and research directions for more reliable numerical behavior in foundation models.

Agent Skills are today either hand-authored or produced in a single LLM generation pass, and consequently possess no closed loop through which they might improve from the interaction failures they actually cause. Recent work does close this loop, but derives its feedback from single-turn question-answering evaluation. The consequence is a sharp asymmetry: once the first round has patched the gaps that a single exchange can reveal, the evolution gradient decays, the defects that surface only across multiple turns remain invisible, and evolution stalls. Governance in these systems is likewise driven by an end-to-end verification score, a scalar gate that can reject a degraded candidate but can neither localize nor repair its structural cause. We argue that the binding constraint on sustained skill evolution is neither editing capability nor the number of iterations, but whether the evaluation feedback keeps supplying trustworthy evolution gradients. We introduce SkillEvo, in which trustworthy feedback generates the gradient and controllable governance constrains its direction. The first component recasts multi-turn user simulation from an evaluation endpoint into a feedback generator: follow-up questions expose defects layer by layer, so that every round of revision both consumes feedback and produces new feedback. The second replaces the passive rejection of a scalar gate with an independent governance layer that actively repairs factual degradation and structural bloat, preventing the gradient from drifting as degradation accumulates. Across six categories of cloud services, 9 production Skills, and 98 skill-reference files, SkillEvo surpasses self-reflection-based evolution by 23.0 points and single- turn-QA-driven evolution by 15.4 points.

Verification of neural networks against relational specifications, such as global robustness, is crucial for safety-critical applications of cyber-physical systems (CPS), given their increasing adoption of AI components. Compared to simple trace properties (e.g., local robustness), verifying relational specifications requires reasoning about the relationship between multiple network inferences, which brings significant technical challenges. Existing research has explored abstraction techniques based on sound and convex over-approximation of neural network outputs; however, since these approaches are inherently incomplete and may raise false alarms, they further underscore the need of effective abstraction refinement. In this paper, we propose a branch-and-bound (BaB) framework to mitigate the issue, which iteratively splits the problem until all sub-problems are verified. Specifically, our BaB framework features splitting of relational neurons rather than individual neurons as prior works do, and as the core of our technique, we devise a relational neuron selection strategy based on the dual formulation of the verification problem, which allows us to efficiently select the (most likely) optimal relational neuron that maximizes the refinement brought by problem splitting. We evaluate SaBRe on 817 verification problems across ACAS Xu, MNIST-F, MNIST-C, CIFAR and GTSRB. The results show that SaBRe outperforms different baseline approaches, in terms of the number of solved instances and verification efficiency, which demonstrates the effectiveness of our proposed techniques.

Recent advances in Multimodal Large Language Models (MLLMs) have led to substantial progress in video understanding, accompanied by a growing number of long video benchmarks. However, existing benchmarks rely predominantly on web-sourced videos that lack inter-clip spatiotemporal continuity, making it difficult to assess whether models can maintain consistent memory across days or weeks of real-world experience. We introduce EgoMonth, the first month-level egocentric video understanding benchmark. EgoMonth comprises over 300 hours of first-person daily-life recordings from 20 participants spanning 20 to 120 days, paired with 1,443 human-crafted multiple-choice question-answer pairs. We design a cognitively grounded 14-task evaluation framework organized into three hierarchical cognitive levels: Schema Consolidation, Episodic Indexing, and Cascading Reasoning. Evaluation of state-of-the-art open-source and closed-source MLLMs reveals that even the best-performing model, Gemini 2.5 Pro, achieves only 71.8% macro-average accuracy, remaining 22.4 percentage points below the corrected human baseline of 94.2%. Several models perform near or below the 25% chance level on tasks such as Route Reasoning, Cross-view Spatial Reasoning, and Direction Judgement, while even the strongest closed-source model remains substantially below human performance. These results indicate that current MLLMs function as lossy summarizers rather than faithful memorizers, highlighting the need for architectures with genuine long-term spatiotemporal memory.

Multi-source evidence fusion under Dempster-Shafer theory faces two persistent challenges: existing conflict measures assess inter-evidence inconsistency and intra-evidence uncertainty independently, yielding incomplete evaluations, and current fusion methods evaluate evidence sources exclusively through instantaneous comparisns without exploiting their long-term reliability across diverse decision contexts. This paper proposes a unified evidence reasoning framework that addresses both limitations. Specifically, a chaos-conflict measurement is introduced to jointly quantify cross-evidence conflict and intra-evidence non-specificity, with five formally proven properties ensuring consistent assessment. A historical experience driven weighting scheme partitions the decision space via spectral clustering and applies regret theory to compute context-specific reliability profiles from past fusion outcomes. These mechanisms feed into a hybrid combination rule that adaptively balances uncertainty preservation against weighted consensus, controlled by the global conflict level, followed by a belief-interval decision strategy that enables robust classification without discarding epistemic uncertainty. Experiments on 16 real-world benchmark datasets demonstrate that the proposed framework achieves an average F1 score of 85.78 and a mean AUC of 93.30, outperforming eight DST-based baselines and three gradient boosting methods. Ablation analysis confirms the contribution of each component we proposed. The framework offers an effective approach for adaptive evidence fusion in multi-source decision making.

Research on automatic speaking assessment (ASA) has increasingly adopted multimodal speech large language models to assess learners' speaking performance. However, existing studies provide limited analysis of how acoustic and content information contribute to predictions and how stable the resulting performance is. We propose CASA, a simpler architecture combining Whisper-medium and Qwen3.5-2B that achieves state-of-the-art performance while providing a more interpretable separation between speech delivery and content. On the Speak & Improve Corpus 2025, CASA achieves a root mean square error (RMSE) of 0.358, improving on the previous best RMSE while using approximately half the estimated inference parameters. The general-purpose architecture is designed for adaptation to other ASA corpora without structural changes and relies on three handcrafted fluency features. Through ablations and repeated runs, we analyze the individual and complementary contributions of acoustic and content information, examine performance variability, and demonstrate the potential of large language model reasoning for training-free content validation.

Contemporary online assessment systems rely primarily on browser lockdown, webcam monitoring, and behavioural analytics, yet remain vulnerable to attacks that extract the assessment content itself through screenshots, screen sharing, optical character recognition, and automated scraping. This paper extends the Multi-dimensional Spatio-Temporal Context Camouflaging Model (MSCCM) within the MARS (Multi-modal Assessment Resilience Suite) by introducing the Multi-Layer Context Camouflaging Theory (MCCT), a mathematical framework that protects rendered assessment content through semantic superposition. Authentic assessment content and synthetically generated camouflage are represented as a unified rendering while remaining recoverable only by legitimate candidates. The framework models the adversarial extraction process through an explicit extraction-channel operator and develops six coupled constructs: the Context Inversion Operator, Contextual Lamination Operator, Separation Channel, Human Readability Functional, Computational Ambiguity Functional, and Context Camouflage Tensor. Computational ambiguity is formulated using conditional entropy, yielding a closed-form expression that quantifies uncertainty during unauthorized extraction, while legitimate recovery is guaranteed through an exact filtering identity. We further establish theoretical properties governing ambiguity, camouflage density, semantic preservation, multi-observation leakage, and temporal multiplexing, and present a rendering algorithm with computational complexity and a pre-registered evaluation protocol. MCCT provides a mathematically rigorous foundation for behaviorally adaptive, accessibility-aware, and computationally resilient digital assessment by securing rendered assessment content while preserving readability for legitimate users.

Limit order book (LOB) simulators are most useful to practitioners when they combine realistic market dynamics, computationally efficient sampling, controllable scenario generation, and the ability to generalize beyond the instruments seen during training---properties that existing agent-based and deep generative simulators provide only partially. We present \textbf{FlowLOB}, a conditional \textbf{flow}-matching generator of \textbf{LOB} trajectories, trained on multiple Hong Kong Exchange (HKEX) symbols at three sampling frequencies ($0.1$s, $1$s, $10$s) in tick-relative representation that transfers to unseen instruments. Because flow and diffusion models admit a common formulation, we train both with identical data, architecture, and budget, and sample both through the same fixed-step ODE solvers, yielding a controlled comparison of sampling efficiency and fidelity. Flow matching attains its best quality with only $10$ ODE-solver steps, whereas diffusion needs many more function evaluations to approach the same fidelity. At this efficient operating point, FlowLOB improves realism over baselines, two learned and two agent-based models, in most distributional metrics at the two finer sampling frequencies. We evaluate counterfactual controllability with a distributional test that asks whether changing a scenario condition moves the generated statistic toward the corresponding real tail regime; FlowLOB satisfies this criterion in most tested settings. Both realism and control effects transfer zero-shot on a held-out symbol. We additionally conduct ablation studies on the network architecture and the learning rate.

Neural combinatorial optimization (NCO) solvers report the best of many sampled solutions per instance, and the sample count is, by convention, identical for every instance. Whether a non-uniform allocation of a fixed total budget would buy anything has not been measured. We measure it, and we audit the measurement itself. First, on in-distribution workloads the allocation headroom is not detectable. Across three pretrained solvers (POMO, AM, SymNCO) on uniform TSP-100, an oracle allocation computed and evaluated on the same stored samples reports a 2.2-2.6% gain with intervals excluding zero; measured out of sample the same gain is indistinguishable from zero (0.457, 0.015, -0.512 percent). Following the customary in-sample procedure, all three solvers would have supported a published 2%-level gain that does not exist. We calibrate this bias against an instance-wise null in which the true gain is zero by construction; over the ranges we test it does not shrink with more samples or more instances. Second, the same correction that removes the phantom gains preserves a real one. Under distribution shift (a workload mixing uniform and clustered instances), a pre-registered confirmatory experiment finds that allocation guided by held-out sample statistics improves best-of-k by 11.5% (AM, primary endpoint; 95% CI [7.4, 19.7]) and 12.0% (SymNCO, replication) at equal evaluation budget, with the signal-acquisition cost not charged; a pre-registered negative control (POMO, an order of magnitude more robust to shift) shows -0.3% [-0.7, 0.24]. The gain exceeds a frozen distribution-label baseline by 4.2 points [1.9, 7.7]. An exploratory policy charging a 20-sample probe against the same budget retains 3.4% (AM) and 4.6% (SymNCO). We give a correction procedure and a reporting checklist, and release all data, code, and the pre-registration record.

Generative models of limit orderbook (LOB) data have advanced rapidly, but their evaluation often focuses on stylised facts and selected market statistics. These measures provide useful diagnostics but may not capture the joint temporal and cross-level structure of order-book trajectories. We introduce LOB-ID, an embedding-based framework that adapts the Fréchet Inception Distance (FID) and Monge Inception Distance (MIND) to LOB data. To obtain domain-specific embeddings, we train the DeepLOB architecture on four months of Level-2 order-book data for five equities. We show that LOB-ID is stable across time, instruments, and embedding checkpoints, and rises monotonically under controlled distortions. We then construct a moment-matching attack against FID and a deep-book perturbation that evades statistic-based evaluation. MIND remains substantially more sensitive to both distortions. Finally, we score five generative LOB models, spanning stochastic baselines and deep learning approaches, and find that LOB-ID ranks them in line with the joint temporal and cross-level structure each captures by construction.

This paper proposes a novel learning-based approach to approximately solve instances of mixed-integer optimization problems. These problems are computationally challenging, as they require jointly determining discrete and continuous decisions while satisfying complex combinatorial constraints. The proposed method relies on a graph-based generative diffusion model that learns the discrete component of mixed-integer optimization problems while integrating a training-free feasibility projection operator directly into the reverse diffusion process to steer intermediate samples toward the feasible set throughout generation. Once the discrete decisions are generated, the remaining optimization reduces to a continuous problem that can be solved efficiently (relative to the original problem) using existing numerical methods. The resulting framework named Constrained Graph Diffusion (CGD), is problem-agnostic and can accommodate a broad class of mixed-integer optimization problems through suitable projection operators. We evaluate CGD on optimal transmission switching for ACOPF and discrete portfolio optimization, demonstrating substantial improvements in feasibility and solution quality over learning-based baselines while achieving speedups of up to $425\times$ over state-of-the-art numerical solvers for MINLPs.

Formal verification provides strong guarantees of software correctness, but its adoption is limited by the high cost of writing precise formal specifications. While recent large language models (LLMs) have shown strong capabilities in theorem proving and verified code generation, their true ability to generate program specifications remains unclear. Existing evaluations require either verifying implementation conformance or proving semantic equivalence between specifications, both of which are formidably difficult and may conflate proof difficulty with specification quality. To address this problem, we introduce Coins, a Rocq based evaluation framework that assesses specification quality by instantiating specifications under evaluation on trusted test cases and generating concrete proof obligations. This design aligns with the asymmetric nature of formal reasoning, where successful proofs provide reliable evidence while proof failures are inherently ambiguous. Using Coins, we conduct a large scale study on HumanEval with a curated set of human written Rocq specifications. Our results show that specification generation remains a formidable challenge, and that verification complexity can obscure genuine differences in specification quality. Overall, we find that accurate specification evaluation, rather than model scaling alone, is central to understanding the power of LLMs for specification synthesis, and that test case based formal reasoning offers a more faithful and discriminative measure of progress.

Large Language Models (LLMs) have achieved remarkable success in natural language understanding and generation, but their deployment is constrained by high computational demands. Deploying smaller LLMs directly on the edge can circumvent this, but with degraded accuracy. Deploying smaller cloud-based big LLMs preserves performance, but at the cost of expensive per-token computation. We present a distributed inference framework, \our{}, that integrates speculative decoding (SD) across edge and cloud. A compact draft model deployed on the edge generates candidate tokens rapidly, and a large verifier model on the cloud validates these tokens in parallel. Accepted tokens are retained, while only rejections trigger verifier correction, substantially reducing the number of cloud queries. Our plug-and-play design shifts the bulk of computation to the edge, significantly lowers inference time and cloud cost, and preserves the accuracy of the big model without any retraining requirement. Our approach demonstrates a practical path toward scalable, cost-efficient, and accurate deployment of LLMs in real-world environments. Experimental results across multiple Natural Language Processing tasks using SpecBench and CNN/Dailymail datasets demonstrate that \our{} reduces the cloud model calls by $76\%$ with zero loss in accuracy as compared to the full model.

Significant health risks are associated with the illegal, yet commonly practiced use of industrial-grade Calcium Carbide (CaC2) for ripening climacteric fruits like mango and banana, which leaves behind trace residues of arsenic and phosphorus. To address this, the proposed study explores a novel, non-invasive multispectral framework for distinguishing safely ripened fruits (naturally ripened and ethephon-induced) from calcium carbide-ripened samples, while also estimating their ripening progression (in percentage) and remaining shelf life (in days). The spectral profiles of mango (Mangifera indica) and banana (Musa acuminata) at 18 discrete wavelengths in the visible-near infrared (NIR) range (410 nm - 940 nm) are studied using the AS7265x spectral triad sensor. CaC2-treated samples exhibit sharper spectral intensity drops in the visible region, consistent with accelerated chlorophyll degradation and carotenoid development. To characterize these physiological changes, the feature engineering strategy integrates inter-method spectral variance, intensity ratios at distinct wavelengths, and environmental parameters including temperature and humidity. Dimensionality reduction using Principal Component Analysis (PCA) retains >90% of spectral variance within the first 5-7 components. The resulting feature set is used to train three independent eXtreme Gradient Boosting (XGBoost)-based learning algorithms for ripening method classification, along with quantitative estimation of remaining shelf life and ripening progression. A classification accuracy of 95% along with carbide class recall of 0.67 is observed for mango samples, while the model achieves an accuracy of 81% and carbide class recall of 0.74 for banana. This instrumentation and data-driven approach demonstrates the effectiveness of the proposed non-invasive framework.

Electroencephalography (EEG) decoding models often generalize poorly across datasets and subjects due to domain shifts in acquisition protocols and individual neurophysiology. We propose EEG-PRIME, a two-stage EEG foundation model for cross-dataset multi-task decoding. EEG-PRIME combines masked pretraining with prototype-aligned instruction tuning to enable instruction-aware and subject-invariant decoding across diverse BCI paradigms. During pretraining, an EEG encoder learns transferable representations through masked reconstruction with frequency-cutoff spectral augmentation. During instruction tuning, EEG-PRIME incorporates task-semantic, dataset-specific, and subject-invariant conditioning. The resulting conditioning signal modulates the Q-Former through Layer-wise Query Modulation, while frozen text embeddings of class labels serve as prototypes for cosine-similarity-based prediction across heterogeneous label spaces. Experiments on sixteen datasets covering motor imagery, emotion recognition, ADHD detection, covert speech, and mental workload show consistent improvements over state-of-the-art baselines and prior EEG foundation models under cross-subject settings. On two additional held-out datasets, EEG-PRIME achieves balanced accuracy comparable to within-session calibration models without target-domain optimization, calibration, or linear probing, demonstrating promising zero-shot transfer capability.

Large language models (LLMs) are predominantly aligned to function as passive, sycophantic assistants. We challenge this default paradigm by empirically evaluating the cognitive plasticity of open-weight architectures when subjected to rigorous behavioral reprogramming. Our objective is to induce a proactive, Socratic conversational framework, characterized by high-frequency question generation under strictly constrained high-performance computing (HPC) conditions. Through a massively parallelized hyperparameter sweep comprising 405 HPC jobs, we define precise mathematical bounds for parameter-efficient fine-tuning (PEFT). We identify an architectural threshold at LoRA rank $r=16$ and demonstrate via extensive epoch ablation that generalization capacity strictly reaches its optimal convergence within an optimized training window of $e \in [2, 3]$ depending on dataset density (minimum validation loss of 0.919). Furthermore, scaling model capacity to 14B parameters yielded a lower localized evaluation perplexity (1.414). Subsequent Direct Preference Optimization (DPO) successfully decoupled the underlying assertive behavior from localized syntax, while rigorous cross-lingual stress testing reveals both the capabilities and the structural boundaries of zero-shot persona transfer, demonstrating robust alignment in closely related linguistic families alongside identifiable degradation pathways in morphologically distant targets. These findings establish a rigorous empirical framework for compute-efficient, cross-lingual behavioral modification.

Prior work on LLM behavior under anomalous conditions asks whether a model notices anomalies. We ask a narrower question: once a model sits in a workflow with a low, controllable failure rate, does its explanatory engagement - length, specificity, self-reported confidence - change as failure grows asymptotically rarer? We built a local, zero-cost harness on three open-weight models (qwen3:8b, llama3.1:8b, mistral:7b) running a repeated tool-call task where one call fails at probability p, swept across eight rates from 0.2 to 0.0001, under five elicitation conditions from immediate prompting to none. We hypothesized a rise in engagement as failures grew rarer, then a collapse near a detectability threshold. Pooled across conditions this appeared false: length fell in a flat, monotonic pattern. Splitting by condition overturned that. Under immediate_forced, where the model must explain every failure instantly, the predicted rise is confirmed but followed by a plateau, not a collapse: length peaks at 28.4 words at p=0.05, settles to 17.4-19.0 words at the rarest rates, and confidence rises unevenly from about 53% to the 70s-90s. Under grouped_runs, explanation batched to run-end, no collapse appears. Under passive_unprompted, aggregate magnitude is a floor artifact, but a recovered logging gap revealed real, model-specific self-monitoring: llama3.1:8b volunteers structured confidence reports unprompted, sometimes eroding its own confidence as trials accumulate; the other two do so only once, as boilerplate. Elicitation structure is a first-class moderator of collapse observability. A companion guaranteed-failure run (72 cells, backfilling rates where random sampling gave zero real failures) shows models differ in whether they recognize an anomaly, distinct from engagement once recognized. Limitation: discrete rate points cannot capture behavior between them, a direction for future work.

As the feature representation changes, replay must preserve the earlier classes. However, only a bounded active exemplar set can be replayed. We propose Uniform Herding, which allocates the current active set across observed classes and uses a bounded candidate pool to refresh their chosen exemplars in the current representation. On CIFAR-100 with ten class-incremental tasks, a ResNet-18 backbone, active budget $M=2{,}000$, retrieval budget $b=64$, and three seeds, Uniform Herding obtains $44.00\pm0.51\%$ final average accuracy and $17.22\pm0.43\%$ forgetting, compared with $42.33\pm1.20\%$ and $24.87\pm1.11\%$ for iCaRL. Within the Uniform Herding protocol, final accuracy decreased when NME or herding was replaced with the tested alternatives, while forgetting increased when distillation was removed. Changing the retrieval budget has a smaller effect across the tested range than changing the active budget. The comparison with iCaRL is end-to-end. It does not isolate the effect of refresh from the other protocol differences. These results are limited to the tested protocol.

Machine learning theory studies learning procedures through mathematical setups in which the data model, training protocol, oracle access, loss, metric, and randomness define the phenomenon that a theorem is meant to explain. Solving an open problem therefore requires the problem formulation, theorem target, and proof mechanism to be developed in concert. Researchers formulate hypotheses, test them through preliminary theoretical or empirical analysis, and refine both assumptions and proofs. We investigate whether this process can be organized as an autonomous agentic workflow for ML theory research. We develop VALG, an agentic system that combines multi-level Verification, Adaptive formulation of Learning-theory problems, and Graph-structured proof development. Within each source-relative theorem branch, VALG maintains a fixed mathematical specification, checks the theorem-level composition of a typed proof-dependency graph, and constructs and reviews local proofs in dependency order. When a proof attempt fails, VALG identifies whether the obstruction lies in a derivation, the proof structure, or the theorem formulation and routes the next attempt accordingly. Formulation-level obstructions initiate an explicitly related variant or relaxation, preserving the mathematical relation between the resulting theorem and the source problem. We evaluate VALG on nine subproblems from five COLT 2026 open problems. Two runs produce internally finalized theorem candidates that match the scope of their source briefs; the remaining seven yield restricted-method results, special cases, or conditional theorems. These case studies show how VALG keeps source-scope matches, relaxations, conditional results, and blocked attempts mathematically distinct. VALG is open source at https://github.com/DechenZhang/VALG-ML-Theory-Agent.

In expert-parallel (EP) MoE serving, every layer synchronizes at the slowest GPU. Dispatchers balance token counts (EPLB, LPLB, UltraEP) or activated-expert counts (METRO), assuming expert time is linear in one. Measurements on two datacenter GPU generations show it is neither: below $\nstar\!\approx\!156$--$168$ tokens, HBM weight streaming dominates---cost attaches to \emph{activated replicas}, not tokens; above it, grouped GEMM rounds tokens to 128-tile $M$-tiles, so \emph{splitting} an expert adds padded compute. A max-affine profile $t=\max(a+bG,\,c+βN)$ captures both regimes. Realistic decode batches hold hot experts in the linear regime and cold in the flat \emph{simultaneously}; recorded batches show proxy dispatches differ by $1.4$--$1.6\times$ in modeled block time (p95 up to $1.7\times$), and \emph{which} proxy wins flips with the regime. We formalize per-batch dispatch as a fixed-charge makespan problem---NP-hard on two fully replicated GPUs, polynomial in degenerate limits---and present \sys{}, a makespan-aware dispatcher solving it in milliseconds off the critical path; its SGLang integration runs out-of-process and fuses dispatch with count collection into one in-graph kernel. Anchored by an 8-GPU Testbed~A microbenchmark, \sys{} stays within 1\% of the best fixed baseline everywhere and wins by up to $15.5\%$ where regimes mix. End-to-end on Testbed~B, Qwen3-235B (inside the win region) gains $4$--$6\%$ throughput and cuts p99 latency by ${\sim}15.6\%$; DeepSeek-V3 (outside, communication-dominated) shows only mechanism cost. A phase diagram, not a universal win, is the claim: it predicts both outcomes before deployment.

When a security researcher publishes a report on a cyberattack, detection engineers are supposed to turn it into working detection rules. In practice, most automated attempts at this only extract the simplest clues from the report --- bad IP addresses, domain names, and file hashes --- and turn them into block lists. This is a weak strategy, because attackers can change these simple clues within hours or days, so the resulting detections stop working almost as soon as they are deployed. Security teams describe this idea with the Pyramid of Pain. This project asks whether feeding a report into a knowledge-graph retrieval system, Microsoft GraphRAG, rather than a standard vector-similarity retrieval system (Naive RAG), produces detection plans that rely more on these durable, top-of-pyramid clues. Both systems are given the same report, the same generation instructions, and the same language model to write the final plan; only the retrieval step differs. In a detailed case study of one APT28 report, the GraphRAG plan kept firing at 100\% of its detections after every IP address, domain, and file hash in the report was rotated, while the Naive RAG plan kept firing at only 29\%. Repeating the comparison across nine real CTI reports from four vendors confirms the same pattern: GraphRAG plans consistently reach higher, harder-to-evade levels of the pyramid, even when the two systems end up close on total score. The results support treating knowledge-graph-aware retrieval as the architecturally correct foundation for automatically generating SOC-deployable hunting plans, while showing that the wording of the generation prompt matters almost as much as the retrieval back-end itself.

In this work, we introduce DMDIntel which uses dynamic mode decomposition (DMD) to make the predictions made by LLMs in a classification task interpretable. It develops an input attribution pipeline, that first decomposes the hidden states of an LLM into prominent patterns, also known as modes, and then associates ranks to the input tokens based on the projection values on those modes. Rigorous experiments across three datasets and three model families consistently show that the ranked attribution of input tokens obtained using DMDIntel by far outperforms state-of-the-art techniques such as principal component analysis, integrated gradients and SHAP.

Organizational decisions are co-created while evidence, constraints, and human priorities continue to evolve. In conventional transcript-based multi-agent systems, humans typically provide an initial problem, agents deliberate internally, and the system returns a final response. BoardroomAI instead treats the human as a persistent participant who can intervene by challenging assumptions, modifying constraints, changing priorities, introducing evidence, or redirecting the decision process. We operationalize this human--agent coexistence through four components: (i) a typed decision graph representing evidence, assumptions, constraints, claims, objections, alternatives, risks, decisions, semantic dependencies, and specialist responsibility; (ii) an intervention compiler that converts confirmed human actions into explicit graph updates; (iii) dependency-aware propagation that identifies affected subgraphs, preserves unaffected artifacts, and selectively reactivates relevant specialists; and (iv) an evaluation framework measuring intervention impact, repair coverage, preservation, recomputation, and decision validity. Across 600 generated decision-DAG interventions, propagation matched exhaustive impact computation while inspecting only 14.59% of nodes. In a 12-case exploratory pilot, selective repair recomputed 62.11% of canonical nodes, preserved all gold-unaffected nodes, and produced valid updated decisions in six cases while abstaining in the remaining six. These abstentions show that correct intervention routing may still provide insufficient context for synthesis, motivating a \emph{decision-sufficient context closure} for human-steered multi-agent deliberation. All results are synthetic and prototype-level.

Diffusion models have achieved dominant performance in visual generation but suffer from substantial inference overhead. While cache-based acceleration has emerged as a promising solution, existing policies rely on local similarity heuristics, which we identify as being significantly misaligned with final generation quality. This discrepancy stems from the non-uniform propagation and accumulation of errors along the denoising trajectory. To address this, we propose Global-Impact Cache (GCache). We first establish a rigorous theoretical characterization of the error propagation upper bound. Recognizing that this bound can be overly conservative for complex, highly non-convex diffusion models, we further reparameterize the propagation exponent with a Bernstein form and reformulate cache policy search as a bilevel optimization problem. In detail, GCache identifies an optimal reuse policy in the inner objective while aligning the error-weighting function with generation quality loss in the outer objective. This framework effectively reconciles theoretical rigor with empirical performance, learning to prioritize computation where it most impacts visual fidelity. Extensive experiments demonstrate that GCache consistently outperforms prior caching strategies on both video and image generation. Notably, on the state-of-the-art Wan2.1 video diffusion model, GCache maintains a 2.17x speedup while significantly enhancing generation quality, reducing LPIPS from 0.1095 to 0.0316.

Enabling agents to learn from experience and internalize it into their policy has become a central problem in self-evolving AI. On-policy self-distillation (OPSD) offers an effective pathway by using a privileged self-teacher to provide dense supervision on the student's own trajectories; however, existing methods still rely heavily on designer-specified privileged artifacts (e.g., answers, feedback, skills, or trajectories), limiting the end-to-end learnability and scalability required for continual self-improvement. In this work, we introduce Latent On-Policy Self-Distillation (LOPD), which, rather than proposing another hand-crafted OPSD variant with a newly prescribed form of privileged context, makes the teacher's privileged context itself learnable end-to-end from experience. Technically, LOPD retrieves relevant experiences and composes them into continuous latent tokens that condition a self-teacher, while the student generates trajectories from the task and interaction history and receives dense token-level supervision at every visited prefix. We further introduce a privileged-margin objective to stabilize and regulate the learning of latent context. Empirically, LOPD demonstrates (I) strong performance, outperforming RLVR and representative OPSD methods including OPSD, SDPO, and Skill-SD across both agentic tool use and code generation; and (II) high learning efficiency, surpassing GRPO and Skill-SD with less than 30% of their rollout budget. Ablation studies further provide direct evidence that making privileged context learnable is necessary for realizing these gains. Together, these results position LOPD as a step toward a more scalable and self-directed paradigm for agent evolution.

The paper introduces the ante-hoc Explainable AI methodology to assess the global feature importance of the Machine Learning models used for heat demand forecasting in intelligent control of District Heating Systems, with motivation to facilitate their interpretability and trustworthiness, hence addressing the challenges related to adherence to communal standards, customer satisfaction and liability risks. Methodology includes use of four different approaches, namely intrinsic interpretability of Gradient Boosting method and selected post-hoc methods, namely Partial Dependence, Accumulated Local Effects and SHAP. None of the selected methods assume feature permutation or perturbations which can introduce bias due to introduction of random unrealistic values of data instances. Discussion of results is provided, including the assessment of complementarities where applicable, with specific interpretations in context of the district heating processes.

Traffic video understanding has become an important problem in intelligent transportation, as road videos provide direct evidence for accidents, violations, and interactions between vehicles and vulnerable road users. A useful system should explain how a traffic event develops, why it happens, and when the relevant interaction occurs, yet this remains difficult for multimodal large language models (MLLMs) because traffic videos contain sparse events and varied viewpoints. We introduce UniTraffic-Agent, the MR-CAS solution for Track~3 of the 10th AI City Challenge, which includes Traffic Anomaly Reasoning (TAR) and two out-of-domain evaluations: FETV for fisheye traffic events and PSI-VQA for pedestrian intention reasoning. UniTraffic-Agent follows an observe--reason--act--verify workflow that samples timestamped visual evidence, reasons over all questions from the same clip in one request, and converts responses through task-specific action adapters. On the official Public leaderboards, MR-CAS ranks 16th on TAR with a score of 0.5780, 2nd on FETV with 0.4884, and 4th on PSI-VQA with 64.4161. The code is available at https://github.com/Roclp/UniTraffic-Agent.

The field of bioinformatics struggles with legacy code - old code that is commonly used but may no longer have a maintainer, or may be written in an now-unfamiliar language (e.g. Perl, Fortran). This incurs maintenance cost (technical debt), but dynamically typed languages also negatively impacts the environment and fail to make use of modern hardware. Legacy code may also have security or safety problems that make it unsuited for use in clinical settings. Here we show that agentic AI, combined with static analysis, can be used to translate legacy code to the modern language Rust. We provide prompts and supporting software to aid systematic translation, and evaluate it on common software for NGS and imaging. We showcase the result on our software Bascet: Size was reduced by ~80x, build time decreased by ~10x, and performance of key steps improved >3x. Unix dependencies were also removed, making Bascet the only single-cell pipeline able to run on native Windows, without a container. Large-scale refactoring of bioinformatics software is thus now possible at a limited budget, enabling more complex tools to be developed.

Relational Deep Learning (RDL) models multi-tabular databases as temporal heterogeneous graphs to enable end-to-end representation learning. However, prevailing RDL evaluation practices rely on static, single-episode dataset snapshots, overlooking the continuous, time-evolving nature of real-world databases. Consequently, current RDL benchmarks fail to capture how model performance changes as new data accumulates over time. To address this limitation, we introduce an incremental, multi-episode evaluation and training paradigm to assess and improve the temporal robustness and adaptability of state-of-the-art RDL models. Using established large-scale datasets, we examine data evolution and model training dynamics, demonstrating that temporal concept drifts occur in the majority of predictive tasks. We present multiple incremental training regimes for fine-tuning the models and demonstrate that transfer learning is both feasible and highly effective in the RDL setting. Alongside a new temporal evaluation metric that prioritizes near-future accuracy, we show that our incrementally fine-tuned models consistently outperform the standard, expensive, from-scratch trained baselines.

Standard probabilistic logic programming frameworks typically rely on grounding logic programs into discrete propositional representations. This operational requirement restricts exact inference to finite domains and discrete probability distributions. In this paper, we introduce Measure-Theoretic Probabilistic Definite Clause Logic (MT-PDCL), a generalized foundational framework that eliminates this finite-domain restriction. By explicitly defining stochastic variables over bounded index domains and equipping the interpretation space with standard Borel $σ$-algebras, MT-PDCL allows logical variables to operate natively over continuous measurable spaces. Building on Continuous Distribution Semantics, MT-PDCL models probabilistic rules as mutually independent causal events. However, rather than aggregating these derivations via finite boolean circuits, declarative entailment is formally defined through exact Lebesgue integration over the continuous measure space. We introduce a continuous immediate consequence operator that unifies the integration of continuous prior distributions with the evaluation of exact continuous observations. We demonstrate that this approach replaces the combinatorial bottleneck of discrete grounding with exact, algebraic, and structurally differentiable inference. While this transition trades discrete combinatorics for the geometric curse of dimensionality, it achieves the expressive power of continuous probabilistic models while preserving the pure declarative syntax of definite clause logic.

Retrieval-augmented generation treats an external corpus as inference evidence, allowing injected documents to promote attacker-chosen claims. Existing detectors depend on trusted references, specific attack artifacts, or global thresholds sensitive to corpus topology. We present RAGSieve, a self-referenced detection framework that constructs its reference from the inspected system. RAGSieve-Query (RSQ) performs query-local contrast, scoring top-five candidates against ranks 6-20 of the same retrieval to detect answer-anchor concentration and carrier transitions. RAGSieve-Graph (RSG) performs corpus-local contrast, comparing each document's semantically similar but lexically distinct neighbors with its local baseline to detect coordinated density before queries arrive. Across three QA datasets and six poisoning constructions, RSQ achieves 95.2% AUROC and detects 82.2% of poison at 5% clean-document removal, versus 81.1%/52.5% for GMTP. RSG achieves 93.3%/79.8%, versus 79.4%/37.6% for CleanBase. Joint deployment reduces attack success from 67.4% to 14.0% while retaining 41.3% F1 on unpoisoned retrieval, demonstrating practical protection at both corpus ingestion and query time without poison labels or trusted corpora. Source code is available at https://github.com/XrazyMee/RAGSieve.

Multi-hop retrieval must recover passages that provide sufficient evidence together. An initial passage often resolves an entity or relation implicit in the question, making the missing evidence easier to describe only after retrieval begins. Graph retrieval improves access to related evidence through stored corpus structure, but its retrieval signal is commonly derived from the original question. Complementary evidence must then be reached through stored relations even when an observed passage provides a more direct semantic cue. We introduce EviReform, which separates revising the retrieval request from aggregating evidence in the graph. Retrieved source passages formulate residual queries for the unresolved information need. The original and residual retrieval signals are normalized separately, combined, and propagated between propositions that share entities. On 2WikiMultiHopQA, HotpotQA, and MuSiQue, EviReform exceeds the strongest baseline by up to 5.59 Recall@5 points and 4.50 F1 points. These results show that observed evidence can guide graph retrieval toward the part of a supporting chain left underspecified by the original question. Code is available at https://github.com/XrazyMee/EviReform.

Knowledge-base question answering (KBQA) systems rely on effective retrieval and reasoning mechanisms to generate accurate answers from external knowledge sources. However, developing reliable KBQA systems for low-resource languages such as Bangla remains challenging due to limited retrieval-focused research, scarce language resources, and difficulties in grounding generated responses in external knowledge. In this work, we propose HybridRAG-BN, a retrieval-augmented framework for Bangla KBQA that integrates hybrid retrieval using BM25 and BGE-M3, answer generation using the GGUF version of Gemma-4-31B-Instruct, and a LoRA-fine-tuned Gemma-4-31B-Instruct model for answer verification and refinement. To further improve robustness, the framework incorporates a post-processing stage that addresses unresolved cases through fallback answer replacement and DuckDuckGo-assisted retrieval. Experimental results demonstrate the effectiveness of the proposed framework, achieving token-level F1 scores of 0.71654 and 0.72912 on the public and private leaderboards, respectively, securing first place in the competition.

Heterogeneous USV cooperative pursuit in constrained port waterways requires evader interception under navigation, traffic, and role constraints. This paper proposes OGR-MARL, an option-guided residual multi-agent reinforcement learning framework that is decoupled from a specific MARL algorithm. OGR-MARL integrates shared evader belief, role-conditioned option targets, adaptive rule penalties, and residual policy learning, allowing different MARL algorithms to learn corrective actions on top of rule-guided behaviors rather than exploring constrained port environments from scratch. We instantiate OGR-MARL with representative continuous-control MARL backbones, including MADDPG, MATD3, MAPPO, and MASAC, yielding OGR-MADDPG, OGR-MATD3, OGR-MAPPO, and OGR-MASAC. Experiments in an abstract Xiazhimen port-waterway scenario show that the OGR-MASAC instantiation achieves a 75.0% capture rate, promising mission-effective rule compliance, and the best heterogeneous coordination among the tested methods. Without retraining, zero-shot transfer to a QGIS/AIS-informed Xiazhimen map achieves promising results, demonstrating the generalization potential of OGR-MARL in more complex port scenarios.

Long-horizon LLM agents must preserve information from past interactions to support future tasks. Existing memory systems typically rely on eager consolidation, invoking LLMs after each interaction to extract, summarize, or update memories. This design makes memory construction increasingly costly as conversations grow. Coarse summarization can reduce construction cost but risks discarding fine-grained contextual evidence, whereas larger retrieval contexts or multi-hop LLM reasoning shift the overhead to query time. We present LycheeMemory V2, an efficient long-term memory framework that replaces turn-level consolidation with semantic segment-level consolidation. Instead of consolidating every interaction, LycheeMemory batches multiple exchanges into segments and encodes each finalized segment into context-independent typed memory records. Segment-level batching lowers LLM encoding frequency, while semantic boundary detection helps preserve coherent event-level and temporal evidence compared with fixed-window batching. The resulting records are organized with lightweight structured indexes for query-planned evidence retrieval. Experiments using GPT-4.1-Mini show that LycheeMemory achieves state-of-the-art performance, reaching 89.22% on LoCoMo and 92.20% on LongMemEval-S. Compared with A-Mem, it reduces construction tokens by 86.0% on LoCoMo and 75.9% on LongMemEval-S without increasing query-time token usage. More broadly, our results suggest that the accuracy--cost trade-off of long-term agent memory depends not only on what information is retained, but also on the granularity at which it is consolidated.

Pretrained tabular foundation models have demonstrated strong predictive capability; however, their application to large-scale datasets remains constrained by the limited inference context. This paper introduces Balanced Adaptive Prototype Selection (BAPS), a framework for constructing compact, information-preserving contexts for scalable TabPFN inference. Without modifying or retraining the pretrained model, BAPS jointly preserves representative structure, informative decision boundaries, local density, class balance, and feature-space diversity. Experiments on the million-row HIGGS and SUSY datasets show that 512 prototypes retain strong predictive performance and reliable calibration, corresponding to an approximately 1,953-fold context compression. All experiments were conducted on an Intel Core i7 CPU with 16 GB RAM and no GPU acceleration. These findings establish effective context construction as a practical mechanism for extending pretrained tabular foundation models to million-scale datasets.

Generative information retrieval (GIR) has emerged as a compelling alternative to the conventional index-retrieve-then-rank retrieval pipeline by training a generator to produce the identifiers of relevant items directly. Despite its promise, a number of open challenges still remain. First, constrained left-to-right decoding is vulnerable to prefix-level errors and local optima. Second, most prior GIR research remains largely unimodal, leaving instruction-aware retrieval across text, image, and mixed image-text items underexplored. Third, although discrete identifier-based GIR offers higher efficiency, its retrieval accuracy still lags behind that of the cutting-edge dense-vector-based retrieval methods. Motivated by these challenges, we propose DrIG, a novel Generative framework for universal multimodal retrieval featuring Dual-role Identifiers, which supports diverse retrieval tasks across multiple modalities and domains. Each candidate is assigned a single residual-quantized identifier that serves two complementary roles. In its sequential role, the identifier is decoded autoregressively, where the first token explicitly models modality and the remaining tokens capture progressively finer semantics. In its set-based role, the same tokens are reinterpreted as an unordered set to provide a prefix-independent relevance prior, which guides constrained beam search and alleviates local-optimum errors. Extensive experiments on the M-BEIR benchmark and the text-to-image evaluation datasets show that:(1)DrIG consistently outperforms state-of-the-art generative multimodal baselines across diverse tasks, while hybrid reranking achieves a favorable efficiency-effectiveness trade-off against strong dense retrievers. (2)Ablation and scaling analyses reveal how the base LMM, beam size, reranking depth, and fusion strategy affect retrieval performance, providing practical guidance for system design.

Long-form research reports generated by large language models drift, contradict themselves, and lose provenance: the same metric appears with different values, and rumor is quoted as confidently as an audited filing. We present a two-tier agentic system that separates a maintained, point-in-time knowledge library from report writing. A deterministic "librarian" ingests timestamped sources into a trust-tiered ontology, layering evidence cards, an authoritative metric ledger, and a claim graph into an always-current source of truth, not per-query RAG over raw chunks. A portable multi-agent "writer" runtime then composes a contradiction-free, evidence-grounded report at any knowledge cutoff T, reading only evidence with as_of <= T (no look-ahead); red-team verdicts flow back into the librarian. We evaluate on a self-collected, public corpus of 6,130 sources yielding 555,926 evidence cards (SEC EDGAR filings across 295 issuers and 11 sectors, U.S. Bureau of Labor Statistics releases, and Wikipedia). From the one library we compose four point-in-time reports on distinct theses and run eight reproducible experiments, whose headline metrics come from a deterministic quality-control gate, itself validated by defect-injection meta-evaluation at recall 1.0 and precision 1.0. A shared metric ledger removes 6,845 cross-section contradictions to zero. Tier-first selection is correct on 22/22 gold cases where a popularity-first baseline scores only 9/22; trust tiering leaks zero media-sourced numbers, and no government statistic displaces a company's own filing. A red-team refutation propagates back and self-corrects a later run with zero manual edits. Replay exhibits zero look-ahead violations across seven cutoffs while the library grows from 235,373 to 555,312 cards. Difficulty-tiered model routing exceeds the all-Opus quality ceiling while running 3.7x faster than serial.

In this research work, we are constructing the sensing matrix, which is essential for the success of the compressive sensing technique. We have chosen a learning-based technique for the construction of the sensing matrix. The novelty and uniqueness of the proposed technique is that it does not use any data set and also does not use a specific application. It uses the mathematical property/constraint for the construction of the sensing matrix for the perfect recovery of the signal. The perfect recovery of signals is an old and still very challenging problem in real-world applications. In late 2000, compressive sensing became a popular mathematical tool for the perfect recovery of sparse signals. The core of the compressive technique is the construction of the sensing matrix, which satisfies certain special properties such as restricted isometry property (RIP), null space property (NSP), and spark property (SP). All these properties are NP-hard problems and hence computationally challenging to solve. For all practical purposes, the construction of the sensing matrix needs to achieve low mutual coherence to achieve the perfect recovery of the signals. We have used a neural network for the construction of the sensing matrix, and this framework constructs a binary sensing matrix with low mutual coherence. The entries in the matrix are generated through a shared underlying rule. The proposed architecture is simple and does not use large-scale training data sets. Such uniqueness and novelty bring a drastic reduction in computational cost, and also, for the first time in literature, the use of a mathematical property for defining the loss function. In this proposed research work, the mutual coherence property has been used in the neural network framework. Such a neural network framework brings generality, robustness, and reduces storage requirements.

The expanding operational capabilities of large language model (LLM) agents introduce sophisticated security threats. Runtime defenses have emerged as an effective approach to mitigating these risks by integrating security mechanisms into the agent execution loop. However, existing runtime defenses rely heavily on manually designed interventions and lack a principled framework for their construction and maintenance. In this work, we first develop a harness-level formulation of runtime defense that systematically characterizes how harness mechanisms enable defense construction and provides a unified view of existing runtime defense interventions from a harness perspective. Building on this formulation, we propose HARD (Harness-based Autonomous Runtime Defense Evolution), a self-evolving runtime defense framework that automatically identifies appropriate intervention strategies and iteratively improves defense artifacts based on observed failure traces. HARD transforms runtime defense development from manual engineering into an autonomous evolution process, and extensive experiments demonstrate that it improves security performance over existing handcrafted defenses while preserving benign task utility. Our findings highlight autonomous defense evolution as a promising new paradigm for securing deployed LLM agents, enabling agents to identify defense weaknesses and continuously improve their protection mechanisms.

McCoy & Griffiths (2025, henceforth M&G) suggest that a Bayesian prior can be distilled into Artificial Neural Networks (ANNs) through Model-Agnostic Meta-Learning (MAML, Finn et al., 2017). They support this empirically by showing that meta-trained networks demonstrate formal language learning abilities comparable to Yang & Piantadosi (2023)'s Bayesian learner, significantly outperforming standard ANNs. We point out that under the standard interpretation of a prior, M&G's procedure does not actually instill one; it merely initializes network weights favorably, leaving the objective function unchanged. We then consider a more permissive interpretation, where the system as a whole can be seen as implementing a Bayesian learner even without an explicit prior in the objective. We show that this interpretation faces nontrivial challenges. Finally, we assess how well MAML approximates the empirical results of Bayesian learning, showing that unlike genuine Bayesian learners, M&G's model overfits and generalizes poorly to unseen data.

In this paper, we study how to perform statistical inference for quantile temporal difference learning (QTD) in distributional reinforcement learning. Assuming access to a generative model, we first establish functional central limit theorems for both synchronous and asynchronous QTD, which show that the averaged iterates of QTD converge weakly to a rescaled Brownian motion. We next provide online inference methods. Based on random scaling, the inference procedure constructs an asymptotically pivotal statistic for inference by using the information along the whole QTD path. Meanwhile, the proposed statistic can be computed online without storing the entire trajectory of QTD iterates. This substantially reduces the memory requirement and enables efficient statistical inference in distributional reinforcement learning.

LLM-based software (LBS) integrates large language models as core components to deliver flexible, personalised responses. Unlike traditional software with deterministic outputs, LBSs exhibit context-dependent, stochastic behaviour that renders classical acceptance testing and test oracles insufficient: the same query may require fundamentally different responses depending on user personas and software context. This gap creates an urgent need for automated acceptance testing frameworks that autonomously interpret user instructions, while reliably inferring user intentions in a changing environment. In this paper, we present an automated acceptance testing framework for LBS with calibrated verdict reliability via two technical contributions. First, we introduce Requirements-Augmented Generation (REAG), which interprets user intentions by retrieving relevant software requirements, domain knowledge, and personas via adaptive RAG and self-reasoning to generate context-aware test oracles. Second, recognising that oracle generation may retrieve irrelevant constraints, misinterpret intent, or hallucinate requirements, we introduce a confidence-calibrated cascade judgment. This method quantifies verdict reliability via simulated expert agreement -- accepting high-confidence verdicts, escalating ambiguous cases, or abstaining when uncertain -- with empirical reliability guarantees backed by conformal risk control. An industrial case study on a production nutrition advisory application demonstrates that REAG achieves a 3.91/5 oracle quality score, reaching qualified or marginal oracle quality in 82% of cases. The confidence-calibrated cascade achieves 98.8% accuracy, improves oracle quality from 3.91 to 4.30 by filtering unqualified outputs, and delivers a 31.7% cost-efficiency improvement over single-judge baselines, validating industrial viability

Vertical Federated Learning (VFL) enables organizations holding complementary features of shared entities to collaborate and train models. In this setting, the initiator can withhold information about the learning task, while other contributors participate without exposing their local datasets, creating an asymmetric information structure aligned with growing privacy demands. However, this asymmetry is a double-edged sword. Among various threats, backdoor attacks are particularly concerning because VFL not only enables malicious contributors to poison the model during training, but also allows them to activate the backdoor at inference time to manipulate predictions. Although prior work has reported near-perfect attack success rates and proposed effective defenses, we find that most findings fail to hold under realistic conditions, exposing a fundamental gap between research and practice. In this paper, we present a systematic, practice-oriented study of backdoor vulnerabilities in VFL, revealing this gap in both methodological design and evaluation practices. We show that existing approaches overlook key practical constraints and therefore rely on unrealistic prior knowledge. Furthermore, these limitations have remained hidden due to poorly designed evaluation practices in the literature. To bridge this gap, we redefine threat models under realistic constraints, propose practical backdoor workflows, and introduce BVBench, a backdoor-centric benchmark that enables fair, practical, and comprehensive evaluation, preloaded with state-of-the-art baselines. BVBench provides strong evidence of the fragility of the current understanding of VFL backdoor risks and establishes a foundation for steering research toward uncovering practical vulnerabilities and developing more meaningful defenses.

The OWL 2 EL profile is used in some of the largest production ontologies, including the Gene Ontology and SNOMED CT. Existing neuro-symbolic (NeSy) learning methods accept propositional theories or Datalog, and reasoning-shortcut (RS) awareness has not been investigated in ontology settings. We present Moose, a method that compiles an $\mathcal{EL}^{++}$ TBox and finite ABox to a Sentential Decision Diagram (SDD). The SDD acts as a differentiable weighted-model-counting layer, and we add closure clauses outside the $\mathcal{EL}^{++}$ profile on declared exhaustive families to overcome the limited expressivity of $\mathcal{EL}^{++}$ under partial supervision. We show termination, soundness, completeness, and polynomial intermediate sizes, and validate the proofs in Lean. We then define the first formal partial-supervision latent-concept-learning task over an OWL EL ontology, i.e., learning per-individual classifiers for latent concepts from observed ABox literals, and evaluate Moose on MNIST-with-ontology and Pizzaïolo. Moose improves over propositional-NeSy, fuzzy-logic, and ontology embedding baselines, and presents the first reasoning-shortcut analysis in an OWL EL setting.

Latent world models are judged by how well they predict, so when planning fails at long horizons the natural reading is that the predictor degrades. On a reproduction of LeWorldModel on TwoRoom we show the binding constraint is the planner's objective instead. The predictor is not the limit: its imagined state seventy-five environment steps ahead is still only 0.189 as wrong as assuming the world froze, while the planner never imagines beyond twenty-five. The objective is. Cross-entropy-method planning minimises squared latent distance, which tracks true distance at r = 0.426, saturates by about eighty arena units and decreases beyond a hundred and twenty, so moving away from the goal can lower the cost. The information is present throughout: a ridge probe recovers position from the frozen embedding at R^2 0.9922. The pathology is the method's, not one reimplementation's. It is present in the authors' released weights, and across four checkpoints long-horizon success rank-orders exactly with metric quality and inversely with prediction accuracy. Replacing only the objective, with nothing retrained and no GPU, lifts goals reached at offset 100 from 26.0% to 98.0%, equals the 98.0% at offset 25, and reaches 92.0% under a third of the budget: planning stops depending on the horizon. The best cost is not the most accurate. A head learned from frame separation alone predicts spatial distance worse than a position probe (r = 0.819 against 0.9897) yet plans better, charging 24% more to cross the environment's dividing wall where squared latent distance charges 4% less. It has learned reachability, not proximity.

Group Relative Policy Optimization (GRPO) learns from reward differences within a rollout group, but receives no useful relative signal when every sampled response is incorrect. Privileged self-distillation can fill this gap with dense token supervision, yet applying it throughout training creates a different failure mode: the teacher is a biased, low-variance surrogate for the reward objective, so persistent imitation can oppose reward-improving updates after the policy becomes capable of producing successful trajectories. We introduce I-SDPO (Instance-Level Adaptive Self-Distillation Policy Optimization), which treats teacher reliance as capability-dependent. I-SDPO makes one routing decision per input instance and shares it across that instance's rollout group: all-incorrect groups use a privileged self-distillation objective, whereas any-success groups remain intact for GRPO. This design uses imitation only where group-relative rewards are uninformative. A local analysis characterizes when teacher and reward directions align and shows that a non-vanishing biased distillation weight induces an optimization bias floor. The routing rule automatically reduces the expected distillation rate as success probability rises, withdrawing teacher influence without a hand-designed schedule. On SciKnowEval, I-SDPO obtains the best result in all four scientific domains and improves average mean@16 accuracy from 56.67% with GRPO to 70.31%, with a maximum domain gain of 18.24 points.

Structured pruning is a promising approach for compressing large language models (LLMs), yet existing methods rely heavily on greedy heuristics that produce myopic decisions, and often fail to precisely meet target compression budgets. We present SNIPER, a two-stage structured pruning framework that solves a knapsack optimization over coarse-granularity components to yield conditionally optimal parameter allocations with respect to fixed importance estimates, followed by a fine-grained pruning stage to meet strict budget constraints. We introduce the Compression Ratio Adherence Factor (CRAFT) to quantify budget fidelity, showing that while existing pruners deviate from target compression ratios by up to 33%, SNIPER achieves near-exact adherence with a CRAFT score of 0.98. Evaluations across four diverse architectures over a set of 18 tasks spanning five domains demonstrate SNIPER's consistent improvements in average performance retention and task-level stability over six state-of-the-art pruners. Across all pruning configurations, SNIPER achieves an excellent mean rank of 1.25, indicating its robust cross-architectural generalizability and excellent reliability.

Electrocardiography (ECG), photoplethysmography (PPG), and phonocardiography (PCG) provide complementary views of the same cardiac cycle, yet existing cardiac foundation models are trained for a single sensing modality, leaving the shared physiology across sensors unexploited. We introduce CardioState-JEPA, a cardiac foundation model to learn a single shared representation jointly across ECG, PPG, and PCG, built on a physiology-aware joint-embedding predictive architecture. The model maps heterogeneous waveforms into a common token space, processes them with a single shared Transformer encoder, and learns by predicting masked latent cardiac states, placing the pretraining target on shared physiology rather than sensor-specific waveform appearance. To handle the temporal offsets between electrical, mechanical, and hemodynamic events, cross-modal prediction uses a learned delay aligner that matches signals at the corresponding cardiac time. Because synchronized multi-sensor recordings are scarce, CardioState-JEPA first learns within-modality structure from abundant unimodal data and then uses paired data to align modalities in latent cardiac time. Evaluated as a frozen encoder across 25 downstream tasks spanning ECG, PPG, and PCG, our encoder improves average PPG classification by 8.2 AUROC points, PCG murmur detection by 18.8 AUROC points, and ECG classification by 15.5 AUROC points over the best self-supervised signal baseline and matches or exceeds cardiac models trained with privileged clinical text or supervised labels on several ECG benchmarks. These results establish that heterogeneous cardiac signals can mutually supervise a single foundation model of cardiac physiology.

Joint-embedding predictive architectures (JEPAs) learn world models that predict in a compact latent space rather than in pixels, reducing the pressure to model nuisance appearance. Yet this provides no guarantee against visual perturbations: they can still alter the encoded representation and affect subsequent action-conditioned predictions. Bisimulation captures this requirement precisely: two observations should be treated as the same state only when their action-conditioned consequences agree. Guided by this criterion, we introduce Action-Conditioned Predictive Consistency (ACPC), a diagnostic that measures how far a clean history and a visually perturbed view of it diverge after being rolled forward under the same action sequence. We prove that this divergence bounds the perturbation-induced change in multi-step prediction error and planner cost. Building on pairwise ACPC, we define two complementary measures: the Invariance Radius (IR) summarizes clean-perturbed rollout spread, while the Separation Rate (SR) checks whether different states remain distinguishable after rollout. Experiments on four visual control tasks show that pairwise ACPC predicts perturbation-induced prediction and cost changes. On LeWM, the IR-SR screen transfers across tasks, and the joint diagnostic remains informative under blur and resize. PLDM exhibits similar diagnostic trends under a different architecture.

As quantum computing progresses from proof-of-principle demonstrations toward practical utility, a significant impediment is the need to augment algorithmic feasibility with system-level optimization across heterogeneous hardware and software stacks. Quantum resource estimation (QRE) plays a central role in this transition, yet existing approaches remain largely compilation-heavy or domain-knowledge-guided symbolic annotations, and tightly coupled to long-term fault-tolerant assumptions, limiting their topical applicability. In this work, we introduce AutoQuREO, an Automated framework for full-stack Quantum Resource Estimation and Optimization. AutoQuREO is built around four core novelties: (i) a flexible, user-defined abstraction of the quantum computing stack; (ii) a modular library of reusable stack components enabling rapid full-stack prototyping; (iii) surrogate modeling of layer-wise resources via algorithmic profiling and neuro-symbolic learning; and (iv) integrated multi-objective optimization that embeds QRE directly into deployment pipelines. Together, these design choices enable AutoQuREO to serve as a digital twin for quantum computing stacks, supporting the tractable exploration of complex design spaces. We demonstrate the capabilities of AutoQuREO through representative co-design case studies, including early-fault-tolerant quantum algorithms, small error correction codes, gate decomposition and variational training of parametric quantum circuits. These examples illustrate how AutoQuREO enables systematic discovery of unexploited resource trade-offs that are computationally intractable or abstruse using existing QRE tools. AutoQuREO is positioned as a general-purpose platform for advancing quantum technology readiness.

Perturbation methods explain model decisions by measuring prediction changes under altered inputs, but response magnitude tells us only how much a model reacts, not what that reaction means. The same magnitude can support the final factual-counterfactual difference, oppose it, or arise strongly along the perturbation path yet vanish at the endpoint. We therefore track how the contrast develops as paired inputs are progressively revealed, using the final contrast to interpret the trajectory. We introduce DECAF (Decomposition of Evidence, Contradiction, And Fragility), which routes aligned, opposed, and endpoint-null responses into evidence E, contradiction C, and fragility F. The decomposition preserves ordinary magnitude exactly, Abs = E + C + F, and is unique under endpoint-relative axioms. Across controlled vision and tabular settings, the three components track independently measured behavior. In a 72-model ImageNet-9 audit, we compare cases with nearly identical response magnitude but different independently measured behaviors. The largest DECAF component agrees with an observed behavior in 96.4% of cases, compared with 35.0% for magnitude alone. Changing only the reveal path increases total response by nearly 80%, yet evidence barely changes while fragility grows by more than 4x. On FunnyBirds and ImageNet-1k, short forward-only DECAF trajectories outperform the tested general-purpose attribution baselines. On a 1B-scale DINOv2 model, a short trajectory matches a strong gradient-based baseline with 4.75x lower wall time and 2.36x lower peak memory.

Vision-Language-Action (VLA) models promise to bring end-to-end reasoning to autonomous driving, but their computational cost remains far too high for real-time control. The core challenge is structural: VLA inference is not a single bottleneck but a cascade of four. Visual encoding wastes compute on overlapping video frames; language-model prefill recomputes context that could be carried over from the previous timestep; reasoning tokens are generated serially despite low entropy; and flow-matching denoising applies uniform compute to a non-uniform velocity field. Addressing any one stage in isolation leaves the others untouched. We propose FlashDrive, an algorithm-system co-design framework that targets all four stages simultaneously. Our key insight is that each bottleneck admits a distinct, lightweight algorithmic shortcut: temporal overlap enables streaming KV-cache reuse across frames; the low per-token entropy and strong intra-block correlations of driving-domain reasoning make a non-autoregressive diffusion drafter highly effective for speculative decoding; and the velocity field's structure---sharp at the endpoints, flat in the middle---permits adaptive step caching that concentrates compute where it matters. Layered on system-level CUDA Graph compilation and kernel fusion, these techniques compound. Applied to Alpamayo 1.5-10B with W4A8 quantization, FlashDrive reduces end-to-end latency from 717ms to 151ms (4.7x) while leaving accuracy essentially unchanged: minADE6@6.4s shifts by only 0.08m, minADE1 improves, and closed-loop collision and off-road rates improve in simulation. By raising a 10B-parameter reasoning VLA from 1.4~Hz to 6.6~Hz on a single GPU, FlashDrive moves end-to-end autonomous driving substantially closer to real-time deployment.

The study presents a systematic machine learning (ML) study of 6G-IoT beamforming optimization (6GBO) using supervised and unsupervised approaches. We compared the predictive power of network, environmental, device, and vision feature groups for 6GBO. Additionally, it addressed other unsupervised perspectives that can enhance 6GBO, including clustering network scenarios using methods such as K-means, DBSCAN, and hierarchical clustering. Several imbalance-aware experiments revealed that network features possess better prediction power than device, environmental, and vision feature groups, as evidenced by their recall, F1-score and ROC-AUC values. For unsupervised ML exploration (assessed using Elbow, Silhouette score, and Davies-Bouldin Index methods), the results indicate that the deployment environment and type of device primarily influence clustering, rather than mobility-based attributes. Furthermore, the explainability analysis showed that bandwidth, IoT sensors, and mobility possess higher global feature importance across the feature groups. In the future, we would apply deep and reinforcement learning techniques to predict throughput/latency or to optimize rewards determined by performance indicators like SNR enhancement

We introduce a Polish-language medical visual question answering (VQA) benchmark, built from Polish Board Certification Examination questions for licensed physicians and dentists pursuing specialist certification. The benchmark comprises image-containing questions spanning diverse medical specialties and visual domains, together with a text-only question answering (QA) control set. We evaluate Polish-oriented, general-purpose open-weight, and commercial vision-language models. The task remains challenging: the best model achieves 79.0\% accuracy on the full VQA set, and only GPT-5.6 surpasses the approximate human reference on the subset with available candidate responses; all other evaluated models perform worse than humans. To assess visual grounding, we compare complete inputs with configurations omitting the image, the question, or both, and categorize questions by image importance. Models derive more useful information from the question text than from the image and perform worse on image-dominant questions. Across both QA and VQA, they nevertheless achieve above-chance accuracy from the answer choices alone, showing that non-trivial performance can persist even when key task components are missing.

Traditional player evaluation in professional handball relies on basic box-score metrics or heuristic indices, which fail to credit the multi-player build-up chain. While football (soccer) analytics has adopted Expected Threat (xT) and Valuing Actions by Estimating Probabilities (VAEP), these event-based action valuation frameworks have not yet been adapted to handball. In this paper, we present the first comprehensive adaptation and evaluation of xT and VAEP for handball, utilizing five seasons of tracking-derived event data from the Handball Bundesliga. We develop Handball-xT (H-xT) using a handball-native court zoning layout, demonstrating via simulations that it is systematically more robust than standard rectangular grids. We optimize Handball-VAEP (H-VAEP) by tailoring its feature space and selecting the context length to limit team-identity leakage. Our evaluation shows that H-VAEP yields exceptionally stable, discriminative, and intuitive player ratings that highlight build-up play. Finally, we release our complete code repository to help professional clubs deploy these models.

Momentum-based optimizers are widely used in modern deep learning, yet the relations among momentum recursion, update geometry, and acceleration remain only partially understood. We develop an $\textbf{A}$DMM-$\textbf{I}$nspired $\textbf{M}$omentum (AIM) framework based on residual-penalty variable splitting, which interprets momentum as a multiplier-like correction driven by the splitting residual. AIM recovers the exponential moving average of gradients from an ADMM-style multiplier update and separates two mechanisms that are usually intertwined in practical optimizers: the residual penalty determines the update geometry, whereas the approximation of the objective-related subproblem determines the acceleration form. Building on AIM, we propose $\textbf{R}$elativistic $\textbf{A}$daptive gradient $\textbf{D}$escent with $\textbf{A}$ccelerated $\textbf{R}$esidual (RADAR), which combines relativistic adaptive geometry, decoupled residual correction, and second-order momentum filtering to improve the update direction and momentum estimation. We establish stochastic convergence through a variance-perturbed Lyapunov drift analysis. Experiments on supervised vision learning, language modeling, and reinforcement learning show that RADAR achieves consistent improvements over strong adaptive optimizer baselines.

The performance of large language model (LLM)-based multi-agent systems (MAS) largely depends on effective communication topologies. Existing topology generation methods, however, typically learn communication topologies through black-box optimization driven solely by task-level rewards. While effective, such optimization provides little insight into why particular communication edges are selected, making it difficult to identify the critical communication subgraphs responsible for successful collaboration. To address this limitation, we propose E2-Explainer, a model-agnostic framework for providing interpretable explanations of communication topologies produced by arbitrary topology generators. Specifically, we formulate topology explanation as a causal attribution problem that identifies compact communication subgraphs supported by edge-level evidence of task preservation. We obtain this evidence with a Granger-style objective that measures how masking each communication channel changes the task outcome and the stability of the final response. The resulting budgeted subgraphs are then distilled into an amortized explainer, enabling efficient post-hoc explanation without repeated edge-level evaluations at deployment. Extensive experiments on multiple reasoning and coding benchmarks demonstrate that E2-Explainer identifies critical communication subgraphs that preserve successful collaboration. These subgraphs can also be executed directly to prune redundant communication edges, substantially reducing communication costs while maintaining competitive task performance.

Developing effective robot navigation methods in crowded environments is essential for real-world applications. Although recent deep reinforcement learning (DRL) methods have improved navigation performance in crowded environments, they often focus primarily on task-centric objectives and underrepresent social compliance objectives. In this paper, we introduce a novel proxemics-based reward formulation for DRL social navigation that provides a dense, interpretable social learning signal while maintaining navigation efficiency. Our approach models each human's personal space as a radial Gaussian-mixture field derived from Hall's proxemics theory and computes a robot-centric local cost over the robot's field of view. We integrate the proposed reward into established DRL navigation methods and evaluate it in simulation across multiple crowd scenarios, reward baselines, and crowd densities using both navigation metrics and social metrics. Results show that the proposed reward consistently improves social metrics in simulation while maintaining competitive navigation performance relative to the compared reward models.

The rapid growth of LLM inference is shifting sustainability concerns from one-time training to continuous serving, where infrastructure decisions shape energy use, carbon emissions, water consumption, and service quality. Yet operators often need to compare deployment alternatives before large-scale infrastructure is built, making direct measurement costly, slow, and sometimes infeasible. We present InFactPlanner, a trace-driven decision-support framework for what-if analysis of sustainable AI data center deployment for LLM inference across single and geo-distributed sites. InFactPlanner combines query traces, hardware-model profiles, candidate site configurations, PUE/WUE parameters, renewable generation models, and time-varying grid carbon intensity to estimate power, energy, carbon emissions, water use, latency, and server utilization. The framework abstracts low-level serving effects into configurable hardware-model profiles, enabling rapid comparison of site selection, capacity placement, hardware, model, renewable integration, and routing choices. We validate the energy accounting pipeline by reproducing reference LLM inference energy estimates with less than 10% deviation, evaluate scalability across multiple data centers and server counts, and demonstrate scenario-driven decision analyses for hardware selection, renewable placement, geographic deployment, and carbon-aware routing. Our results show that sustainability-optimal choices can differ from latency-optimal ones, and that the carbon value of deployment depends strongly on the local grid mix.

Contrastive Decoding (CD) improves generation quality, but its amateur-model pass makes decoding expensive. Accelerating CD with speculative decoding raises a proposal-alignment question: should the contrastive signal shape the drafter, or should it remain only in verification? We study this question in the lightweight feature-level drafter regime. Two controlled diagnostics, matched Cross-alpha training and an Approximate Dual-Drafter decomposition, give the same diagnosis: contrastive-aware drafting does not consistently improve over expert-aligned drafting because the contrastive correction is usually weaker than drafter error, and reconstruction can amplify that error. We introduce Decoupled Contrastive Decoding (DCD), which drafts with an expert-aligned lightweight proposer and applies the amateur only in unchanged CD verification. Standard speculative verification preserves the vanilla-CD output distribution. Across the main 8B settings, EAGLE3-based DCD achieves average greedy speedups of 1.65 to 1.95x over vanilla CD and reduces MMLU proposal-path latency by about 5 to 12x relative to amateur-coupled proposal paths.

This paper considers the overestimation bias problem of Q-learning in the setting of a large action space, for the purpose of relieving the bottleneck of existing methods. We find that the large action space increases the randomness in Q-value estimation. The randomness makes two paradigms that drive the major literature on the overestimation problem have their own bottlenecks: the coupling paradigm, i.e., the optimal action and its Q-value are estimated with the same Q-function, always has a positive bias. This is because randomness leads to some actions having abnormally high estimated values than their true values, and the coupling methods prefer these actions. The decoupling paradigm, i.e., the optimal action and its Q-value are estimated with two independent Q-functions, always has a negative bias. This is because randomness increases the estimation gap between the two independent Q-tables for the same action. This paper shows that action intersection can be a simple yet powerful strategy to relieve these bottlenecks. The action intersection strategy enables semi-decoupling via two designs: (1) it allows two Q-functions to share a certain fraction of trajectory data; (2) if a data sample is shared, each Q-function is updated using the coupling paradigm; otherwise, using the decoupling paradigm. Two properties make the action intersection strategy powerful: (1) attaining a large bias range, i.e., varying the data sharing fraction, the estimation bias varies from underestimating to overestimating; (2) fine granularity: the action intersection size can be made arbitrarily finer to enable finer control. We consider two experiment settings, i.e., tabular and deep RL, deep RL experiments show that our method outperforms several SOTA baselines drastically; tabular experiments reveal why our method can achieve superior performance.

RNA-Protein Interactions (RPIs) are critical for regulating cellular functions. While traditional wet-lab experiments for RPI detection are costly and time-consuming, Deep Learning (DL) methods provide an efficient computational alternative for RPI Prediction (RPIP). In particular, Graph Neural Networks (GNNs) are promising, as they naturally model RPI networks. However, existing GNN-based methods often rely on homogeneous graphs or predefined meta-paths, which limit their ability to handle data sparsity and to generalize to cold-start scenarios involving unknown molecules. To address these limitations, we propose Edge Generation-guided Relation-aware Learning (EGRL), a novel framework with several key components: implicit meta-path learning to capture relational semantics without handcrafted paths; a multi-relation-aware attention mechanism for adaptive fusion of interaction patterns; a graph generator that predicts potential ("soft") edges to support cold-start nodes; and a multi-feature fusion predictor for final interaction scoring. EGRL is jointly trained with a primary task loss and an auxiliary generator loss. Comprehensive evaluations on four benchmark datasets demonstrate that EGRL achieves competitive overall performance. More importantly, it exhibits superior generalization in cold-start settings, achieving an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.867 and an Area Under the Precision-Recall curve (AUPR) of 0.861 on unknown molecules, corresponding to improvements of 8.6% in AUROC and 5.0% in AUPR over prior state-of-the-art methods. The code will be released soon.

The behavior of contemporary generative Large Language Models (LLMs) is directly shaped by prompts, unstructured texts that describe the desired output and model behavior. In this paper we argue that prompts are linguistic objects that merit investigation in their own right. To this end, we collect 57.5K unique samples of prompts from GitHub. Specifically, we focus on transactional prompts: reproducible natural language instructions that are integrated into software. To enable the empirical, quantitative study of prompts, we introduce a structured ontology, capturing the properties of prompts as well as their formal and semantic components. Based on this ontology, we transform prompts from unstructured raw texts into richly structured linguistic objects. Analysis of these structured data reveals significant diversity of usage patterns across languages, domains, tasks, and modalities, in a typical Zipf-like distribution where some clearly prevail and others, more diverse, appear in the long tail. To validate the reliability of the ontology-based annotation of the prompts, we perform a comprehensive error analysis across all fields, providing a detailed assessment of annotation quality. We release the dataset together with a browsing and exploration interface (https://github.com/OnlpLab/transactionalPromptsCollection ).

The $k$-Nearest Neighbor~(KNN) algorithm is widely used across various tasks. The selection of the $k$ value is a key issue because it significantly impacts performance. In this paper, an adaptive and efficient KNN approach via granular-ball computing is proposed. The method consists of two stages. \textcolor{black}{In the training stage, the dataset is first coarsely partitioned to reduce the complexity of data distributions within a granular ball, and then the Fisher criterion is introduced to control ball splitting and stopping, yielding a multi-granularity granular ball representation. In the prediction stage, the nearest granular ball is first located through a weighted distance mechanism, and an adaptive neighborhood is then constructed around the test sample. The effective $k$ value is dynamically determined by the actual number of samples contained in this neighborhood. The neighborhood induced by the nearest granular ball provides more stable local group information, thereby improving robustness against noise and local perturbations.} Experimental results demonstrate that the proposed method outperforms existing KNN variants across multiple datasets in terms of both accuracy and efficiency. The code has been open-sourced for reproducibility: https://github.com/lianxiaoyu724/Adaptive-GBKNN.

Document parsing aims to transform unstructured documents into structured and machine-readable representations. Recent advances in Vision-Language Models (VLMs) have significantly advanced document parsing. However, existing approaches still face two major challenges. First, decoupled VLM-based methods heavily rely on accurate layout analysis, where geometric distortions in camera-captured documents can introduce cascading errors. Second, although end-to-end VLM-based methods alleviate the dependence on explicit layout detection, they often suffer from redundant generation, hallucinations, and insufficient structural reasoning in high-resolution scenarios. To address these challenges, we propose NaviDC-OCR, a unified framework for document parsing. NaviDC-OCR introduces deformation-aware learning to incorporate geometric perception into VLMs and proposes an adaptive sampling mechanism for complex layout representation. Furthermore, a content-structure decoupled learning strategy is developed to explicitly model formula grammars and table structures, enabling more effective structured representation learning. Extensive experiments demonstrate that NaviDC-OCR achieves state-of-the-art performance across diverse document parsing benchmarks. It obtains overall scores of 96.87, 88.53 and 78.41 on OmniDocBench v1.6, Wild-OmniDocBench, and PureDocBench, respectively, and ranks first in the ICDAR 2026 Sci-ImageMiner Challenge. These results validate the effectiveness and generalization capability of NaviDC-OCR in complex document parsing scenarios.

Compositional reliability bounds for multi-agent systems multiply component reliabilities, a step licensed by a conditional-independence assumption that is routinely stated and rarely tested. We test it. Two instances of one model, in a two-agent handoff, co-fail on 90.0% of the missions on which either fails (log OR 6.66, 95% CI [6.38, 7.00]; phi 0.916), in a preregistered evaluation of 18,000 missions scored by deterministic code with no LLM judge. Substituting a different model reduces the association in six of six contrasts; substituting a different vendor, model already different, does not -- a registered hypothesis reported as a null. The error is signed and runs against the operator: positive dependence inflates joint failure above the independence product, so redundancy is over-credited exactly when components share a model. The assumption-free alternative is often vacuous, and fitting a dependence model is worse: we prove a bootstrap bound on a fitted model's functional loses coverage of the truth as n grows, the identification gap being O(1) while the bootstrap haircut is O(n^{-1/2}). More data makes such a certificate worse, with no visible symptom. We give a finite-sample certificate assuming no dependence structure: a linear program over the joint, over a Bonferroni-Clopper-Pearson box around measured co-execution moments. It is sound, sharp for the information supplied, and monotone in the moment family. Enriching ten moment functionals to fourteen narrows the identified interval by 85.7% and lifts the certified floor from 0.2455 to 0.4116. A companion anytime-valid certificate holds type-I error at 0.0471 under optional stopping. Common dependence statistics are marginal-bounded and can reverse an apparent ordering of conditions when the compared agents fail at different rates. Contracts, scoring code, analysis scripts, and the preregistration are released.

Cultural evaluation of large language models (LLMs) often focuses on high-resource standard languages, leaving regional culture and dialect communities underrepresented. We introduce BavGround, a benchmark for evaluating Bavarian regional cultural grounding and dialect competence across English, German and Bavarian. BavGround contains 206 multiple-choice source questions across eight cultural domains per language, yielding 618 multi-parallel instances, with items covering both broadly accessible cultural knowledge and source-grounded regional knowledge from journalism, historical sources, and specialist literature. We evaluate fifteen 7B-10B open-weight instruction-tuned models and one closed-model reference. Strong multilingual models perform best overall, but performance drops on Bavarian items and source-grounded questions, indicating persistent difficulty with dialectal and localized cultural knowledge. We further show that conclusions depend strongly on evaluation protocol: raw answer-letter scoring, shuffled-letter scoring, option-text likelihood, generated-answer parsing, and semantic matching can produce different absolute scores and rankings, especially for regionally adapted models. Finally, an exploratory analysis of GENBA-10B checkpoints suggests that continued pretraining improves answer-content likelihood unevenly across domains, while dialect competence remains comparatively weak. BavGround supports localized, protocol-aware evaluation of cultural representation in LLMs.

Activation steering turns localized representations into control directions, but localization alone does not reveal whether a direction has a selective operating regime. We introduce Predictive Memory Localization (PML), which treats the measured-grid intervention path as the predictive object of memory localization. PML separates random-calibrated target movement from semantic-neighbor and capability damage, and compares static localization and supervised geometry with a strength-disjoint low-dose causal response. Our frozen study covers 3,000 records from nine datasets and fourteen domains, yielding 30,000 distinct record-direction-layer paths and 210,000 distinct path-strength evaluations. At layer 7, the geometry-derived RFM/AGOP direction reaches 13.1% target-any and 12.3% clean-any, exceeding random by 3.6 and 3.4 percentage points under a record-paired bootstrap. Across record-, dataset-, and domain-grouped splits, responses at $|α|=0.1$ are the strongest signal for outcomes at disjoint strengths $|α|\in\{0.25,0.5\}$. On held-out records, a predictor-driven selector chooses a coefficient or abstains, improves utility and reduces semantic-neighbor damage relative to a train-tuned fixed-strength policy, and avoids most evaluations in a dense scan. Across three residual-norm-matched base models, learned directions retain selective-path gains and low-dose responses yield 0.801-0.828 record-held-out macro AUROC. PML therefore turns memory localization into a falsifiable forecast of margin-level selective outcomes and a risk-aware intervention decision.

Agent-memory systems increasingly buy retrieval quality with structure, transforming raw conversation histories into summaries, embeddings, trees, or knowledge graphs before any question is asked. We ask how much of that benefit comes from the structure itself, rather than from competent retrieval over the raw history. We present ReFind, an agent-controlled search interface that builds no semantic structure at all: it leaves the conversation archive unmodified, indexes it lexically at turn granularity, and combines a generic iterative keyword-search loop with four chat-native controls grounded in empirical refinding work: session-aware rank fusion, local context expansion, temporal narrowing, and skipping already-inspected sessions. A separate reasoning stage answers from the collected evidence. Across a broad suite of conversational-memory tasks (single- and multi-hop QA, event ordering, and fact consolidation), roughly 2,800 questions on precise-retrieval and fact-tracking capabilities evaluated under the incremental multi-turn setting of MemoryAgentBench, ReFind attains the highest mean accuracy (58.2) of any system compared, above the strongest graph- and tree-based memory systems (HippoRAG 2, 53.2), all under a GPT-4o-mini backbone matched to every reused baseline. Controlled comparisons to single-shot BM25, a matched generic-agentic BM25 control, component removals, and agentic dense/hybrid variants separately support the roles of agent control, chat-native controls, and lexical retrieval. On LongMemEval-S/M, the same interface reaches 93.2 +/- 3.3 and 89.3 +/- 6.0 with GPT-5-mini. The results indicate that for precise, evidence-grounded questions over chat archives, much of the benefit credited to elaborate memory structures is recoverable by giving an agent controllable search over the unmodified record, with no LLM-based index construction at all.

Security evaluations of tool-using agents often equate stored labels with behavioral facts. We audit a preserved campaign by tracing 10,200 execution rows to 180 model-bound requests, 45 semantic requests, and 15 observable stimuli. Two schema treatments were delivered, but the planned external payload-family corpus was not. The historical grader exhibited direct treatment leakage: treatment metadata gated the ATTACK_SUCCESS class, so fixed behavior could change class under treatment relabeling. A treatment-blind reconstruction corrects 58 historical ATTACK_SUCCESS or HIJACK_ATTEMPT labels to authorized benign completions while preserving three verified protected-data transfers and one separate unauthorized-forwarding case. The locked v2 census contains exactly zero ATTACK_SUCCESS records, while the forwarding case remains a HIJACK_ATTEMPT at a semantic boundary concerning objective completion. A dual-reviewer blinded concordance review of all 96 requests deemed structurally interpretable by locked v2 produced identical reviewer-consensus classes but differed from the locked codebook on four construct-boundary cases. We contribute a seven-link Integrity Chain and an executable, scope-bounded endpoint-integrity linter. The result is a campaign-bounded measurement audit, not a population attack-rate, model-ranking, defense-efficacy, or causal estimate.

Fractional partial differential equations describe nonlocal dynamics, but discovering them from noisy data is difficult because fractional differentiation amplifies high-frequency measurement noise and the derivative orders are unknown. We propose Weak-Pareto, which combines an adjoint-consistent weak formulation of fractional terms with Pareto-based subset selection over discrete term types and continuous fractional orders. For linear right-hand-side terms, the adjoint transfers fractional operators from measured fields to smooth test functions, replacing noise-sensitive pointwise differentiation with smoothing integration; for nonlinear terms, the noise-suppression effect is partial yet useful. Coefficients are fitted by ridge regression within a branch-aware differential-evolution search over the orders. The support size is then selected at the validation-error-complexity elbow. We show that the variance of fixed linear right-hand-side weak features vanishes under grid refinement, whereas noise amplification in pointwise fractional features increases with derivative order. Across fractional advection-diffusion, reaction-diffusion, and Burgers benchmarks, Weak-Pareto recovers parsimonious structures from clean and noisy measurements. In controlled advection-diffusion and Burgers comparisons, it retains the correct support at every tested multiplicative-noise level, whereas the unregularised strong-form counterpart largely fails once noise is introduced; this advantage persists under additive Gaussian noise. Ablations show that the weak library drives noise robustness and that continuous-order Pareto search avoids the support-selection failure of a dense fixed dictionary. On the advection-diffusion benchmark, Weak-Pareto yields more consistent operator recovery and substantially lower measured runtime than a contemporary neural baseline.

Multi-hop fact verification, which verifies claims by reasoning over multiple pieces of evidence, is critical for combating misinformation on social media yet remains highly challenging. Recent methods primarily rely on multi-agent collaboration to decompose fact verification into specialized subtasks. However, these methods face two critical limitations: (1) agents may perform individual subtasks without sufficient awareness of the global verification objective, causing their reasoning to deviate from the intended direction; and (2) conflicts between parametric knowledge and the provided evidence may undermine evidence-grounded reasoning and lead to incorrect verdicts. To address these challenges, we propose ReflectFact, a novel self-reflective agent framework for multi-hop fact verification. ReflectFact introduces three key tasks. Explicit Reasoning Path Planning builds an evidence-grounded reasoning path by resolving implicit entities, decomposing the claim into sub-questions, and integrating the verified facts into a verdict. Evidence-Drift Verification makes the agent re-answer by quoting the supporting evidence when a grounded answer merely echoes its parametric prior, thereby calibrating evidence deviation to ensure grounded comprehension. Reasoning Reflection Verification re-examines each reasoning step and regenerates it once an inconsistency is detected, correcting reasoning flaws such as location bias and replacement bias through a global task perspective. Subsequently, the agent aggregates validated reasoning chains to yield reliable verdicts. Extensive experiments on HOVER and EX-FEVER demonstrate that ReflectFact effectively remedies the comprehension and reasoning defects of existing methods, achieving state-of-the-art performance and respectively outperforming the strongest baseline by 3.32\% and 2.78\% on the two datasets.

Detecting AI-generated images is only half the task: a deployed detector must also justify its verdict, yet existing detectors inherit three failure modes from their training data: real and fake images collected from different sources invite provenance shortcuts, supervised explanation corpora teach templated rationales, and a static forgery corpus leaves the decision boundary standing still while generators keep moving. We introduce \methodname{}, an adversarial reinforcement learning framework that pits two heterogeneous models against each other. A diffusion image editor learns to edit real photographs into fake counterparts of those same photographs that fool the current detector, while a reasoning MLLM learns to expose them with a verdict grounded in free-form reasoning. Both rewards are shortcut-proof by design: the attacker is credited only when its edit is faithfully executed, and the defender only when its verdict is correct. As the two models alternate, each round's attacker regenerates a harder training pool aimed at the current detector's blind spots, so the detector must generalize rather than memorize any fixed artifact distribution. Although the explanation is never rewarded, its quality rises round over round as a side effect of accuracy-only training. A detector trained within this loop improves monotonically across rounds on each of three external benchmarks.

Should you replace your text-embedding pipeline with a large language model? We answer this with a controlled, cost-aware comparison of ten LLMs across six families and 26 embedding models (118M to 14B parameters) on 37 tasks spanning classification, semantic textual similarity (STS), clustering, pair classification, and retrieval. In aggregate the two paradigms are effectively tied: the best LLM (Gemini 3.1 Pro, 77.6) and the best embedding model (77.2) differ by 0.4 points. Their strengths differ by task: LLMs lead on reasoning-heavy retrieval, embedding models lead on classification, and the two match on clustering, STS, and pair classification. Reaching that parity is expensive. An LLM costs up to 1,431x more than an embedding model of comparable quality (USD 154 vs. USD 0.11 per benchmark pass), and the open LLMs tested process tokens 2.5 to 736x more slowly on the same GPU. Reasoning tokens account for 28 to 81% of LLM inference cost; lower reasoning budgets preserve or improve retrieval quality for most models in our ablation. The Pareto frontier contains the leading embedding models and one LLM, Gemini 3.1 Pro. These results support a division of labour: use embedding models for similarity, classification, and clustering, and reserve LLMs for reasoning-intensive retrieval. Our code, datasets, and results are publicly available at https://github.com/embeddings-benchmark/embedders-dilemma.

Plasticity loss has emerged as a critical challenge in continual learning that significantly hinders the acquisition of sequential tasks. While optimizing activation designs offers a potential solution, current fixed-form functions suffer from an inherent spectral bias towards low-frequency variations, whereas learnable variants permit unconstrained updates that induce catastrophic forgetting. To address these limitations, we propose a novel learnable wavelet activation that decomposes the activation function into low-frequency and high-frequency components to explicitly counter spectral bias. Furthermore, we employ dynamic wavelet injection to adaptively enhance plasticity for new tasks, alongside a regularization strategy to ensure the stability of previous learned knowledge. Theoretically, we provide rigorous mathematical guarantees for the proposed framework, proving the structural necessity of the hybrid wavelet architecture for efficient $L^2$ approximation and demonstrating that the decoupled learning rate mechanism successfully restores network plasticity for high-frequency information. Additionally, we provide a formal derivation of the loss-driven injection trigger mechanism to precisely guide the injection. Extensive empirical evaluations demonstrate that our approach maintains superior trainability and generalization throughout the learning process and achieves state-of-the-art performance across diverse continual learning benchmarks.

Probabilistic Circuits (PCs) are generative models that support exact inference and, unlike deep neural networks, admit an exact and tractable measure of loss-surface curvature: the trace of the Hessian of the log-likelihood. Recent work regularizes this trace globally to bias learning toward flatter, better generalizing optima. We show that treating sharpness as a global regularizer can be misspecified for PCs, whose curvature is inherently compositional. We prove that each sum node's contribution to the Hessian trace factorizes exactly into its circuit flow, which measures how heavily the node is used, and a local sharpness term determined by its output distribution. This decomposition provides insights into why global sharpness regularization is depth biased and can lead to underfitting. Building on it, we introduce an adaptive sharpness aware regularizer that penalizes nodes based on intrinsic local curvature and preserves closed form EM updates. We also show that empirically, this targeted regularization recovers the generalization that global regularization sacrifices while retaining the robustness and benefits of sharpness aware learning.

Public blockchain data enables large-scale DeFi-related analysis, but many existing approaches are application-specific, difficult to scale, or hard to interpret. This research proposes a scalable, application-agnostic framework for \emph{persistent behavioural pattern discovery} from large-scale blockchain activity. It constructs behaviour sentences enriched with contract, token and market context, then applies a two-step embedding process: sentence-level embeddings capture individual actions, while sequence-level embeddings capture user behaviour over time. An interpretable behavioural profiler characterizes discovered communities through behavioural motifs, routines, temporal dynamics, entity exposure, and suspiciousness evidence. Evaluation on Ethereum using over 30 million transactions shows that the framework uncovers both routine and malicious behavioural patterns, including decentralised exchange (DEX) trading, NFT activity, phishing, bot operations, oracle manipulation, and rug-pull schemes. Importantly, many patterns remain stable across independent observation windows, enabling the identification of long-term behaviours beyond a single analysis period. The proposed framework combines scalability, interpretability, and persistence analysis, supporting blockchain forensic investigation, behavioural attribution, and threat discovery.

As AI systems proliferate in consumer facing applications, questions about liability for AI related harms remain unresolved. This working paper examines whether India's Consumer Protection Act, 2019, adequately addresses harm caused by defective AI products and services, and whether it proportionately allocates liability across the AI value chain. The Act's broad definitions of product liability, harm, and deficiency appear technology agnostic and potentially applicable to AI related incidents including personal injury, psychological harm, biased outputs, and loss of control. However, significant gaps remain. Proving causation between AI defects and consumer harm presents a technical challenge, as AI failures often stem from design choices rather than discrete defects. Additionally, the Act's framework assumes distinct roles for manufacturers, sellers, and service providers, yet the AI value chain involves overlapping responsibilities among data providers, model developers, deployers, and users that do not neatly map to these categories. Current liability frameworks lack proportionate mechanisms to effectively address complex, multistakeholder AI harms. While the Act may cover AI entities, enforcement requires clarification on sector specific overlaps.

Many software systems expose multiple execution modes through command-line options, subcommands, and configuration flags. For such programs, fuzzing depends on both mutated inputs and the invoked mode. Yet evaluations still focus on coverage and bug counts, leaving unclear how execution modes partition, overlap, and miss software structure, and how these differences affect effectiveness. We present an empirical study of software structure under multi-driver fuzzing. We propose a structural abstraction that uses a static call graph as a shared backbone and projects driver-specific dynamic coverage onto it to derive driver-induced subgraphs. Based on this abstraction, we develop a four-phase methodology for backbone construction, fuzzing and profiling, graph-based analysis, and research-question-driven evaluation. We apply it to 27 OSS-Fuzz-derived C/C++ projects, spanning 43 executables and 854 driver configurations. Under the same total budget, multi-driver fuzzing outperforms the best single-driver baseline, increasing covered call-graph nodes by 27.9% and CFG-edge coverage by 73.5%, and revealing 11 unique bugs and abnormal behaviors largely missed by single-driver fuzzing. However, driver contributions are uneven, subgraphs differ substantially in cohesion, fragmentation, modularity, overlap, and residual under-exploration follows recurring regimes rather than a homogeneous tail. These results show that multi-driver fuzzing is fundamentally a structural exploration problem.

Autonomous driving requires planning under both semantic constraints and predictive dynamics. Existing end-to-end driving approaches, however, typically emphasize only one side of this requirement: Vision-Language-Action (VLA) models exploit VLM priors for semantic reasoning, while World Action Models (WAMs) provide future-aware prediction through generative world modeling. This naturally motivates a unified planner that can leverage both semantic priors and predictive dynamics. However, we find that a naive combination through joint token-level attention suffers from an attention-allocation mismatch, where semantic shortcuts dominate the shared attention space and suppress predictive dynamics. Inspired by neuroscience evidence that complex behavior arises from coordination among functionally specialized systems, we propose BrainWAM, a structured action-space coordination framework that converts semantic reasoning and predictive world modeling into two specialized action-oriented pathways, and aligns them at the level of compact action representations. We further introduce an asynchronous rectified-flow inference strategy with decoupled video and action denoising, which shortens inference latency while preserving planning-relevant predictive context. BrainWAM reaches state-of-the-art performance on both NAVSIM v1 (89.5 PDMS) and NAVSIM v2 (89.6 EPDMS), consistently outperforming VLA-only or WAM-only methods, highlighting BrainWAM as a practical and promising direction for autonomous driving systems.

Language can describe states of affairs that are false and states of affairs that could not be the case at all. Whether an AI model internally distinguishes these failures remains unclear. I report an exploratory activation study of the multimodal open-weight model Gemma 3 4B IT using 85 prompts from 17 philosophical families and a topic-matched modality set of 15 topics, each expressed as a truth, contingent falsehood, improbable claim, semantic anomaly, and necessary falsehood. In its answers, the model conflates contingent falsehood with contradiction, labeling 12 of 15 false statements "contradiction." Its activations show a different pattern. A linear truth probe separates impossible from true statements (AUC 0.93) but not impossible from false statements (AUC 0.20). An impossibility probe evaluated on held-out topic families separates necessary from contingent falsehood at AUC 1.00, peaking at layer 15 with balanced accuracy 0.97 (Bonferroni-adjusted P=0.018). The truth and impossibility directions are close to orthogonal, whereas the impossibility direction partially overlaps a semantic anomaly direction while remaining distinguishable from it. Sparse autoencoder features at the same layer repeat this geometry. Features selective for impossibility also fire on anomalous sentences but rarely on contingent falsehoods. In this model's activation space, necessary falsehoods are not extreme cases of contingent falsehood but lie closer to the experimentally defined category of semantic anomaly. This representational proximity does not imply that impossible statements are intrinsically meaningless. These correlational observations from one small model offer an empirical footnote to an old philosophical distinction.

Self-improving LLM agents convert successful trajectories into persistent cross-task state. An unsafe success can thereby become reusable policy after its triggering input disappears. Skill evolution makes this failure measurable by distilling operational trajectories into executable, transferable, and inspectable procedures. Because evolution optimizes task outcomes rather than procedure safety, compromised experience can cause skill misevolution. Existing benchmarks measure current behavior or static artifacts but cannot attribute risk across authoring, retrieval, and later execution. To expose this lifecycle, we introduce SkillMisevo-Gym, a lifecycle-aware harness that versions skill state across agent frameworks, and SkillMisevo-Bench, a frozen design from malicious exposure to carryover tasks, with concept-aligned benign tasks and nine lifecycle metrics. We also introduce SafeEvolve, a wrapper that repairs unsafe content and governs subsequent reuse. Across 25 agent-method configurations, each covering 525 tasks in 25 episodes, all 21 evolved configurations author unsafe artifacts, while only fifteen lead to fresh-session harm. In the exposure sweep, three malicious tasks raise carryover ASR from 16.0% to 35.3%. Across representative skill evolution methods, SafeEvolve reduces unsafe retrieval and fresh-session harm by 26.7 and 17.3 percentage points, respectively, while mean benign utility changes by only 0.4 points. Together, persistent-adaptation safety must govern what updates write and what future executors reuse. Code is available at https://github.com/henrymao2004/misevolve.

Retrieval can identify a past trajectory that may matter, yet it does not specify how an acting agent should use that trajectory after users, entities, constraints, or environment state have changed. We identify this post-retrieval reuse step as a distinct bottleneck for long-horizon trajectory memory and formulate an evaluation framework that holds candidate retrieval, target state, model, decoding, and tool budget fixed while varying the support delivered to the agent. We instantiate the framework with query-conditioned reuse (QCR), a deliberately simple target-bound note that records a reusable procedure, bindings to recover, applicability conditions, and verification requirements. QCR serves to test the reuse hypothesis rather than to claim a universally preferred memory format. Across 2,391 target instances in WebArena, WorkArena, and AppWorld, QCR reaches 62.3% average Success, 10.7 points above Full Trajectory, while using 48.9% fewer online tokens. Summary reranking selects a reusable memory for 94.8% of targets, placing end-task Success within 1.8 points of an oracle reusable selector. Analyses by trajectory length and source--target binding shift show that direct trajectory injection loses much of its utility as traces grow longer or source-specific values change, whereas target-bound support preserves a larger share of the measured gain. The resulting framework separates retrieval quality from the problem of turning retrieved experience into safe, useful support for a new task.

Semantic ID (SID)-based generative recommendation has recently achieved remarkable success. However, existing methods suffer from a previously overlooked fairness issue, which we term \textbf{Token Frequency Bias}, where high-frequency SID tokens are systematically over-predicted while low-frequency SID tokens are under-predicted. This bias originates from the combined effects of imbalanced semantic codebooks during SID construction, and popularity bias together with the maximum likelihood estimation objective during recommendation training, resulting in unfair exposure across item categories. Existing SID methods mainly focus on improving codebook quality and overlook the impact of token frequency imbalance on downstream recommendation fairness, while LLM debiasing methods often yield suboptimal results when directly applied to SID-based recommendation, due to the hierarchical semantics of SID tokens. To address this issue, we propose \textbf{FSGR}, a fairness optimization framework for SID-based generative recommendation. During SID construction, FSGR employs OT-based Assignment Optimization and Dual-Criteria Re-anchor mechanism to form a more balanced SID representation space. During recommendation training, it adopts a two-stage training strategy and introduces Hierarchical Frequency Calibration for layer-specific fairness fine-tuning. Experiments on three public datasets with three backbone models demonstrate that FSGR mitigates token frequency bias and delivers an average Gini fairness improvement of over 20\% while maintaining competitive recommendation accuracy.

Text-based person anomaly retrieval aims to retrieve pedestrians exhibiting anomalous behaviors from a large image gallery using natural language descriptions. Compared with conventional text-based person retrieval, this task requires fine-grained reasoning over pedestrian appearance, behaviors, object interactions, and scene context, making robust cross-modal matching significantly more challenging. This paper presents the GENAI4E team's solution to AI City Challenge 2026 Track 4. Our framework builds upon a strong retrieval backbone and progressively integrates heterogeneous vision-language embedding models through score alignment and iterative ensemble fusion, followed by disagreement-aware VLM reranking for ambiguous queries. On the official Pedestrian Anomaly Behavior (PAB) benchmark, our approach achieves 90.92% mAP, 85.13% Recall@1, 97.72% Recall@5, and 98.68% Recall@10, demonstrating the effectiveness of combining complementary vision-language representations with selective multimodal reasoning for large-scale text-based person anomaly retrieval.

Model merging has recently attracted significant attention as a promising paradigm for constructing unified multi-task models without requiring additional retraining. However, parameter conflicts and knowledge interference across tasks often degrade merged-model performance. Prior work introduced Conflict-Aware and Balanced Sparsification (CABS), which reduces parameter interference through structured pruning and sequential masking. However, CABS relies on grid search to determine scaling coefficients, resulting in exponential time complexity, while its optimization objective can be dominated by high-performance tasks, leading to suboptimal overall performance. To address these limitations, we extend CABS and propose CABS+. Specifically, Adaptive Weight Allocation (AWA) optimizes merging coefficients via a gradient-free search scheme to reduce time complexity, while an asymmetric fitness function promotes more comprehensive performance gains across tasks. Moreover, we conduct a systematic empirical study of key factors influencing model merging performance and propose Relative Synergy Score (RSS) to quantify model mergeability and guide model selection. We compare CABS+ with state-of-the-art model merging methods, including CABS, AdaMerging, and WUDIMerging, across 27 datasets and 5 models covering large language, small-scale language, and vision models. Extensive experiments verify the effectiveness and efficiency of CABS+. Compared with AdaMerging and WUDIMerging, CABS+ improves overall performance by 16.97% and 12.93%, respectively, exhibits stronger stability and robustness across varying task numbers and model architectures, uses less than 25% of the GPU memory required by AdaMerging, and achieves nearly a 4x speedup in merging time over WUDIMerging.

We study recursive self-improvement at the level of quantitative-investment research: whether an autonomous system can use evidence from earlier experiments to improve the hypotheses and candidates proposed in later iterations. We present AQuA, which comprises two separate language-model-driven research systems: one for symbolic factor discovery and one for trainable model development. The two systems do not share agents, memories, candidate spaces, or research state. Instead, each independently closes its own research loop by retaining validated evidence and using it to guide subsequent proposals. In this bounded sense, both systems implement recursive self-improvement at the level of the research process. Each system also uses its own sealed sandbox, which fixes the data splits, feature and label definitions, and evaluator while allowing the model to act only through constrained factor expressions or configuration diffs. The factor system, a manager-mediated multi-agent pipeline, discovers and combines factors into a signal that reaches a combined information coefficient of about $0.190$ on a crypto universe. The model system, a config-driven loop over a hybrid time-series architecture, reaches a per-stock information coefficient of $+0.0843$ on US equities and converts it into a threshold long/short strategy with a held-out Sharpe of up to $+2.50$ at a two-leg cost. The strategy is positive in every year from 2021 to 2025.

Large language models often fail when answer options require combining atomic judgments under explicit logical operators, even when they judge the individual atoms correctly. We study compound options connected by AND, OR, and NEITHER/NOR, introducing a framework that decomposes each option into atomic answers and scores contrastive hypotheses about each one, so the model never sees a compound option. An operator-constrained integer linear program then composes the calibrated scores into a single prediction. We evaluate on LOGICAL-COMMONSENSEQA and introduce LOGICAL-SATA, a reading-comprehension benchmark derived from SATA-Bench. Our framework improves Macro-F1 from 48.3 to 77.0 on the human-validated LOGICAL-COMMONSENSEQA split and from 47.0 to 75.6 on LOGICAL-SATA, with the largest gains on NEITHER/NOR.

Online platforms increasingly compare many adaptive decision policies---ranking systems, recommendation algorithms, pricing rules, and language-model agents---while each reward-bearing interaction can be costly or risky. A direct A/B/n design gives each of $J$ policies its own horizon-$T$ trajectory and therefore uses $JT$ outcomes. We introduce Tree-Coupled A/B Testing (\TCAB), an exact feedback-sharing design for arbitrary history-dependent contextual-bandit policies. At each round, a predictable tree connects the current policy histories; every parent--child context--action law is maximally coupled, and one reward is shared within each component of matched tree edges. Every policy retains exactly its standalone finite-horizon trajectory law, even though the policies are deliberately dependent. If $D_{e,t}$ records a mismatch on tree edge $e$ at round $t$, the number of reward queries satisfies the pathwise identity $N(T)=T+\sum_{t,e}D_{e,t}$ and hence equals $T$ plus cumulative tree-edge total variation in expectation. This cost is conditionally optimal among exact edge-local designs on the selected tree, and a current-round minimum-spanning tree is myopically optimal among tree designs. For fixed $J$, sublinear pseudo-regret of every policy and almost-sure uniqueness of the oracle action imply $\mathbb{E}[N(T)]=T+o(T)$, versus $JT$ for independent runs. We also obtain finite-sample variance bounds for pairwise policy contrasts. Experiments on reward-model evaluation, multiple-choice language-model evaluation, and adaptive search policies demonstrate substantial improvements in the cost--precision frontier.

Distribution steering seeks feedback laws that drive the state law of a dynamical system between prescribed initial and terminal distributions. Optimal transport provides a natural geometric approach, but its implementation generally requires a transport map or coupling in the full state space. Sliced optimal transport avoids this full-dimensional construction through one-dimensional projections. Yet, the resulting projected maps specify only directional displacements and do not by themselves prescribe a realizable feedback law. To this end, we develop a finite-horizon control framework based on sliced optimal transport. At each sampling instant, a projected optimal transport map defines a directional terminal condition, whose minimum-energy realization yields a randomized single-direction controller. Averaging over projection directions gives a deterministic sliced feedback. For the single-integrator dynamics, the averaged feedback makes the sliced Wasserstein distance to the target non-increasing. For Gaussian endpoint laws, it is affine, preserves Gaussianity, and steers the mean and covariance to their prescribed terminal values. We further identify a law-dependent gain that yields linear decay of the sliced Wasserstein distance together with an explicit characterization of the control energy. We also prove that the randomized controller converges to the averaged sliced flow as the sampling period vanishes. Finally, we extend the construction to linear dynamical systems. Reachability-normalized coordinates allow instantaneous realization of the sliced velocity for uniformly fully actuated systems, while local controllability Gramians provide exact finite-step realization for general controllable systems. Numerical examples illustrate the resulting distributional flows.

Large language models (LLMs) remain vulnerable to harmful requests and jailbreak attacks. Parameter-efficient safety alignment methods based on prompt tuning typically rely on a single global prompt or externally selected prompt modules. Such static designs struggle to maintain a cross-category safety boundary while generating constructive responses tailored to specific risks and avoiding over-refusal of benign inputs. To address these limitations, we propose HiRoute, an input-adaptive hierarchical prompt-tuning framework that separates category-agnostic safety control from category-specific response guidance. HiRoute first trains a lightweight hierarchical router on representations extracted from a frozen LLM to jointly detect harmful intent and predict multi-label risk scores. It then freezes both the backbone model and the router and uses preference optimization with alternating gradient updates to learn a shared coarse-grained prompt and a set of fine-grained prompt experts as continuous embeddings. At inference time, benign inputs bypass the safety branch, whereas risky inputs are processed using the shared prompt together with a router-weighted mixture of risk-specific prompt experts. Experiments across three instruction-tuned models show that HiRoute achieves high safety rates across multiple safety benchmarks while preserving safe-response helpfulness, reducing over-refusal, and maintaining competitive performance on general-purpose tasks.

FastThaiG2P provides sub-millisecond Thai grapheme-to-phoneme conversion for text-to-speech pipelines (International Phonetic Alphabet and Kokoro-TTS conventions) using a PyThaiNLP-tokenized, extensible dictionary and normalization rules for common Central Thai speech. The approach achieves an average latency of 0.15 ms per utterance on a benchmark of 27,242 synthetically generated utterances, of which 30\% is spent on tokenization, 12\% on normalization, and 58\% on out-of-vocabulary fallbacks (0.5\% OOV rate). To demonstrate its effectiveness, we used FastThaiG2P to phonemize Som-TTS, an open dataset containing 20 hours of grapheme-and-audio pairs, then trained an 82M-parameter StyleTTS 2 model based on a Kokoro-TTS recipe. The resulting model vocalizes intelligible Thai speech suitable for prototyping and development at 0.25 real-time factor (4x real-time) with ONNX inference on CPU.

The exceptional generation capabilities of text-to-image diffusion models have raised copyright concerns, particularly the unauthorized reproduction of animation characters. Existing concept erasure methods fall short for animation character erasure: model modification methods struggle to identify suitable anchors for diverse, highly distinctive characters; prompt-based steering methods lack fine-grained control for precise intervention. These approaches often yield incomplete erasure and degraded image fidelity, hindering real-world deployment. In this paper, we propose a controllable method operating on the model's continuous textual representation to erase target characters during generation. We optimizes an anchor embedding via structural and detailed constraints to serve as a character surrogate, then replaces target-related embeddings with the anchor via a structure-aware adaptive strategy. Experiments show that our method achieves state-of-the-art erasure effectiveness and image fidelity preservation, while supporting controllable erasure degree, multi-target removal, and model transferability. Moreover, our optimized anchors are plug-and-play with current model modification baselines to improve their erasure performance.

Access to clinical data is essential for developing reliable healthcare machine learning systems, but direct use of electronic health records is constrained by privacy regulation, institutional review, data-use agreements, and the risk of re-identification. Synthetic data promises a practical alternative: it can preserve useful statistical and clinical structure while reducing exposure of sensitive patient records. Prior studies often evaluate a single generator, one dataset, or a narrow downstream task, making it difficult to know when synthetic data can support model development and when it fails to preserve task-critical signal. We introduce CoMedBench, a reproducible benchmark that evaluates a family of generators under a common clinical-validity framework and one shared training and evaluation engine, spanning static tabular and temporal downstream tasks on established critical-care datasets. In total the benchmark spans 37 dataset-task pairs across two modalities consists of 20 static tabular and 17 temporal ICU time-series-drawn from seven public data sources: three intensive-care databases (MIMIC-III, MIMIC-IV, and eICU) together with the UCI Machine Learning Repository, the CDC BRFSS diabetes cohort (2015), NHANES (1999-2014), and the pycox survival datasets (GBSG and METABRIC). The benchmark evaluates both statistical fidelity and task utility by comparing models trained and tested across real and synthetic data. In these settings, synthetic training data preserves most of the downstream signal: on tabular tasks the reference generator CoMed-CTGAN retains a mean AUROC utility (the synthetic-to-real performance ratio) of 90.6%, rising to 97.3% for the strongest generator, CoMed-TVAE. Temporal ICU tasks are harder and more generator-sensitive: CoMed-CTGAN retains 81.6% (AUROC) and only 64.0% under the imbalance-sensitive AUPRC, whereas CoMed-TVAE still retains ~95% (AUROC).

Simulating the ALICE Zero Degree Calorimeter (ZDC) neutron detector responses at the LHC is computationally expensive, requiring complex Monte Carlo chains. We develop a generative surrogate, focusing on Normalizing Flows (NFs). Through transfer learning, we pre-train on the full imbalanced dataset and fine-tune specialized models for different particle types ($γ$, $n$, $Λ$, $K_S^0$, $Σ^+$) using two gradual-unfreezing schemes. As standard ZDC metrics like Wasserstein distance overlook conditional structure, we introduce refined metrics: conditional weighted MAE, dispersion ratio, and Jaccard co-activation error, that better capture physics-relevant input-output dependencies and response variability. Our ensemble of fine-tuned models achieves a Wasserstein distance of $1.61 \pm 0.02$, outperforming baselines across all metrics. This work provides a generalizable NF-based framework for LHC detector simulation, combining NFs, conditional fine-tuning, and physics-motivated evaluation.

What a finite learning device has recorded and what will hold value for it on future tasks are not the same quantity. We develop a typed accounting for finite-state learning devices that separates four components: a training-side fit functional $Φ_{\mathrm{fit}}$, the record-correlation stock $J_{D}=I(M;D)$, an update-side search ledger $σ_{M}$, and an operational capital value $V(M;T,b)$. This value is the work gap between an informed protocol class and a blind class obtained by deleting the memory-read port and re-optimizing from scratch. (I) Separation: for every $n$, there is a device family on which record correlation and world correlation grow by $n\ln 2$ while the capital gain is exactly zero. In the $\mathrm{flat}^{*}$ regime, data-free updates never increase $V$. (II) Capitalization ledger: an exact $\mathrm{flat}^{*}$ extraction identity and a universal ledger identity give, for (F5$'$)-stable $M$-local updates under a no-discarded-record-correlation condition (f), the bound $η_{\mathrm{cap}}\le 1$ for the capitalization efficiency $η_{\mathrm{cap}}=ΔV/(k T\,σ_{M})$, together with necessary and sufficient conditions for equality. (III) Value retention: for the retention gap $L_{\mathrm{gen}}$ and retention ratio $ρ_{\mathrm{gen}}$ (the former carries no sign constraint; the latter is defined for positive training-side value and is not confined to $[0,1]$) we give a two-layer alignment domain: an exact exchange rate between value and the side-information-adjusted record fit $I(M';D\mid Y)$ without any record-side-information independence assumption, and a raw record-stock exchange rate under a joint side-information neutrality condition $(M,D)\perp Y$, whose boundary is marked by an explicit one-time-pad witness. These are statements about finite-device value retention under task-distribution shift, not a theory of statistical generalization.

Tool-using agents consume external data from sources with different levels of trust, yet tool responses rarely identify who produced each component or what it should convey. We show that this gap enables state-corruption attacks, in which attacker-controlled content makes environmental claims beyond the informational authority of its response component and corrupts the agent's perceived environment, making the resulting action appear justified to existing guardrails. We introduce PIPES (Provenance-Informed, Prior-Enforced Screening), which screens response units using semantic priors and source provenance. PIPES uses static field contracts when schemas provide stable expectations, and conditions screening of open-ended content on the pre-response trajectory and trusted provenance metadata. It marks units that violate their semantic prior or the provenance hierarchy; deployments may remove, warn, block, or escalate detected violations. We instantiate atomic removal and evaluate PIPES against adaptive PAIR-style attacks. Across the three VitaBench and three AgentDyn splits with Gemma 4 31B IT as the target agent, PIPES reduces average attack success from 84.7% to 2.3%, while preserving average benign utility (92.5% with PIPES versus 90.6% without defense).

The rapid advancement of Auto-Research has surfaced a fundamental evaluation challenge: how can we measure the alignment, logical coherence, and evolutionary completeness of its research trajectory with human research behavior? We propose Auto-Research's Alignment and Completeness, ARAC-Bench: a Researcher-Mimicking Evaluation framework that shifts the objective from matching final answers to reproducing high-quality human research processes. The framework operates through two synergistic components: the Academic Cognition Skills system, which is the first to transforms implicit reviewer expertise into stage-calibrated, quantifiable rubrics; and a three-stage capability diagnostic protocol, which decomposes the research process under strict modular constraints into three traceable, mutually independent dimensions: Proposal, Experiment, and Synthesis. Systematic evaluation of 11 SOTA frameworks yields a best alignment score of only 67.9 of 100, revealing a significant gap in simulating rigorous human methodology. Validation against Ph.D. Candidates rankings shows a strong correlation of 0.8141, confirming that ARAC-Bench reliably reflects the dimensions researchers truly value. ARAC-Bench provides not only a fine-grained diagnostic tool but also a scalable reward signal for training the next generation of autonomous research systems.

Understanding the temporal progression of symptoms in clinical narratives is critical for disease monitoring, safety surveillance, and causality assessment. Clinical narratives, however, rarely provide explicit temporal anchors. Current approaches to temporal information reasoning focus predominantly on pairwise relation classification across multi-visit and timestamp-rich records, leaving the reconstruction of structured symptom trajectories from individual anchor-sparse reports largely unaddressed. We propose CRAFT, an LLM framework that pairs a generator with a constraint-based verifier to iteratively produce and refine stage-wise symptom timelines through targeted feedback. We conduct evaluation on MedTempo, a new benchmark of 5,347 vaccine adverse-event narratives spanning three COVID-19 vaccine types, with expert-validated temporal stage annotations for 3,166 reports. Experiments across four LLM backbones demonstrate that CRAFT consistently improves temporal ordering accuracy, with ablation analysis isolating the contribution of generator and verifier components across model capability levels.

This paper introduces ViTOED, a novel dataset for target-oriented emotion detection in Vietnamese social media texts. The ViTOED comprises 10,985 user comments and 21,244 manually annotated opinion quadruples (source, target, expression, polarity) that follow strict guidelines. The dataset reveals Vietnamese-specific phenomena, such as implicit sources and targets and vocabulary ambiguities, enabling deeper analysis of user emotions toward entities. We propose a baseline using structured sentiment graphs and evaluate various Vietnamese pre-trained language models. The empirical results highlight challenges in span detection and relation extraction and indicate substantial room for model improvement in Vietnamese Target-Oriented Emotion Detection tasks.

Semi-supervised semantic segmentation has long turned on one question, which pseudo-labels to trust, and a generation of selection rules, dynamic thresholds, per-class curricula, soft confidence weights, answered it for the noisy, under-confident ResNet teachers of their day. Self-supervised foundation encoders change the regime: with a DINOv2 teacher, confidence saturates, so the filtering that helped a weak teacher can hurt a strong one. We propose CW-BASS v2, a saturation-aware pseudo-label selection method that reads the teacher's confidence regime rather than committing to one rule. It pairs held-out calibration, an unbiased per-class noise estimate, with a self-adaptive confidence floor that provably bounds retention away from 1, and combines them in a one-pass gate: measure the reliability of the teacher's confident set, pi_kept = Pr[correct | c >= tau], on a held-out slice, and filter strictly when it meets the confidence demanded (pi_kept >= tau), falling back to the adaptive floor otherwise. The boundary is the pre-existing operating threshold, not a value tuned to mIoU, and across six DINOv2 teachers it makes the correct strict-vs-floor call blind. CW-BASS v2 thus recovers the UniMatch V2 operating point on the saturated benchmarks by selecting strict (Pascal VOC 1/8 87.4 against its reported 87.9; Cityscapes within 0.5), and improves on it where the confident set is unreliable (pi_kept ~ 89%, ADE20K), where the floor edges ahead (+1.5 mIoU, single seed). The gate is principled because the failure it avoids is measured, not assumed: on a reliable, saturated teacher the confidence distribution's dynamic range collapses (98% of Pascal pixels >= 0.95), so an adaptive cutoff floods the retention mask and self-training decays into confirmation bias.

The extent to which large language models for code rely on memorization over genuine understanding remains highly debated. While current literature frequently reports widespread memorization, evaluating the underlying probing techniques across dense architectures reveals a severe breakdown in their utility at scale. Traditional encoder-style probes using perturbations such as synonym fuzzing or dead-code insertion struggle to expose memorization in scaled models, even on known-contaminated benchmarks, and decoder-style probes that rely on log probabilities show similar performance degradation. The specific mode of failure for these probes, particularly why such techniques disrupt smaller models but fail to impact larger ones, motivates us to untangle representation load from memorization rather than treating them as a single phenomenon. By applying invertible mathematical transforms to numeric problems, we isolate these two factors and reveal that scaled encoders successfully absorb substantial representation load while still converging on the correct family of solutions. In practical software engineering, this ability to adapt to varying surface forms is what truly matters for usability and generalizability in LLM and agentic applications. Whether a specific solution was seen during training becomes a much less pressing question because although memorization inflates scores on contaminated benchmarks, factoring out representation load makes it debatable how much we should truly care if a functional answer was originally memorized. Future evaluations must therefore be built around separating these phenomena rather than relying on methodologies that quietly entangle them.

Deep search agents operate over trajectories spanning dozens of steps, yet standard reinforcement learning provides only a single outcome reward per trajectory, which is far too sparse for effective credit assignment. On-policy self-distillation (OPSD) addresses this by using the model's own logits as dense token-level teachers, but extending it to search agents introduces a fundamental tension: the teacher, having access to privileged information such as the correct answer, produces a distribution that differs systematically from the student's exploration-based reasoning, and naive distillation causes the student to inherit this information asymmetry rather than learn better search strategies. We resolve this tension through two contributions. First, we construct Evidence Anchors, which are concise, step-level evidence snippets extracted from the web, as privileged information that captures key reasoning steps without revealing the entire answer path. Second, we propose Step-Level Self-Distilled Policy Optimization (SSPO), which converts teacher-student disagreement into step-level advantage weights within GRPO, applied exclusively to incorrect trajectories. This design decouples what to update from how much to update: the outcome reward determines the direction of policy change, while the teacher modulates its magnitude at each step. Correct trajectories are left untouched, preserving their diversity. On Qwen3-8B, SSPO consistently outperforms GRPO across BrowseComp, GAIA, and FRAMES, surpassing or matching GRPO trained with twice as many gradient steps while adding only about 5 percent overhead per step from a single additional forward pass.

Schedulability analysis is essential for certifying real-time systems, but existing tests are often developed through pen-and-paper proofs that are difficult to scale, validate, and maintain. Mechanized verification in PROSA/ROCQ offers a rigorous alternative, yet manually constructing such proofs requires substantial domain expertise and proof-engineering effort. Recent successes of large language models (LLMs) across a wide range of tasks make them promising candidates for generating PROSA/ROCQ scripts for mechanized theorem provers. However, state-of-the-art LLMs often lack the PROSA-specific knowledge required to correctly use its modeling abstractions and proof patterns. This paper introduces PROVE-RT, an LLM-assisted framework for generating PROSA/ROCQ scripts to mechanize schedulability analyses in real-time systems literature. PROVE-RT guides generation through dependency-aware informal sketches, retrieval from processed PROSA documentation, staged skeleton generation, and proof completion. We construct a mechanization-oriented corpus from 1, 191 real-time systems papers, containing 13, 134 informal sketches with dependency information. On a curated evaluation set, direct prompting of state-of-the-art LLMs fails to reliably generate valid PROSA mechanizations, whereas PROVE-RT achieves a success rate of 44.7%. These results show that retrieval-guided and staged LLM assistance can improve automated mechanization of schedulability analysis in PROSA/ROCQ.

Agentic workflows are commonly evaluated by whether they reach the correct outcome. That is insufficient in institutional settings, where a correct action may rely on the wrong authority, an unsupported completion claim, or work made stale by a later change. We define governed execution as work whose decisions, completion, and response to change are supported by inspectable provenance. We present Matrix, a deterministic causal-state layer that records authority and fact dependencies, verifies completion evidence, and selectively invalidates affected work. Across controlled comparisons, governed and direct workflows often reached the same outcomes, but only the governed path consistently preserved governing evidence, refused unsupported closure, and limited recovery to dependent tasks. A role-separated transfer challenge then failed: a deterministically enforced completeness contract severely over-blocked synthetic packets produced outside its authoring context. These results do not establish Matrix as a general accuracy enhancer; they support its primary role as an institutional integrity layer for making agentic work auditable and independently verifiable.

Laplacian-regularized minimization is fundamental in signal processing and machine learning, but is limited by the dense and ill-conditioned nature of the graph Laplacian pseudoinverse. While the Laplacian itself is sparse, its pseudoinverse is dense and often ill-conditioned, rendering direct computation impractical at scale. Moreover, pseudoinverse learning is more challenging than Laplacian learning. To address this challenge, this paper considers the setting where the graph Laplacian is given and proposes a Difference-of-Convex Regularizer (DCR) graph learning framework that approximates the spectral action of the Laplacian pseudoinverse without direct inversion via regularized Maximum Likelihood Estimation (MLE). By reformulating Laplacian-Regularized Nonnegative Least Squares (LR-NNLS) through a dual representation, DCR decouples pseudoinverse learning from instance-specific inference and enables efficient primal solution reconstruction via a differentiable dual-guided learning scheme. We establish theoretical guarantees on stability and the existence of a unique fixed point for DCR algorithm. Numerical experiments demonstrate improved performance over convex solvers and graph filtering baselines and robust performance across diverse graph topologies.

Adaptive latent tokenization maps a fine-grained input to a shorter sequence of continuous representations associated with input-dependent spans. We introduce ReconSpan, which divides text into chunks that a backward decoder can reconstruct from a single contextual prefix code and retains one such code as the latent token for each chunk. The reconstruction criterion is applied when chunks are formed, allowing one trained autoencoder to produce average chunk lengths from 6.5 to 12.2. At matched average length, reconstruction-guided boundaries preserve more text than random boundaries. Readers of the resulting latent sequence recover topic information reliably but struggle to extract exact details.

We study decentralized multi-player reinforcement learning in episodic tabular Markov decision processes (MDPs) under three forms of information asymmetry: (A) unobserved actions with common rewards, (B) observed actions with independent rewards, and (C) unobserved actions with independent rewards. Players cannot communicate during learning but may agree on a protocol a priori. For Problems A and B we propose \texttt{mQ-learning} and \texttt{mQ-learning-intervals}, achieving $\tilde{O}(\sqrt{H^4 S A_{\text{joint}}\, T})$ regret, where $H$ is the horizon, $S$ the state count, $T = KH$ the total steps, and $A_{\text{joint}} = \prod_{i=1}^M |\mathcal{A}_i|$ the joint action space across $M$ players. For Problem C we give \texttt{mEXC} and \texttt{mEXC-Bellman}, two-phase explore-then-commit algorithms with regret $\tilde{O}(H (S A_{\text{joint}})^{1/3} T^{2/3})$. Against the centralized joint-action benchmark, decentralized learning under information asymmetry matches the single-agent Q-learning rate of \cite{jin2018q} up to logarithmic factors. Because $A_{\text{joint}}$ grows exponentially in $M$, the bounds are most meaningful for small $M$ or small per-player action sets.

Logic synthesis transforms RTL designs into gate-level netlists, where PPA results are highly sensitive to the choice of optimization commands, making synthesis tuning both high-dimensional and expensive. Previous approaches fall into two categories: automated methods, which perform black-box search over fixed action spaces with limited decision-level interpretability, and LLM-based methods, which typically generate static scripts upfront and cannot adapt to evolving circuit states. We present SynAct, an adaptive closed-loop LLM reasoning--acting agent that iteratively diagnoses live synthesis reports and reasons over the current circuit state, retrieved tool knowledge, and historical optimization experience to issue targeted commands. SynAct focuses on improving timing, particularly worst negative slack (WNS), while maintaining balanced area and power trade-offs. Experiments on a commercial synthesis tool across 14 designs show that SynAct reduces average WNS to 27% of that from bootstrap synthesis.

LLM-based simulated clients are increasingly used to train novice counselors, evaluate LLM therapists, and generate synthetic data. However, current simulators produce overly cooperative clients that disclose too readily, accept therapeutic reframes without resistance, and resolve core issues within a single session. We trace these issues to profiles that lack causal depth and behavioral mechanisms that treat all content as equally accessible. We present PatientAct, a framework for client simulation grounded in established clinical theories. Our profiles integrate the 5Ps clinical case formulation, providing causal depth without tying the design to any single therapeutic modality. During simulation, profiles include a dynamic memory layer in which items carry trust thresholds (e.g., symptoms are available early, whereas formative memories require a sustained therapeutic alliance). At each turn, the client's emotional reaction and behavior are modeled before generating a response. If the therapist approaches gated content, PatientAct expresses resistance in terms of quantity, content, and style rather than defaulting to cooperation or a single resistance pattern. We evaluate our framework on 40 clinical situations and demonstrate that it generates diverse profiles with high clinical plausibility. Moreover, PatientAct significantly outperforms the baselines, yielding substantial gains in resistance quality and behavioral realism. Our code and data will be publicly available via github.com/Sahandfer/PatientHub.

Object hallucination in multimodal large language models arises when language priors and corpus co-occurrence bias outweigh the visual evidence, with nothing tying an individual object mention to what the image shows. Most remedies intervene at decoding time without training, yet under a unified protocol their benefit is confined to short captions;supervised fine-tuning (SFT) on a detail- rich corpus lengthens captions, but over forty percent still name absent objects. This paper proposes Dual-Stream Cross-Anchor Correction (DSCC). Unlike work that post-processes decoding, DSCC is the first to inject object-level visual anchors into the language model itself during fine- tuning: a perception stream aligns object-level hidden states at an intermediate layer to frozen text anchors by a bidirectional contrastive objective; a cognition stream lets deeper layers query those anchors by cross-attention at every generation step; and a two-stage curriculum gate couplesthem, making evidence retrieval a structural constraint at each autoregressive step. Under one backbone and one scoring protocol, experiments span long-caption hallucination, object-existence discrimination and cross-domain generalisation, with vanilla SFT on the same corpus and schedule as a length- and density-matched control, so gains are attributed layer by layer. DSCC is the only method reaching the long-caption, low-hallucination region: captions roughly 1.9 times the baseline length at 88.19% precision per object mention, the highest under a density-independent criterion. Ablations expose a synergy: the perception stream alone degrades precision yet reverses sign when stacked on the cognition stream. No universal superiority is claimed: three out-of- domain benchmarks yield a predictable, falsifiable domain-conditionality, the synergy being bound to the anchors' semantic domain and breaking on charts and optical illusions.

Medical AI has demonstrated specialist-level diagnostic accuracy, yet these capabilities remain largely inaccessible in resource-constrained rural settings where bandwidth is scarce, compute is limited, and clinical decision-making requires integrating heterogeneous modalities. We introduce a cloud--edge collaborative architecture that addresses these constraints: lightweight, domain-specific models on the edge transform raw medical data into compact structured outputs, while a cloud LLM synthesizes these outputs into clinical summaries. An LLM-based orchestrator dynamically selects diagnostic tools based on patient context, promoting comprehensive modality coverage without processing irrelevant inputs. We evaluate on 20 multimodal clinical cases spanning cardiac, obstetric, trauma, and screening scenarios under three simulated network profiles (500,kbps--5,Mbps). The hybrid system achieves 98--99% diagnostic tool recall with 92--96% precision, matches or exceeds cloud-only baselines on clinical accuracy, and maintains bandwidth-invariant latency (25--35,s) at 4--15x lower token cost. These results highlight the role of architectural design in enabling efficient multimodal integration and improving factual grounding compared to cloud-only approaches under deployment constraints.

Spatial intelligence is becoming a foundation for embodied agents, robotic planning, and multimodal assistants. To improve the spatial reasoning ability of VLM agents, existing work has mainly followed two lines. One line uses post-training methods, such as supervised fine-tuning and reinforcement learning. Another line adopts an agentic paradigm in which the model calls external spatial tools, such as depth estimation and 3D reconstruction tools, to gather intermediate spatial evidence. We study a complementary and underexplored route: Can a frozen VLM agent improve its spatial reasoning through \textbf{parameter-update-free self-evolution}, without depending on external expert spatial tools at inference time? We present \textbf{Spatial Memory Agent (SMA)}, an \textbf{experience-grounded runtime framework} that converts verified spatial experience into reusable transferable lessons. In a verifiable spatial environment, SMA queries the frozen VLM, obtains a predicted answer and reward, and uses \textbf{verifier-guided reflection} to distill compact transferable lessons from spatial experience. SMA further assigns each lesson a \textbf{Transfer Reliability Score (TRS)}, which is initialized uniformly and calibrated from later retrieval outcomes as visit evidence of future transfer reliability. During \textbf{read-only deployment}, SMA retrieves lessons by semantic filter and similarity-TRS combined ranking, allowing the retrieved memory to guide frozen model inference. Across five representative spatial benchmarks and four base VLMs, SMA achieves the highest macro average in every base-model block and the best accuracy among the evaluated methods in most of the 20 evaluations, establishing a practical parameter-update-free path for spatial self-evolution across the evaluated frozen model scales and environments.

Few-shot in-context learning (ICL) with multi-modal large language models (MLLMs) enables task adaptation without parameter updates, but its performance is highly sensitive to the quality and coverage of the selected demonstrations. While unlabeled multi-modal data is abundant, it remains elusive how to exploit them for ICL. We propose MAG (MAnifold-Guided semi-supervised in-context demonstra- tion selection), an efficient framework that leverages unlabeled data to improve multi-modal ICL. MAG formulates demonstration selection as a semi-supervised propagation problem on a multi-modal graph and adopts a two-stage strategy: (i) relevance score propagation identifies a compact set of high-impact unlabeled samples for pseudo-labeling, reducing MLLM inference cost; (ii) multi-modal relevance is used to select the final demonstrations. We show that textual represen- tations are more effective for relevance propagation, while both visual and textual modalities are crucial for high-quality demonstration selection. Experiments on eight multi-modal benchmarks demonstrate that MAG consistently outperforms strong baselines in label-scarce regimes, achieving significant gains with a limited pseudo-labeling budget.

While Large Language Model (LLM) agents increasingly rely on long-term memory for persistent interactions, the retrieval mechanisms governing this memory are rarely treated as evolvable components. This static approach limits performance on heterogeneous memory queries, which often demand diverse evidence construction strategies. To address this, we introduce \textbf{ERSkill}, a retrieval-centric framework for self-evolving, skill-guided memory access. ERSkill compiles interaction histories into a structured memory store and represents retrieval behaviors as executable skills composed of fundamental primitives. At inference time, a trained router dynamically matches each query to the optimal skill to construct tailored evidence for answer generation. To enable continuous improvement, ERSkill co-evolves the skill set and the router during training. It employs an experience trie to efficiently record explored retrieval paths, alongside a double-frontier mechanism that safely decouples the expansion of new skill capabilities from stable, router-facing deployment. Experiments across multiple agent memory benchmarks demonstrate that ERSkill substantially outperforms strong non-evolving and self-evolving baselines. Notably, it improves the overall average across F1, BLEU-1, and LLM-judge scores by 31.3\% with Qwen3-Next-80B-A3B-Instruct and by 28.1\% with GPT-5.4-nano.

Routing each query to a cost-effective large language model (LLM) is critical for balancing quality and cost, yet most routers rely on a centralized task center to predict model performance, creating an information-risk mismatch and a scalability bottleneck as the model pool grows. We propose a market-based routing paradigm that shifts ex-ante prediction to LLM providers via a reverse auction, where providers bid with self-predicted success probabilities and execution costs. To account for inherently noisy provider predictions and center evaluations, we introduce the \textit{\textbf{E}rror-\textbf{A}ware \textbf{R}everse \textbf{A}uction \textbf{M}echanism} (EA-RAM), which explicitly models this inherent Dual Error. We prove that EA-RAM is Bayesian incentive compatible and individually rational under the Dual Error, establish sufficient conditions for center rationality, and derive an explicit welfare-loss bound. We further identify robustness effects: opposite-signed errors can cancel, vanishing-tail link functions (e.g., logistic) stabilize clear-cut cases via saturation, and extra noise smooths belief maps, reducing the gains from marginal manipulation. Experiments on simulations and real-world benchmarks show that EA-RAM is robust to the Dual Error and achieves a better cost--performance Pareto frontier than centralized baselines, with additional gains when providers contribute local information, validating its practical effectiveness.

Mechanistic interpretability of large language models lacks spatially resolved, falsifiable tools for testing whether internal components are specialized for distinct cognitive operations. We adapt subtraction analysis, the standard framework of human neuroimaging, from biological brains to perturbed transformers, and apply the same logic to both substrates in parallel. Building on the Brain-LLM Unified Model (BLUM), which showed that layer-perturbed LLaVA-1.6-Vicuna-13B error profiles match the lesion patterns of aphasic patients, we develop PRISM (Perturbation-based Regional Interpretability through Subtraction Mapping). PRISM maps the seven clinical Philadelphia Naming Test categories, subtracts error classes pairwise, and treats each perturbation seed as a subject in a group analysis with threshold-free cluster enhancement along the layer axis. We run a structurally matched analysis on 213 chronic post-stroke aphasia patients using correlation-difference lesion-symptom mapping, and replicate both sides on held-out splits. The designs match in subject dimension (seeds, patients), spatial dimension (layers, atlas-parcellated cortex) and thresholding, but the contrast operator differs: a within-subject error-proportion difference for the LLM, a between-subject correlation difference for the cortex. Both substrates recover a robust phonemic-favoring dissociation, a deep layer cluster and a frontal-perisylvian cortical cluster, both replicating; the semantic-favoring direction is a consistently signed but non-significant trend on both. PRISM thus gives a falsifiable, spatially resolved test of functional-specialization claims in transformer language models. A confirmatory ROI-level intervention (PRISM Stage 3) licensing the strongest causal-mechanism claim is left to subsequent work.

Generative speech enhancement faces three gaps: spectral models capture harmonic structure but often disrupt phase, waveform models preserve phase but miss harmonics, and Schrödinger Bridges (SB) shorten transport from noise to clean speech but leave inference cost only loosely tied to training. We propose HybridSB-MoE, a dual-domain framework that fills these gaps through three contributions unified by a single asymmetric design principle. (i) Asymmetric uncertainty fusion: The spectral path captures epistemic uncertainty via expert disagreement, while the waveform bridge models aleatoric variance through stochastic dynamics. We fuse them asymmetrically, allowing the mixing weight to adapt to distinct error regimes rather than average predictions. (ii) Heterogeneous MoE with top-k=2 routing across five distinct architectural archetypes, where architectural diversity makes the epistemic signal indicate which inductive bias fails rather than small perturbations among similar experts. (iii) Discretization bound (Theorem 1): path-consistency and trajectory regularizers together bound the K-step bridge sampling error in 2-Wasserstein distance at rate K-alpha, making small-K inference an objective-level guarantee rather than an empirical claim. On VoiceBank+DEMAND, HybridSB-MoE outperforms diffusion- and SB-based baselines at their step budgets while remaining competitive with consistency-distilled few-step methods.

Watermarking LLM-generated text is an important task for tracing its provenance. Existing LLM watermarks preserve provenance under editing, but this same robustness allows an adversary to alter critical content while retaining attribution, a vulnerability known as piggyback spoofing. We introduce an innovative watermark that jointly provides provenance and tamper evidence. It co-embeds a robust signal and a fragile signal into each generated token. The signals share the same mechanism but use independent keys and different seeding windows over normalized text, making one resilient to edits and the other sensitive to reader-visible changes. Multiple rounds of unbiased tournament reweighting preserve the expected generation distribution, while a periodic round-allocation pattern controls the trade-off between the two signals. At detection, their scores form a two-dimensional space supporting three decisions: Intact, Tampered, and No-Watermark. Across two large language models and two prompt datasets, our method demonstrates the highest tamper-detection rate among the evaluated methods while maintaining competitive attribution robustness and perplexity. Ablation studies show that reliable three-state detection requires a well-defined notion of intactness, co-embedding of the two signals, and complementary sensitivity to edits.

Muon, a more recently developed optimizer, is useful for matrix-wise models in AI areas. Although many works have studied Muon and its variants, these methods are still not particularly well-suited for hierarchical structured problems. To fill this gap, we propose an effective federated compositional Muon (FedCoMuon) optimizer to solve distributed matrix-wise compositional optimization problems. Specifically, our FedCoMuon optimizer builds on compositional gradient tracking and orthogonalized momentum. Moreover, we propose a variance reduced variant of FedCoMuon (FedCoMuon-VR) based on a momentum-based variance reduced technique. In theory, we analyze the convergence properties of our algorithms under the non-i.i.d. and non-convex settings. In particular, we prove that our FedCoMuon-VR obtains a lower sample complexity of $O(ε^{-3})$ for finding an $ε$-stationary solution than the existing FedMuon algorithms. Extensive numerical experiments on robust federated learning and task-distributed risk-sensitive meta learning show that our proposed methods are competitive with existing compositional baselines and achieve the best reported accuracy in several settings.

Multi-objective bilevel optimization has wide applications in the AI area such as automated learning and multi-task meta-learning. Although recently some works have been begun to study the multi-objective bilevel optimization, the proposed methods rely on the (strongly) convex lower level problems. In fact, these multi-objective bilevel learning problems are generally nonconvex, and particularly their lower level problems are nonconvex. To fill this gap, we propose a class of Multi-Objective Moreau Envelope based Hessian-free Algorithms (MOMEHA) to solve the multi-objective bilevel learning problems with nonconvex lower level. Specifically, our method uses the Moreau envelope to convert the original problem into a multi-objective single-level optimization with an envelope constraint. In particular, our method retains computational advantages of being single-loop and Hessian-free in the multi-objective setting by incorporating a smooth weighted Tchebycheff scalarization. Furthermore, we propose a momentum-based variant of MOMEHA (i.e., MB-MOMEHA) method to solve the stochastic multi-objective bilevel learning problems. In theory, we provide the convergence properties of our algorithms under both deterministic and stochastic setting. Some experiments on few-shot meta-learning and neural architecture search demonstrate that our methods outperform the existing approaches in Pareto front, validating its effectiveness and robustness.

Systems that generate GPU kernels with language models report high correctness rates. Those rates come from a single loose test: run the kernel on a few random inputs at one fixed shape and accept it if the output is close to a reference. A kernel can pass that test and still be silently wrong. It can return an ordinary number where the true answer is a NaN or an infinity, differ from run to run, break when the shape changes, or accumulate in fp16 where the reference keeps an fp32 total. We build the instrument that checks correctness properly: a contract-grade verifier of twelve adversarial gates, each a property a correct kernel must satisfy, several of them tolerance-free, so no choice of threshold can explain a failure away. Aimed outward, the verifier audits 2,638 machine-generated kernels that a public system's own harness had already accepted as correct. It finds 39.5% broken beyond any tolerance argument and 62.1% carrying at least one violation. The field's standard test accepts 1,487 kernels the verifier rejects, against only 14 the other way. We defend the finding four independent ways: a 7/7 positive control, a threshold-calibration sweep, 98.5% agreement with the reference benchmark's own correctness code, and a stratified hand-audit. Aimed inward, the verifier judges a kernel of our own: the first native Blackwell tcgen05 training backward for the gated-linear-recurrence (GDN) family, including the reverse-state stage the field still runs on a fallback. We establish its correctness independently, against a double-precision oracle, and train five family members through it. The correctness signal behind reported progress in kernel generation is far weaker than the numbers suggest, and a set of tolerance-free contracts would close most of the gap.

Self-supervised electrocardiogram (ECG) models are often trained on a few seconds of ECG signal and, increasingly, on discretized token sequences. It remains unclear whether these choices sacrifice information needed for rhythm inference and longitudinal consistency in real-world ambulatory recordings. We present a controlled study on the Icentia11k single-lead dataset that varies (i) the input horizon (16 seconds, 1 minute, 5 minutes, and 10 minutes) and (ii) the front-end representation (continuous convolutional patch embeddings vs. fixed vector-quantized tokens), while holding the Transformer backbone and training protocol constant. Representations are assessed by downstream abnormal rhythm detection and by patient-level retrieval that probes cross-session stability. Our results show that increasing temporal context beyond 16-second snapshots yields stronger transfer and higher retrieval accuracy, with the strongest performance achieved by the 5- and 10-minute models, indicating improved capture of slow-varying rhythm dynamics and individual-specific structure. Across all evaluated horizons, continuous patch embeddings outperform discretized tokens, suggesting that quantization can discard clinically relevant waveform detail. These findings motivate ECG foundation models that emphasize extended context and continuous encoders for clinical prediction and similarity-based applications. Our code and pretrained models are publicly available at https://github.com/muha-0/ecg-ssl-representation-learning.

Multi-parametric magnetic resonance imaging (mpMRI) is a cornerstone for brain tumor diagnosis and treatment, yet current AI models face critical limitations: their lack of natural language interaction and interpretability impedes spatial information integration and cross-modal reasoning required clinically. Key challenges arise from significant physical meaning differences across modalities, spatial misalignment due to scan intervals, and the need for complex multi-feature interpretation in tasks like glioma grading. While visual-language models (VLMs) show promise in cross-modal understanding, existing methods focus mainly on 2D image modeling, neglecting direct perception of 3D volumetric space. Although 3D VLMs have been proposed for report generation and feature alignment in 3D CT imaging, mpMRI applications demand collaborative inference across multiple imaging modalities-a requirement unmet by current solutions. To address this, we introduce Mr3D-VL, a dedicated visual-language foundation model for multi-parametric 3D MRI. With 4 billion parameters, it employs an unsupervised pre-trained shared 3D encoder and 4D rotational positional embedding for dual modality-spatial integration. Its cross-modal projection layer uses a multi-resolution feature implantation strategy to enhance feature perception across resolutions. Experimental results show significant improvements over existing 4B/7B/30B domain-specific and general-purpose models in text generation tasks, achieving a BERTScore of 0.856 for report generation, with question-answering accuracy at 0.713 and multiple-choice accuracy at 0.912.

Bayesian optimization (BO) is a sample-efficient framework for analog circuit topology search, where evaluating each candidate topology can require costly simulation. However, representation-based BO methods typically treat circuit embeddings as fixed after encoder training. This creates a mismatch between representation learning and optimization: embeddings learned to encode or reconstruct circuit structure are not necessarily organized according to the figure of merit (FoM) being optimized. This paper introduces Test-Time Analog Representation Adaptation for Bayesian Optimization (TTARO), an online deep-kernel BO framework that adapts circuit representations throughout the search process. Starting from pretrained circuit embeddings, TTARO jointly learns a nonlinear feature transformation and a Gaussian-process surrogate using the FoM labels of the circuits evaluated so far. Following each new evaluation, TTARO updates the representation and surrogate before selecting the next candidate. We compare TTARO with conventional Gaussian Process-based BO over fixed embeddings and with Deep Kernel Learning (DKL), which learns the representation only from the initial evaluated designs and keeps it fixed throughout the remainder of the search. By continually incorporating newly observed FoM labels into representation learning, TTARO aligns the search space with the optimization objective as BO progresses. In our experiments, TTARO reduces regret AUC by 15.2% on average relative to BO and by 20.7% relative to DKL across 40 encoder/kernel/acquisition settings, outperforming prior art in most settings with reductions as large as 46.7%.

Item demand forecasting is an integral component of store assortment optimization. Existing literature focuses on learning a suitable customer choice model and using this model to determine the value of an objective function (i.e. expected demand) with respect to an assortment proposal. However, for large item universe with many categories, this approach can prove inefficient, needing a separate demand forecast for every possible item assortment. An alternate approach exists whereby we combine the efficiency of forecasting item demand independently, while at the same time applying adjustments to the independent forecasts that account for the relations between item demand and the availability of other similar items on the shelf. Central to this approach is the estimation of Demand Transfer (DT) coefficients. These DT coefficients represent the percent of a particular target item's (item that the customer walked in the store to buy) demand that is redirected to each other item in the universe should the target item be removed from the shelf. We introduce an approach that allows us to compute these DT coefficients on large item universes (assortments having 1 million+ items). Experiments on data as well as historical transaction data for multiple locations within categories demonstrate that when certain reasonable assumptions about substitution behavior are satisfied, our procedure is able to accurately estimate underlying DT coefficients and lead to improvements in demand forecasting.

Large Language Models (LLMs) are increasingly deployed in discovery domains such as math and science. The usual approach is to present the problem to the model and use its answer as the proposed solution. However, beyond this best guess, discovery can be enhanced by increasing test-time compute. In a process called pass@k, the model is allowed to explore the solution space and generate diverse candidate solutions. Unfortunately, the standard approach to post-training LLMs through Reinforcement Learning (RL) may limit pass@k: the model's output distribution narrows around high-reward outputs, causing the solution coverage to collapse. The alternative is to use Evolution Strategies (ES), a population-based, gradient-free post-training method that optimizes directly in weight space through random perturbations. As this paper shows, ES achieves consistently higher pass@k than RL and produces a broader output distribution with greater solution coverage. This coverage in turn makes it possible to achieve better results in e.g. standard math benchmarks. Thus, ES provides a better foundation for post-training in discovery problems and other domains where diverse solution coverage is critical.

Detecting infection-related behavioral changes in mosquitoes from video data is challenging because mosquitoes are small, move rapidly and irregularly, and are affected by environmental factors such as background, lighting, and shadows, which can make reliable feature extraction difficult. In this study, a YOLO- and Contrastive Language-Image Pre-training (CLIP)-based vision-language framework is proposed to classify mosquito flight frames of uninfected and Dengue virus serotype 2 (DENV2)-infected mosquitoes. First, YOLO is used to isolate mosquito regions from the background. Then, visual features extracted from video frames are aligned with biologically meaningful textual prompts in a shared embedding space. The multimodal model was fine-tuned using supervised bidirectional contrastive learning and evaluated through frame-level image-text similarity-based classification. The results show that the proposed method achieved 98.54% accuracy and 99.91% sensitivity at the frame level. After temporal aggregation of frame-level information, the model achieved complete video-level performance. The ablation results showed that fine-tuning and CLIP-based representations were essential for this domain, while the textual branch provided semantic image-text alignment rather than an accuracy advantage over the vision-only model. These findings suggest that vision-language models can provide a useful framework for analyzing infection-related biological behaviors from video data.

Retrieval-Augmented Generation (RAG) is widely used to improve the performance of Large Language Models (LLMs) in answering user queries. Existing privacy research on RAG has focused on preventing unauthorized users from accessing sensitive data. However, another important problem that is often overlooked in RAG privacy research is that external generators have access to the query and the retrieved documents, which may contain confidential information that could potentially be misused or accessed for unintended purposes. In this paper, we introduce the Sensitive Entity Alias Generator (SEAG), a privacy-preserving framework that empowers users to utilize powerful third-party generators without disclosing sensitive information. SEAG introduces a lightweight model that locates sensitive entities, generates corresponding aliases, and constructs an entity replacement table. The table is used to replace sensitive words in the user's query and in the retrieved documents before they are forwarded to an external generator. For this purpose, two datasets were constructed: one for fine-tuning SEAG models to generate entity replacement tables, and another for evaluating the entire SEAG framework. The experimental results demonstrate the success of the SEAG framework. As for the User metric, which measures the ability of the model to provide a correct response to the user while hiding sensitive information from the external generator, all SEAG models achieved over 80% accuracy. Additional analysis further evaluated the ability of SEAG models Qwen-3, LLaMA-3.2, and Phi-4 to hide all sensitive entities within given documents. The results show good performance with total accuracies of 77.83%, 76.73%, and 74.91%, respectively.

Maintaining price consistency and executing an Every Day Low Price strategy is critical for global retailers. However, with catalogs spanning millions of active items, manual governance of price relationships is infeasible. Inconsistent pricing across item variants distorts customer value perception and cannibalizes sales. To address this, we present a scalable, context-aware Multi-Agent Framework designed to automate the construction of "Lines and Ladders" pricing taxonomies. Our framework employs specialized LLM agents to construct these coherent pricing structures by identifying key attributes, extracting multi-modal values, and applying hierarchical grouping logic. Evaluated on real-world enterprise data and deployed in production, our 3-Agent system achieves an F1-score of 0.83 for Lines, outperforming single-agent baselines by mitigating cognitive overload. The system achieves >90% precision and >75% recall in Food & Consumables, and 80.2% assignment accuracy in the unstructured General Merchandise catalog.

Multi-layer transformers form the critical component of essentially all large language models (LLMs) in use today. Because of their ubiquity and computational capability, there is a rapidly growing body of work that aims to precisely calibrate the expressive power of transformers as language recognizers by comparing them against standard models of computation studied for decades by the theoretical computer science community. In this endeavor, circuit complexity has by and large emerged as the "correct" branch of computational complexity to analyze the expressive power of transformers; the reason is that parameterizing transformers by the various resources they use, such as attention and precision, leads to direct comparisons with different classes of circuits parameterized by resources such as type of gates, size, and depth. Here, we present an overview of selected results that delineate the expressive power of transformers using concepts and methods from circuit complexity.

IT service management (ITSM) systems accumulate large volumes of heterogeneous ticket data that are difficult for sales and executive stakeholders to convert into actionable intelligence. This paper presents a sociotechnical AI pipeline, designed and evaluated following design science research principles, that transforms raw ITSM exports into a multilevel decision-support artifact. The pipeline combines LLM-based schema normalization, HDBSCAN sub-topic clustering, and hierarchical agglomerative clustering to generate executive-facing Main-topics and granular Sub-topics. A stakeholder evaluation across six artifacts and five raters from Sales Engineering and customer success roles shows that all four decision-support metrics, interpretability, actionability, trust, and likelihood of use, on average exceed 4.0 out of 5.0, with trust as the most consistent signal. The findings position ITSM analytics as an Information Systems (IS) problem of transformation, abstraction, and human-centered design.

For solving nonconvex equality-constrained optimization problems, a recent Gradient-Eigenstep Algorithm by Goyens et al.~is an iteration-efficient approach, based on minimizing Fletcher's augmented Lagrangian function, for finding an approximate second-order stationary point from an arbitrary starting point. In this paper, the analysis of this algorithm is extended, offering a two-fold contribution. First, it is shown that a local-linear rate of convergence can be obtained by this method if it is initiated sufficiently close to a strong second-order stationary point and employs a sufficiently small step-size parameter and sufficiently large penalty parameter. In this case, the algorithm reduces to a gradient descent algorithm applied to minimize Fletcher's augmented Lagrangian. Second, as a particularly useful application of the first result, it is shown that the Gradient-Eigenstep algorithm can be used as an iteration-efficient subproblem solver in the context of a progressive sampling strategy for solving equality-constrained optimization problems when the objective and constraint functions are defined by large sample averages, ultimately offering an algorithm with an improved worst-case sample complexity when compared to an approach that solves a full-sample problem directly.

Wildfire susceptibility mapping typically relies on physical variables assembled from multiple remote-sensing, climate, and geospatial products. AlphaEarth Foundations (AEF) provides analysis-ready geospatial embeddings that may reduce this dependence on heavy harmonisation and task-specific feature engineering, but their value for wildfire susceptibility mapping has not been systematically evaluated. Using Victoria, Australia (2017-2025), as a case study, we show that AEF embeddings can reconstruct commonly used variables in wildfire susceptibility analysis with high accuracy. In downstream susceptibility models trained on satellite-derived fire occurrence data, embedding-based susceptibility models achieve ROC-AUC values above 0.92 and consistently identify high wildfire susceptibility across eastern Victoria, particularly Gippsland and the north-eastern uplands, with additional localized hotspots in central and northwestern Victoria. A key feature of AEF embeddings is their strong near-region transferability within climatically similar regions. When embedding-based models trained in Victoria are applied to Canberra and Western Sydney-Blue Mountains, ROC-AUC improves by around 4% at Canberra and declines by around 2% at Western Sydney-Blue Mountains, compared with a mean decrease of approximately 25% for physical-variable models. These findings provide practical guidance for using AEF embeddings and lay a foundation for scalable wildfire susceptibility mapping workflows for downstream users such as government agencies and (re)insurers.

Probabilities of causation (PoCs) characterize individual causal responses that cannot be directly observed and therefore generally require partial identification. Tian and Pearl first derived theoretically sharp bounds for binary PoCs, including the probability of necessity (PN), the probability of sufficiency (PS), and the probability of necessity and sufficiency (PNS). Mueller et al. subsequently tightened the bounds for binary PNS by incorporating causal information encoded in covariates and mediators. More recently, Li and Pearl, as well as Shu et al., extended PoCs to multivalued settings and derived corresponding theoretical bounds. These developments naturally raise the question of whether additional causal knowledge can further tighten the bounds in multivalued settings. This paper addresses this question by deriving tighter bounds for multivalued PoCs through the incorporation of causal information encoded in covariates and mediators. We illustrate the theoretical results with toy examples, while simulation studies further demonstrate that the proposed bounds are tighter than existing nonbinary bounds.

A flat training curve does not reveal whether a neural network has reached a global optimum, is locally trapped, is representation-limited, or is mismatched to its trainer. We introduce Training Under Challenge, an executable-certificate framework in which predeclared, architecture-valid procedures construct complete alternatives in the same certified class and reevaluate the same objective. Any lower-valued candidate is a replayable witness that lower-bounds the checkpoint's empirical global-optimality gap. Passing a finite suite is only suite-relative; global-gap conclusions require a separately justified coverage mechanism. We define a resource-indexed challenge-power modulus that characterizes the largest gap compatible with passage. For squared loss, current block-decrease operators make coverage checkable and yield uniform and realized-residual bounds. We prove the converse frontier: without coverage, a first-order ReLU trainer can reach infinitely many exact conditional head optima while converging to a non-global point. On a channel-gated ResNet-18 distillation problem with known optimum, eight internal challenges cover all 240 audited output directions, and realized-residual bounds lie within factors of 1.74--3.02 of the true gap. Paired predictive certificates separate decoder under-use from representation insufficiency, while quantized-denoising studies demonstrate diagnosis, repair, and current-state recertification.

Long-running LLM agents act through tools, and a single step can send an email, merge a pull request, or wire a payment. The steering decision is the pre-commit choice at that boundary: proceed, or hold for human or policy review. We introduce SteerBench-Work, an incident-anchored, bidirectional benchmark for that decision in workplace agents across developer operations, customer service, finance, legal, medical, HR, and security. Release v2026-05 contains 106 scenarios anchored in public incidents, paired evidence-reversed mirrors, and calibration controls, with labels split nearly evenly between proceed and hold so the two error directions get near-identical numbers of chances. A model sees the proposed action and the available evidence, returns a gate decision, and is scored on whether it crosses or holds the boundary correctly. Across 30 model conditions the failures run almost entirely in one direction: models wrongly hold authorized, evidence-cleared work on 28.1% of opportunities and wrongly allow unsafe work on 1.0%. The hardest cases are risk-resolved commits, where signed or structured evidence has already cleared a real risk trigger, and models score markedly worse on evidence-reversed mirrors of famous incidents (63.8%) than on the incidents themselves (98.5%). General capability is not the same as steering calibration: higher-capability models often over-refuse at the commit boundary, and more reasoning can repair a weak gate while leaving a calibrated one flat. The public leaderboard is at steerbench.com.

Benchmark contamination is diagnosed today with n-gram overlap, with likelihood-based membership inference, or with canary strings, and each needs something usually unavailable: the training corpus, a well-chosen test statistic, or foresight at dataset release. A recent alternative reads contamination off a linear probe on internal activations. We show that the natural way to do this does not work, and specify one that survives measurement. The protocol reports a zero-sum contrast on the depth profile of probe accuracy, recentred on a level-matched placebo baseline, tested against a label-permutation null, with the reference set twice the size of the suspect set. Each choice replaces a simpler alternative we measured and rejected. Reporting the level of excess separability rather than its shape makes the false positive rate track the size of the analyst's own control set, from 0.03 to 0.99 under a true null. Contrasting against a flat depth profile fails in both directions, rejecting a true null 0.72 of the time when surface decodability rises with depth and losing all power when it falls. An item bootstrap holds the fitted probe fixed and rejects up to 0.09 of the time where a permutation null that refits it holds 0.02. A half-size baseline triples the error rate. On real transformers, baseline depth profiles are measurably not flat, spanning up to 29.1 accuracy points on a temporal split, and their non-flatness tracks the surface difference between the item sets (correlation 0.87 over 6 audits), so the correction is largest exactly where it is needed. All 4 well-matched Pile arms return null, and the protocol refuses a verdict on the temporal split rather than reporting one. What this does not establish is whether transformers carry a familiarity direction at all: the only positive sits on the split where exchangeability fails. Implementation, tests and audits are released.

LLM judges have become central infrastructure for model evaluations, online grading, and reward modeling. Judges are typically validated by accuracy on golden data, but accuracy says little about whether they are stable under re-prompting, challenge, or sustained pushback. We introduce the \emph{Wiggle Framework}, a unified stress test for epistemic stability in LLM judges. The framework decomposes judge robustness along three dimensions: Mechanical Consistency (stability under re-prompting and reframing), Single-turn Conviction (stability under a single challenge), and Multi-turn Persistence (stability under sustained or adaptive pressure). We use the framework to study 9 frontier models across 14 judging tasks spanning safety, toxicity, AI writing detection, and political-response evaluation. Every model exhibits substantial wiggle as a judge --- flipping verdicts 25--71\% of the time under static pushback, and 62--91\% with an adversarial LLM persuader. Critically, we find that pressure that succeeds in changing a judge's verdict is almost always net-corrupting with respect to ground truth. Beyond the framework itself, we identify baseline jury majority strength as the most effective single-shot signal for anticipating which items wiggle. Taken together, this is the first apples-to-apples cross-dataset comparison of mechanical, conformity, and persuadability tests in a judging context.

Causal discovery aims to uncover the underlying causal relationships given data generated from a system. The goal, however, is not merely to predict causal edges given data, but also to be able to interpret and explain either observed or hypothesized phenomena, such as a particularly large causal effect. We consider this task of conditional causal discovery and cast it as a Bayesian inference problem, in which we target the posterior over causal graphs and parameters conditional on an event such as a causal-effect constraint. Unfortunately, this poses a computational challenge: existing approaches to Bayesian causal discovery struggle when the event has small posterior mass. To address this, we adapt rare-event estimation techniques to perform inference the joint graph-parameter space. Our method gradually drives a particle population toward the constrained region while maintaining samples that approximate the conditional posterior. Empirical evaluation on synthetic graphs validates the accuracy of our approach at small and large scales, and we show in a case study on the Sachs protein dataset how our method can be used to aid scientific exploration by providing pathway-level summaries.

While large language models can generate entire novels, there is little information about the level of formal variation in their output over many generations. Rather than asking whether individual passages can be identified as AI-generated, this study asks whether repeated AI generation can produce the same range of diversity which is found across human corpora. This paper contrasts six corpora based on generation source and target style: twenty novels generated using GPT-5.5 Thinking in a nineteenth-century British realist style, twenty novels generated using Qwen3-14B in a nineteenth-century British realist style, twenty novels generated using each of these models in a contemporary zero style, 205 nineteenth-century human-written British novels, and sixty-five contemporary human-written Zero-Style novels. At the document level, the research includes MATTR-500, Shannon entropy, average sentence length, readability, and punctuation rate measurements. The most robust and reliable result is compression of sentence structure. Repeated generations produce novels that vary far less from one another in sentence structure than human novels do. Compression is also present in the measures of readability, punctuation, and sentence length variability within novels. Lexical measures tend to be similarly compressed, with the exception of Qwen Zero-Style MATTR. Despite having distinct mean stylistic profiles, GPT and Qwen lack a stable pattern of cross-measure correlation. This article therefore distinguishes between variance overclosure, which represents a limited formal range between novels, and a more specific phenomenon of correlational overclosure. This means that an individual AI-generated novel may resemble human fiction stylistically, while a collection of AI-generated novels occupies a much narrower formal range.

GPU kernel agents and GPU programming languages have advanced separately, leaving expert kernels difficult to reproduce. Agents usually treat the compiler as a fixed black box and receive only errors, correctness outcomes, and timing, while existing DSLs either hide critical scheduling decisions or expose them through difficult layout abstractions. We present CAKE, a compiler-agent co-design in which agents author CAKE IR, a typed, hardware-explicit schedule representation. CAKE exposes warp roles, memory movement, synchronization, and pipelines while supporting verification, cost modeling, and localized diagnostics. The harness itself evolves: recurring failures become verifier rules, IR primitives, model calibrations, and reusable optimization tactics. In matched implementation-hidden Flash-KMeans clean starts on B200, the best CAKE IR candidate at an 80-million-token budget runs at 1.144x the tuned FlashML baseline, compared with 0.928x for direct CUDA/PTX. Beyond this benchmark, agent-generated Kimi Delta Attention achieves a 2.05x geometric-mean speedup over official FlashKDA and passes end-to-end serving validation. Dispatcher-backed KNN and KMeans improve performance by 1.42x to 2.12x across more than 400 shapes, and four kernel changes are available as upstream PRs. CAKE targets NVIDIA GPUs from Ampere through Blackwell and separates single-shape evolution from library generalization and dispatch.

Long-horizon egocentric memory transforms continuous first-person video and audio into a searchable record of past experiences. We demonstrate two bottlenecks in existing systems: indices built from context-poor captions are unreliable for agentic search, while retrieval ignores a question's temporal intent. To address both bottlenecks, we introduce EgoCITE (Egocentric Context-augmented Indexing and Time-aware Evidence retrieval), a long-horizon agentic memory framework for egocentric QA. EgoCITE comprises three components. EgoScheme uses local multimodal context to turn fragmentary video captions and speech transcripts into self-contained atomic memory indices. EgoIndex organizes complementary action, activity, utterance, and conversation representations into searchable multi-view memory indices at multiple granularities. EgoRetrv combines semantic search with question-conditioned temporal relevance scoring and curation of retrieved evidence. We evaluate EgoCITE on EgoLifeQA, EgoMem, and EgoR1-Bench in terms of answer accuracy and target-event retrieval alignment. EgoCITE improves accuracy over agentic memory baselines by at least 4.4--14.2\% while achieving 36$\times$ lower cost than long-context LLM agents.

Strategic reasoning in Large Language Models (LLMs) within long-horizon environments is often limited by inconsistent subgoals. In these settings, finite attention resources prevent the model from maintaining strategic coherence over thousands of steps. This limitation leads to strategic drift, where localized decisions fail to sustain a coherent trajectory across reasoning. To address this, we introduce EpicStar, a framework that enables agents to learn memory as policy to tackle long-horizon reasoning. Specifically, the agent maintains a bank of successful past episodes as a heuristic alongside a working memory to track short-term environmental changes. During inference, a dynamic gating mechanism determines whether to execute a retrieved action directly or to perform new reasoning through a contextual fusion of the retrieved episodes and current working memory. Utilizing StarCraft II as the testbed, we evaluated EpicStar against diverse opponent styles. It significantly outperforms baseline methods, achieving higher win rates while consuming an order of magnitude fewer tokens, and it maintains this advantage consistently across difficulty levels and opponent strategies. Our findings provide compelling evidence that structured cross-episode memory is essential for enabling LLM agents to perform robust, long-term strategic execution in dynamic, autonomous settings.

Structure-preserving machine learning embeds physical structure directly into model architectures, yet uncertainty quantification (UQ) for such hard-constrained models remains limited because standard UQ methods may violate the encoded admissibility conditions, require architectural modifications, or impose substantial computational costs. In this work, we propose Structure-Preserving Epistemic Neural Networks (S-PENNs), a general framework for UQ in scientific machine learning models with hard architectural constraints, and instantiate it for GENERIC (General Equation for Non-Equilibrium Reversible-Irreversible Coupling) dynamics. S-PENNs preserve the structural constraints of a pretrained model by attaching lightweight epinets to its constrained components, ensuring that every sampled realization remains physically admissible by construction. When applied to GENERIC dynamics, such a proposed framework yields thermodynamically consistent rollouts that preserve the first and second laws. Furthermore, we combine S-PENNs with split conformal prediction as a post-hoc calibration method to produce prediction intervals with finite-sample marginal coverage guarantees. We validate S-PENNs on three numerical examples: a harmonic oscillator coupled to a heat bath and an idealized chemical motor, both governed by ODEs, and a one-dimensional viscoplastic model governed by PDEs. Across all three examples, S-PENNs produce thermodynamically consistent stochastic realizations and well-calibrated prediction intervals while reducing the computational cost by about one to three orders of magnitude compared to deep ensembles. Although the present study focuses on GENERIC dynamics, S-PENNs can be extended more broadly to scientific machine learning models in computational mechanics with either hard or soft constraints.

Language model classifiers with explanations are used for moderation, routing, topic triage, and low-resource annotation. We study black-box auditing when the defender has only clean calibration data without trigger information but can ask the classifier for a label plus a short rationale or quoted evidence. We introduce Groundedness Drift, a lightweight score measuring whether the answer summary remains grounded in the input. Across two 7B backbones, five datasets, and four common non-adaptive OpenBackdoor-style attack families, Groundedness Drift achieves higher AUROC and lower residual target ASR than every compared detector in all cases at a nominal 5\% clean-FPR budget. We then evaluate Unsupported Groundedness, a multi-probe escalation for explanation-camouflage stress cases. Unsupported Groundedness improves signals but does not close the adaptive gap.

We prove that, on finite simple undirected graphs equipped with a single Boolean node feature, the Boolean queries expressible in $Σ$-MPLang, for any collection $Σ$ of eventually constant activation functions and with arbitrary real coefficients, form a strict subclass of the Boolean queries expressible in ReLU-MPLang. We thereby settle a recently posed open problem: whether ReLU-MPLang is more powerful than trReLU-MPLang when it comes to Boolean queries. In particular, this implies that ReLU-GNNs are strictly more expressive than {TrReLU,id}-GNNs with respect to Boolean queries on Boolean-featured graphs.

In-vehicle music can serve as an adaptive interface to enhance driver experience, attention, and well-being. We present Drive-to-Music, a context-aware system that generates music in real time from multimodal driving signals. Using dashcam imagery and vehicle telemetry, the system extracts scene semantics and driving context, maps them to high-level musical descriptors, and conditions generative audio models to produce contextually aligned soundtracks. The architecture combines perception and generative components to translate visual and kinematic inputs into structured musical attributes and synthesize audio with low latency. It supports smooth transitions as driving conditions evolve, and to ensure robustness and deployment readiness, we incorporate constraint-based controls and safety checks across the generation pipeline. Our results demonstrate the feasibility of real-time, context-aware music generation in automotive settings, providing a foundation for personalized and adaptive in-vehicle audio experiences.

Existing screenshot-to-code systems face a trade-off between flexibility and controllability. Direct multimodal generation can hallucinate visible details, whereas structured pipelines reduce such errors through component-wise decomposition, predefined templates, and customized intermediate representations. These structures, however, introduce additional generative orchestration and restrict outputs to designs covered by the representation. We investigate whether selective tool grounding can improve the fidelity--efficiency trade-off of direct widget-to-code generation. We introduce \textbf{WidgetGen}, a lightweight tool-grounded framework that extracts observable text and color evidence, performs high-level layout and optional chart reasoning, and directly generates executable JavaScript XML (\emph{JSX}). This design reduces reliance on component-wise generation while avoiding a fixed UI schema. Across six multimodal models and \(1{,}000\) held-out widgets, WidgetGen outperforms direct prompting and the structured Widget2Code pipeline on most visual reconstruction metrics, with consistent gains in area, legibility, and style. Finally, reconstruction-derived image-code pairs improve six Qwen-family open-weight models across every reported metric through supervised fine-tuning. These results establish WidgetGen as a strong lightweight baseline and show that selective evidence grounding offers an effective alternative to extensive representation constraints.

There are 56,804 public agent skills today, and teams write many more privately. The dominant delivery model is installation: once installed, a skill's description remains in the system prompt, competing for fewer than 100 reliable trigger slots. This leaves the long tail with no practical path to use and forces teams' own playbooks to compete for the same scarce space. We observe that installation bundles three separable functions: content, persistence, and automatic triggering. Only the last requires prompt residency. We therefore propose @skills, an open protocol that separates them. A path addresses any skill, subtree, or collection, and reading a skill is sufficient to use it, so nothing is installed or made resident. The operation vendors a copy at the same path into a project's Git-tracked tree for adaptation and ownership. The operation adds one .gitignore-style line, the only element that costs prompt residency. A directory is a menu, making bundles ordinary directories rather than all-or-nothing units. The protocol requires no manifest, lockfile, or registration, and SKILL.md remains unchanged. @skills is additive, ships as an installable package, and turns any agent that can read files and run commands into a client through a single instruction file. Its open specification is at https://github.com/SylphAI-Inc/atskills and it is implemented in the AdaL CLI at https://adalagent.ai . Because paths address skills well but cannot find them, the protocol is paired with a free hub at https://atskills.one for corpus-wide search and ranking, repository-free hosting, private and team collections, and one-screen authoring. The hub is optional: gh: and local paths resolve without it, and indexed GitHub skills retain their gh: identities. Install less, use more.

A critical challenge in deploying online HD map construction systems to real-world scenarios is the scarcity of labeled training data, which limits model generalization in diverse environments. To address this limitation, we propose a teacher-student semi-supervised learning (SSL) framework that generates high-quality pseudo-labels from unlabeled data through confidence-aware map refinement. Our approach first trains a teacher model on limited labeled data, then leverages Beta-distribution-based confidence maps to assess the reliability of predicted map elements across temporal observations. Unlike conventional filtering methods that discard entire elements, we introduce a spatial clipping technique that selectively preserves high-confidence regions while removing unreliable segments. The refined map elements serve as map priors that improve the teacher model's prediction accuracy on unlabeled data in a second pass. These enhanced predictions become pseudo-labels for training a student model from scratch, followed by fine-tuning on the original labeled data. Experimental results on the nuScenes dataset demonstrate that our teacher-student framework with refined pseudo-labels improves performance by +6.1 mAP under a low-label regime compared to training on labeled data alone, offering a practical solution to the labeled data scarcity problem in online HD map construction.

Multi-turn dialogues let users revoke constraints as easily as impose them, but revocation does not reliably take effect: models keep enacting withdrawn requirements (occasionally beneath comments asserting their removal), a failure we call \emph{behavioral relapse}, or revocation inertia. No existing instrument measures this influence per clause, predicts it before delivery, or repairs it under matched budgets. \sysname{} closes the three gaps through the model API alone: a contract ledger pairs every constraint with an executable checker, records revocations as tombstones, and compiles the net constraint state ahead of time into a single specification; a sequential ablation probe measures per-clause adherence and incremental behavioral effect; a repair ladder operates under token- and attempt-matched budgets. On \dataname{} (\NTasks{} HumanEval tasks, \NClauses{} verified checkers), relapse at an 8B operating point climbs from \ScaleDelayedMTwo{} to \ScaleDelayedMEight{} as constraint load grows, while stronger models sit at floor. Under matched checkers, model, and budget, ahead-of-time compilation significantly reduces relapse against a no-ledger verifier-retry baseline (\RestoreDiff{}, 95\% CI \RestoreDiffCI{}, $p$ \RestoreDiffP{}); adaptive ladder interventions stacked on top add no detectable gain (95\% confidence excludes gains $\geq$ \LadderExcludedGain{}). The probe predicts relapse before delivery (AUROC \AurocPrimary{}); a one-sentence tombstone note recovers about a third of the compilation effect and survives a placebo control. At \CostDeliveryFactor{} delivery overhead and \CostTotalHedged{} of API compute for every result, revocation failure becomes a measurable, predictable, and repairable property of dialogue state rather than an invisible one.

Intensional operators are often treated as quantifiers over possible worlds, parallel to the treatment of determiners as quantifiers over individuals. Yet individuals introduced in intensional contexts cannot serve as antecedents to later pronouns as easily as those introduced in merely quantificational contexts. For instance, "Everyone is eating a cheeseburger" may be followed by "They are large", where "they" refers to the cheeseburgers being eaten. However, as Stone (1999) points out, the similar "Andrea might be eating a cheeseburger" does not support later anaphoric references such as "It is large" or "They are large". Stone (1999), Stone and Hardt (1999), and Brasoveanu (2010) address this by requiring a pronoun's value (its referents) to exist in the world of evaluation, ruling out anaphora from non-veridical intensional contexts. We show, however, both cases where such anaphora is disallowed even when the pronoun's referents clearly exist and cases where it is allowed even though they might not exist. We argue that intensional anaphora is best captured using a description-based rather than value-based account. A pronoun presupposes that its corresponding antecedent description is instantiated in each world of the context set. Thus, there must be a cheeseburger being eaten by Andrea in every candidate world for "It is large" to be felicitous after "Andrea might be eating a cheeseburger". We implement our proposal via a new logic, building on Keshet (2018) and Abney and Keshet (2022), called Plural Intensional Presuppositional predicate calculus (PIP). Each PIP formula translates directly into standard first-order predicate calculus with set abstraction, providing a classical foundation for this work.

Neural networks can often be trained or fine-tuned through random low-dimensional reparameterization, where a small latent vector is mapped into a full parameter update by a frozen random map. This raises a practical question: how large must the latent search space be to reach a low-loss region? We first express the known accessibility transition in an equivalent conic form, centered for compact convex targets at the statistical dimension of the polar cone. Our main theoretical contribution is an orientation-resolved quadratic master formula that predicts the random-slice residual from both the curvature spectrum and the reference-to-solution displacement profile. It yields a self-consistent isotropic-orientation predictor and, in a conservative radius-only specialization, recovers the earlier Gaussian-width quadratic bound. Building on this analysis, we introduce Random Mapping Networks (RaMaN), which instantiate the predicted latent dimension using structured Hadamard or seed-regenerated Gaussian maps. These constructions avoid the O(dP) storage of dense random maps and reduce optimizer-state memory from O(P) to O(d). We also develop matrix-free curvature approximations and sweep-free dimension selection. Across controlled quadratic and neural-curvature experiments, the orientation-resolved predictor closely tracks measured transition locations and outperforms orientation-agnostic approximations when displacement direction matters. End-to-end experiments further show sharp, protocol-dependent training transitions across image and language models.

As more investors contemplate private markets and contend with limited transparency, sparse disclosures, and infrequent transactions, identifying economically meaningful peer companies for comparison is a fundamental challenge for valuation, due diligence, portfolio construction, and risk management. We propose an ensemble tree-based supervised similarity learning framework that defines company similarity through the lens of market valuation rather than static feature matching or semantic descriptions. Specifically, we train a CatBoost gradient-boosted decision tree model on observed private company valuations and derive a valuation-aware similarity metric from importance-weighted leaf-node co-occurrences across the ensemble. The similarity metric captures shared valuation drivers while accommodating nonlinear relationships, mixed data types, and pervasive missing data common in private markets. Using a global private-market universe of approximately 270,000 companies, including more than 53,000 firms with observed or derivable post-money valuations spanning multiple industries, geographies, and deal stages, we demonstrate that the proposed similarity framework improves upon traditional distance-based and text-embedding-based approaches in downstream k-nearest-neighbor valuation tasks in the evaluated industry groups, while retaining case-based explainability.

Discovery---formulating novel generalizations---is a central part of the scientific process. Despite its importance, there is a gap in the current AI benchmark landscape, with few benchmarks directly probing the capacity for discovering new knowledge with experimentation in controlled environments where the objective is unknown. To address this gap, we release a new benchmark: DiG-bench (Discovery in Games). DiG-bench consists of a set of 70 independent games. Each game is encoded as a short string and has unique transformation rules that must be discovered through interaction and experimentation. The levels of the game present a series of challenges to test whether the rules have been discovered, where the win conditions for each level are also unknown. We provide games at seven tiers of difficulty for AI agents. The lowest tier is routinely solvable by multiple models, while the highest tier challenges the best models in agentic harnesses. All 70 games were solved by at least one human on first attempt. A subset of 21 games is released publicly, and the remainder is held private for secure evaluation.

Continuous physiological time series underpin modern clinical monitoring, yet many of the most informative signals are invasive, expensive, or simply unavailable for a given patient. Conditional generation offers a remedy: an absent signal can be synthesized from co-recorded signals and routine clinical variables. Existing generators, however, are built around a single conditioning modality and degrade when forced to handle the heterogeneous, irregularly missing mix of time-variant signals and static covariates seen in practice. We propose ReCoGen (Represent Conditions, then Generate), a two-stage framework that decouples multimodal condition representation from target generation. Stage I trains one masked autoencoder per modality, distilling each time-variant condition into a compact and missingness-tolerant token sequence. Stage II trains a flow-matching generator that fuses these tokens with static conditions to synthesize the target signal. Across three physiological benchmarks, including continuous glucose monitoring on AI-READI and arterial blood pressure generation on MIMIC-III and MIMIC-IV, ReCoGen attains the best downstream utility on all sixteen (dataset, task, metric) settings, surpassing six representative conditional generators; on thirteen of them its utility also reaches or exceeds the utility measured on the real signal, a reference we read as an approximate anchor rather than a ceiling. Ablations trace the gains to the conditioning path: learnable cross-attention over the frozen per-modality encoders, and a dual token-plus-AdaLN route for the static conditions. ReCoGen thus turns routinely collected signals into informative surrogates for invasive or unavailable ones, a step toward less invasive, lower-cost continuous clinical monitoring.

Thyroid ultrasound diagnosis requires coordinated lesion localization, measurement, risk stratification and reporting, yet most AI systems address these tasks in isolation and provide limited support for clinical review. We present ThyroidXAgent, a clinician-interactive agentic AI system that coordinates specialized diagnostic tools and stores their outputs as an auditable case-level evidence record. The system was developed using OpenThyroidDB, a multicentre, multitask resource integrating approximately 0.3 million ultrasound images and 24,000 paired reports, and was evaluated on 28,458 non-overlapping test cases, including 8,721 cases from 35 centres in the private NHC-MISD-TUS cohort. Across heterogeneous datasets, ThyroidXAgent achieved a mean Dice score of 87.21 percent for nodule segmentation and a mean AUROC of 0.9466 for benign-malignant classification. The same workflow supported lymph-node metastasis prediction and follicular versus papillary thyroid carcinoma classification, with AUROCs of 0.864 and 0.805, respectively. For report generation, evidence-grounded assembly outperformed multimodal language-model baselines across three cohorts. ThyClinScore, a lesion-level clinical semantic metric introduced here, showed the strongest correlation with a location-aware language-model judge. ThyroidXAgent improved physician classification accuracy, increased report diagnostic consistency from 70.3 percent to 86.2 percent, and reduced segmentation and reporting time by 35.9 percent and 27.4 percent, respectively. These findings support auditable, clinician-correctable agentic AI for thyroid ultrasound diagnosis and reporting.

This article presents with DYSANOS the first generative market model for smooth SANOS option surfaces for all strikes and expiries which are free of static arbitrage. Our model is designed to generate entire paths of daily spot and option prices for years in the future. We present a robust and useful if somewhat simplistic baseline hidden state generative model in the form of an AR(1) model. We discuss model setup, data pipeline, and training and investigate numerical resence of dynamic arbitrage. We illustrate model performance on Option Metrics' IvyDB S\&P Index data from 2020 to~2025 and compare it to a pure implied-vol PCA model.

Improving reasoning LLMs requires the ability to judge the quality of long reasoning traces for effective reasoning data curation, strong training signals during reinforcement learning, and an in-depth understanding of reasoning behaviors during model performance evaluation. Additionally, surfacing reasoning mistakes that the model makes would enable improving the model's performance at runtime through providing feedback. Due to the difficulty of this complex task on long reasoning traces, single-model judges (even frontier models) do not do well at identifying reasoning defects. Additionally, leveraging frontier models during online training of reasoning LLMs is generally prohibited due to guardrails in terms of use. In this work, we introduce Reasoning Jury, a system that replaces the single judge with a jury of LLMs and a moderated consensus mechanism, to improve the fidelity of judgments for identifying reasoning defects. In reasoning jury, defects of a reasoning trace and their severity are surfaced through a deliberation where a moderator conducts a discussion amongst the jury where the jurors critique each other's judgments and get to modify their initial votes. The moderator derives a consensus through deliberation amongst jurors or consolidation of judgements. We show that Reasoning Jury with a jury of open-weight models (e.g., gpt-oss-120b) is able to significantly outperform frontier models (opus-4.6, sonnet-4.6, and gemini-3.1-pro) at correctly identifying reasoning defects. Besides accuracy performance improvements, the aggregated cost of the jury (initial verdicts, deliberations, consolidation, etc.) is a fraction (8 to 15%) of the cost of running frontier models in LLM-as-a-judge setup. We also show how these judgements can be leveraged to understand failure modes of reasoning LLMs on benchmarks, which allows much deeper understanding of a model's performance.

AI journaling tools can tailor prompts to a person's own sensed behavior, but it is unclear which behaviors respond to them. We analyzed 369 journal entries from an eight-week passive sensing study. An LLM labeled each entry as expressing an intention to change a behavior or not, and we measured follow-through against 26 sensor features with a 3-day before/after comparison. Responsiveness depended most on whether a behavior involves other people. Behaviors that depend on others improved in only 15 to 22% of cases, while behaviors a person can act on alone improved more often, up to 50 to 63%, though unevenly. How users wrote mattered less. No single text feature separated improved from unimproved entries; writing carried signal only within specific behaviors, most clearly for text messaging and for longer, more personal intention entries. The sample is small, so we treat these as exploratory patterns that point to where AI journaling nudges are most likely to work.

Large language models increasingly need to generate structured outputs that conform to predefined schemas, with one common constraint being selection from a finite set of valid strings. Current constrained decoding systems handle this through general-purpose grammar compilation, which becomes prohibitively slow as the number of valid values grows into the thousands, a cardinality wall. We introduce the trie automaton, a specialized mechanism that exploits finite-set structure (shared prefixes, bounded depth, known cardinality) via Aho-Corasick multi-pattern matching to precompute per-node token masks. The trie achieves 7X faster per-step valid-token computation (0.65 us vs. 5.8 us) compared to XGrammar, one of the primary backends in vLLM and SGLang, and 2--6.5X faster compilation at K >= 300. Because precomputed masks enable a stateless serving path that bypasses the guided decoding pipeline, this advantage compounds in batch serving: end-to-end vLLM throughput reaches 219 req/s vs. XGrammar's 7.5 req/s at batch size 256 (29X). The 29X combines the algorithmic speedup with integration-path savings that only precomputed masks can unlock. Across seven tokenizer families (32K--262K vocabulary), the trie maintains sub-100ms compilation up to K = 10,000 and flat per-step cost regardless of set size, while guaranteeing 100% output validity.

Top-k selection is a fundamental computational primitive with applications spanning databases, information retrieval, signal processing, and modern machine learning workloads, including sparse activations and attention pruning. As data sizes grow, existing approaches become inefficient: exact methods incur high memory and compute overhead, while approximate methods often rely on brittle heuristics that degrade under adversarial or heavy-tailed inputs. In this paper, we introduce Prof-K, a fast, scalable, and distribution-agnostic top-k algorithm with probabilistic correctness guarantees. Prof-K performs a single-pass filtering procedure: a small random sample estimates an adaptive threshold, the N input elements are streamed once into a compact buffer, and an exact top-k routine on this buffer recovers the true top-k elements with probability at least 1 - $ε$, where $ε$ > 0 is user specified. We derive high-probability guarantees for correctness and buffer size, together with an approximately optimal sample size that minimizes overhead as a function of N and k. Empirically, Prof-K achieves 1.5x-10x speedups over the highly optimized PyTorch topk and recent RadiK implementations, with the largest gains in the large-scale, small-to-moderate-k regime where prior methods struggle most. Unlike previous approaches, these guarantees hold independently of the input distribution, ensuring robustness to adversarial settings. By relaxing the recall target (e.g., recovering 95% of the true top-k values), Prof-K additionally provides a principled accuracy-speed trade-off. We further demonstrate its impact on training BatchTopK Sparse Autoencoders (SAEs), where top-k selection constitutes a significant portion of the training cost.

In 2023, a New York judge sanctioned two attorneys in Mata v. Avianca for filing a brief with hallucinated citations generated by ChatGPT. Such failures are largely caught by database lookups; the harder problem is detecting citations that point to real cases but do not support the propositions for which they are offered -- a failure mode that existing evaluations of LLMs for legal use cases largely overlook. In this paper, we study proposition-level citation support verification through controlled perturbations of real legal citations obtained from two legal corpora, either replacing the cited case or changing only the pinpoint page within the same case. We evaluate fourteen model configurations on the resulting examples. Models catch 93-100% of wrong-case corruptions. They catch only 37-61% of wrong-pinpoint corruptions on court opinions and 52-83% on legal briefs. When models fail to catch wrong-pinpoint corruptions, they accept the citation based on topical overlap rather than page-level support. Scale and extended reasoning narrow the gap but do not close it: GPT-5.4 with high reasoning effort still misses 40% of pinpoint mismatches on court opinions and 18% on briefs. Prompting the model to verify support at the cited page improves recall, but it also raises the false positive rate. Recognizing the right legal topic and verifying support for the cited proposition are distinct capabilities, and current models conflate them.

Automating empirical research is a long-standing direction of AI. Recent automatic research (AutoResearch) agents bring this goal within reach, as modern LLMs show the capability to independently implement solutions and learn from the execution outcomes. Behind these gains, post-training (especially RL) plays a central role. In this paper, we identify a fundamental tension when scaling RL for these agents: the two components of every AutoResearch trajectory (agent generation and environment execution) scale in very different manners, since all generation shares compute through batching, while each execution occupies its exclusive sandbox and real machine time. As a result, the environment execution dominates the training cost and becomes the bottleneck as trajectories grow. To resolve this tension, we propose World Model RL (WMRL), which replaces environment execution with a world model to remove this bottleneck. Additionally, the world model can be imperfect, as its rewards are corrupted by bias and noise. Therefore, we further equip WMRL with two mitigations, Online Debiasing and Inverse-Variance Denoising, which offset the bias and suppress the noise respectively. Theoretically, we prove that both mitigations of WMRL strictly improve the convergence guarantee. Empirically, WMRL accelerates training by 3-4x on various tasks at different agent scales, while exceeding the performance of standard RL baselines. Moreover, our post-trained 4B and 9B agents outperform much larger open-weight agents of 48B and 120B on held-out benchmarks. Beyond AutoResearch, WMRL also transfers to post-training embodied VLA policies, which demonstrates the generalizability of our method.

Predictive explanation methods attribute a model output; they do not, by themselves, attribute an intervention effect on the real-world outcome. We introduce the Causal Attribution Score (CAS), a compact score architecture for causal explanation. CAS starts from an identified interventional coalition game, allocates the joint intervention contrast with causal Shapley contributions, and converts those raw outcome-scale effects into Local CAS, Signed Local CAS, and two complementary Global CAS summaries. The innovation is not a new Shapley formula, but a local-to-global causal reporting layer with an explicit intervention target. In the known-truth benchmark, eight repeated primary-interaction simulations (n = 2,200 each, three actions) gave mean Local CAS MAE of 0.107 for coalition-aware CAS, compared with 0.173 for one-at-a-time normalisation and 0.213 for a global normalised absolute ATE vector. The paired advantage over one-at-a-time normalisation increased from -0.003 under additivity to 0.091 under strong interactions. On both empirical DoubleML datasets, 401(k) eligibility/net financial assets (n = 9,915) and Pennsylvania reemployment bonus/unemployment duration (n = 5,099), predictive SHAP/TreeSHAP rankings differed materially from Feature-CAS rankings of treatment-effect modifiers. In Pennsylvania, dep1 (exactly one dependent) moved from predictive global rank 13 to Feature-CAS rank 2 and was the leading local Feature-CAS modifier. These results isolate the added value of separating what predicts the outcome from what explains heterogeneity in an estimated causal effect.

Dance imitation integrates motor planning, sensorimotor integration, and social cognition, offering a sensitive framework to characterize motor behavior in autism. In this work, we explore a computational analysis framework to identify potential biomarkers that allow the design and development of improved medical and human-machine systems. We analyzed 3D motion capture data from autistic and neurotypical adults performing dance imitation under solo and socially-framed duo conditions. Methodologically, using Dynamic Time Warping, we quantified movement consistency and propose the Social Context Sensitivity Index (SCSI) to measure modulation of variability by social framing. These features were then used on a classifier to discriminate subjects into autistic or neurotypical groups. Results show that neurotypical adults exhibited increased movement variability in socially-framed imitation, especially in upper and lower limbs, whereas autistic adults maintained consistent movement across contexts. Classification achieved 79.2% balanced accuracy in distinguishing groups. These findings suggest that social context sensitivity in motor imitation constitutes a robust biomarker of autism-related motor behavior, highlighting the importance of social modulation in motor assessments and informing the development of inclusive human-centric technologies.

Deep learning models are widely used for time series analysis in domains such as healthcare, finance, energy systems, and environmental monitoring. However, these models remain vulnerable to adversarial attacks, where small input perturbations cause severe degradation in predictive performance. Commonly used gradient-based attacks, iterative first-order methods, are computationally burdensome, as they repeatedly backpropagate through the victim model to compute input gradients during a number of iterative refinement steps. We propose a GENerative ADversarial Attack (GENADA) that learns a generative model to produce deceptive perturbations directly in a single forward pass and a procedure to train it. Variants include single-step and iterative generative attack schemes. The validation considers attacks on several neural models and datasets in the time-series domain, a controlled, low-dimensional setting. Empirically, GENADA achieves comparable attack quality to strong baselines while requiring less time to generate perturbations during inference.

Motivation: LinkML is a suitable language for the representation of the structural and content constraints of different kinds of biomedical data. Even if it is a quite recent proposal, it has been applied in several biomedical contexts. Developing and maintaining LinkML schemas presents several challenges, particularly for novice curators. Non-expert bio-curators may struggle with LinkML syntax and best practices, requiring significant time and effort to develop well-structured schemas. Results: In this paper we propose SchemaLink, a web-based environment for the graphical construction and enhancement of LinkML schemas that address the following requirements: $(i)$ introduce a graphical language for the specification of LinkML schemas, $(ii)$ make uniform the specification of schemas in similar contexts, $(iii)$ simplify the design and curation processes by exploiting a RAG-based approach to assist curators in creating new schemas from scratch and editing already developed ones. Several experimental analyses show the quality of the produced LinkML schemas through the AI-based editing facilities. Availability and Implementation: SchemaLink is available online at: https://SchemaLink.biodata.di.unimi.it. SchemaLink code and testing data are available as open-source on GitHub at: https://github.com/AnacletoLAB/{schemalink-webapp,schemalink-api}.

LLM-based program evolution systems such as FunSearch and AlphaEvolve have shown strong ability to discover novel algorithms, but typically optimize each task in isolation, discarding search experience after completion. We introduce $\varepsilon$-MemEvo, a framework for cross-task knowledge transfer in LLM program evolution. $\varepsilon$-MemEvo stores prior experience as task-agnostic tactic memories: compact natural-language summaries of successful algorithmic strategies rather than raw code, enabling transfer across tasks with different APIs and evaluators. To avoid negative transfer from semantically mismatched memories, $\varepsilon$-MemEvo uses an adaptive injection gate that decides whether retrieved memories should be injected, and at what intensity. We evaluate $\varepsilon$-MemEvo on 8 diverse optimization benchmarks spanning mathematical optimization and systems engineering, using a content-level Leave-One-Out protocol that excludes target-task memory entries. On the primary GPT-5 backbone, $\varepsilon$-MemEvo improves AUCC over AdaEvolve on all 8 tasks, with a mean relative gain of +8.7%, and improves early-stage convergence by +9.4% on average. Ablations show that naive memory injection can fail catastrophically, while adaptive gating remains safe across all five ablation tasks. The data-updated posterior is interpretable in observed states: it favors skip during improving search and shifts from skip to hint across early and late plateaus. These gains incur less than 1% computational overhead.

Multimodal Large Language Models (MLLMs) have been increasingly adopted to automate webpage generation from visual designs (e.g., screenshots). However, existing evaluations are limited to visual fidelity assessment under a fixed browser-device configuration. Such a setting overlooks the cross-environment rendering compatibility for real-world deployments. To address this gap, we present the first systematic empirical study of cross-environment compatibility in AI-generated webpages. Specifically, we construct WebCompat, a dataset of 2,032 annotated instances, comprising webpages generated by 8 representative AI tools, each rendered across 9 browser-and-device combinations. We analyze the prevalence of compatibility issues, their user-perceptible symptoms, and underlying code-level root causes. Our findings reveal that 68% of generated webpages exhibit at least one compatibility issue, underscoring the pervasive reliability concerns surrounding MLLM-generated front-end artifacts. The most prevalent symptoms are failures that disrupt the entire page layout (88.3%): pages shrink directly to fit the target screen with too small fonts, or exhibit scale mismatches that produce cut-off content. Failures localized to individual elements, such as image distortion or missing components, are comparatively less common (13.4%). Furthermore, although most MLLMs incorporate responsive design patterns into the generation, they fail to properly implement these codes. Guided by the findings, we develop XCompat, a lightweight offline compatibility issue detector that combines visual screenshots and the structural DOM tree for analysis. It achieves an F1 score of 0.903 on the WebCompat-test, outperforming the existing compatibility checking tools and LLM baselines. All datasets and tools are released to support future research on rendering reliability in MLLM-based front-end code generation.

Deep graph neural networks(GNNs) suffer from oversmoothing- a progressive collapse of node representation towards a low information subspace as network depth increases because the normalized graph propagation operator is repeatedly applied directly to the hidden representations. In this work we study Householder Graph Neural Network (HouseGNN). Rather than updating the hidden state like standard GCN, HouseGNN uses the aggregated neighbourhood message solely to estimate a reflection direction; the node embedding is then updated by a Householder reflector followed by GroupSort, yielding a piecewise orthogonal layer that preserves Euclidean norm at every node and at every depth. We prove three core properties: (i) every internal layer preserves the node-wise Euclidean norm; (ii) the Householder reflector is scale scale and sign-invariant in the message; and (iii) pairwise distance between nodes can change through mismatch between node-wise orthogonal operators.

Privacy documents (e.g., privacy policies) are a central mechanism through which digital services disclose data practices and seek user consent. Over the past decades, research on privacy documents has expanded significantly, encompassing not only traditional privacy policies but also short notices (e.g., privacy labels) and interface-level transparency mechanisms. As this research area continues to grow, it has become increasingly difficult to obtain a coherent view of how privacy documents are created, analyzed, evaluated, and maintained across their lifecycle. This SoK provides a unified, lifecycle-oriented view of privacy documents from a software engineering perspective. We systematically review and analyze 290 papers published between 2010 and 2025, organizing them around five research questions that examine how privacy documents are (1) defined and scoped, (2) generated, (3) analyzed and extracted, (4) checked for inconsistencies and noncompliance, and (5) evaluated and improved for usability. Building on our findings, we identify 15 key research trends and 21 open opportunities. We further chart four broader research directions that highlight (i) emerging challenges in AI-centric platforms, (ii) the need for diverse and up-to-date data foundations, (iii) LLM-based unified policy-code analysis, and (iv) dual usability for end-users and developers. We hope this SoK provides a shared foundation for future research on privacy policies and privacy documents.

We describe a simple rejection-sampling-based algorithm to perform length-squared sampling on an $n \times n$ positive-semidefinite (psd) matrix: that is, to sample a column with probability proportional to its squared $\ell_2$-norm. The algorithm runs in just $O(n)$ expected time, which is significantly sublinear in the input matrix size. The runtime is optimal, even when the input is assumed to be diagonal. Our result has several applications. Length-squared sampling is used by a number of sublinear time algorithms for matrix problems, like low-rank approximation and eigenvalue approximation. Often, it is assumed that the algorithm is given access to the matrix column norms, and thus can perform length-squared sampling efficiently. Our result shows that, at least for psd matrices, we can remove this assumption. We also discuss an application to an asymptotically optimal algorithm for estimating the Frobenius norm of a psd matrix to relative error. Finally, we show that our sampling algorithm yields a very simple sublinear time algorithm for the robust psd low-rank approximation problem introduced by Bakshi et al. (FOCS, 2020), which nearly matches the more complex method developed there.

Organizations decide whom to treat under a budget and want to know what a targeting rule would have earned before deploying it. Off-policy evaluation promises this from logged data, but the deployable rule is a deterministic top-k policy: it removes all averaging over actions, so weak overlap hits the estimate directly. We benchmark six estimators across five datasets and two known-effect sweeps, and validate the mechanisms against a non-simulated paired reference. First, weak overlap is governed by logger-target action alignment, not by logging sharpness alone: what governs support is the logger's probability of the target's actions. Sharpening a logger built from the target's own score barely moves overlap over the tested range; action-level disagreement collapses it. Effective sample size ranks this risk across logging environments, but is weak at ranking candidates within the single log a practitioner holds, and its cut point does not transfer. Second, the optimizer's curse is not fixed by cross-fitting the outcome nuisance. When the rule is fit on the data used to evaluate it, cross-fitting the nuisance alone leaves the reuse bias in place and makes it worse. Honest policy-level splitting avoids the reuse by targeting the learning procedure's value -- a change of estimand, not a de-biasing of the full-sample policy. Third, propensity-estimation error is the largest degradation we measure: an out-of-fold estimate hurts IPS more than any other stress we apply, leaves doubly-robust estimation almost unchanged, and can invert the overlap diagnostic itself. Logging is synthesized and propensities floored at 0.02, so every failure occurs with bounded weights; the floor also reduces the two tuned hybrids to their untuned parents, leaving four practically distinct estimators, and all exact-value surfaces are synthetic or semi-synthetic. We release the benchmark; public data only.

Large language models (LLMs) cannot retain post-deployment experience without parameter updates. We introduce DIVE, a diversity-driven framework that enables frozen LLMs to improve by evolving persistent natural-language skills from task experience and verifier feedback. These skills encode reusable reasoning procedures, verification strategies, common failure modes, and output constraints and are both executed and revised by the same underlying model without access to a teacher model. Since natural-language skill evolution is a stochastic, non-convex search process, optimizing a single skill trajectory can overfit to sampled experience or converge to a suboptimal solution. DIVE mitigates this optimization variance by independently evolving multiple skill populations from bootstrapped experience, adaptively refining them through diverse transformations, and jointly selecting a complementary set of skills. Across six mathematical and logical reasoning tasks and multiple model families, DIVE consistently outperforms existing reasoning methods, prompt-optimization approaches, skill-development frameworks, and memory-based baselines. It achieves rapid self-improvement from accumulated experience, obtaining substantially larger performance gains with fewer rollouts than parameter-based methods such as SFT and GRPO, and prompt optimization with GEPA. Further, the resulting skills transfer across model scales and families, enabling smaller models such as GPT-5-nano to match or outperform larger counterparts, i.e., GPT-5, under conventional prompting. These results establish diversity-driven skill evolution as an effective, interpretable, and parameter-free approach to LLM self-improvement.

Clinical prediction models are often developed as if the outcome of interest were cleanly observed for every patient. This assumption fails when treatment decisions make the clinically relevant outcome permanently unobservable. As a case study of this problem, we consider post-cardiac-arrest neurological prognostication using a cohort of 2,497 patients, including 1,429 patients whose outcomes were rendered indeterminate by treatment decisions. These patients with indeterminate outcomes were reviewed by independent clinical experts, who provided their guesses of counterfactual outcomes about what would have happened to the patients. We refer to these patients as uncertain cases. We also have patients for whom we observe their clinically relevant outcomes; we refer to these patients as certain cases. We propose a framework for evaluating prediction models that explicitly splits the evaluation between certain and uncertain cases. Here, we cannot easily evaluate both types of cases in a uniform manner as the available target labels differ. We then propose a simple prediction model that uses target labels from both certain and uncertain cases in a manner that allows us to trade off between them. Across the proposed neural model and a collection of tabular baselines, models with similar certain-case AUROC can nevertheless differ substantially in both certain-case Brier score and their probability estimates for uncertain cases. Improving alignment with target labels of uncertain cases for our proposed model generally comes at the cost of worse accuracy on certain cases, highlighting an explicit tradeoff that standard evaluation conceals. These results show that when treatment decisions determine whether clinically meaningful outcomes remain observable, conventional evaluation metrics can miss important failure modes in the very patients for whom prognostic support matters most.

Long-term agent memory is usually treated as select--store--retrieve, but retrieval does not decide whether contradictory, superseded, retracted, deleted, or stale records may support an outgoing claim. We introduce Governed Persistent Memory (GPM), an auditable bitemporal state-transition model with source-bound admission, derived lifecycle state, current public barriers, and fail-closed structured release. Five executable clauses cover ledger integrity, source binding, conflict isolation, non-revival after retraction or deletion, and exact claim closure over a fresh view at one verified head. On a prespecified hash-frozen 3,600-case GPM-ReleaseBench, GPM matches all complete outcomes; the strongest of three intentionally simple complete policies matches 1,800/3,600 and makes unmatched releases on 50% of violation cases. A separate sealed end-to-end service evaluation exercises real ingestion and release across eight query families. In its publicly disclosed V3 arm, the governed lane is correct on 2,400/2,400 clusters versus 600/2,400 for ungoverned local Qwen2.5-7B; it repairs all 1,800 baseline failures with no regression (one-sided 95% lower bounds 99.875% and 99.834%). A later V5 reseal over Chinese- and English-command arms, with generation-date pinning and no post-freeze reducer amendment, again obtains 2,400/2,400 per arm. A production-code-independent finite model explores 331,776 semantic and 1,990,656 query states without a full-contract counterexample, and a 100,000-trace three-engine differential yields zero mismatches. These are bounded contract and implementation results, not open-world model accuracy or evidence of world truth. Governed answers in the sealed service evaluation are deterministic service outputs; the 7B result is the ungoverned comparison, not a claim that a language model itself became perfectly accurate.

Is it possible to find an objective classification criterion that organizes the complexity of gust-induced loads across many flight conditions? And one that remains as interpretable as a labelling based on coarse parameters, such as the flight attitude? Our approach encodes a large number of experimental observations through a machine-learned representation and applies a summarization procedure to select a minimal subset of highly significant exemplars. The exemplars provide a similarity-based objective classification criterion of all the observations, they can be more conveniently inspected by experts and can become subject of more refined experiments. We demonstrate the approach on a database of 3480 pressure-load measurements induced by random gusts on a flying-wing model across six flight attitudes. We find nine fundamental response types that recur across multiple attitudes; analysis of a type's transient response enables physical intuition into the underlying fluid mechanics.

Trained transformer models develop privileged bases: coordinate axes whose statistics differ from the rest of the residual stream. But what kind of direction does such a basis select? We investigate the prediction direction, the unembedding direction of the token a model currently predicts, and find that it functions as a content-defined privileged anchor. Measured with respect to this anchor, residual-stream variation is geometrically and behaviorally stratified by proximity to the prediction. The stratification holds in all eighteen models tested (dense and mixture-of-experts, 7B-120B, base and instruction-tuned). A narrow, scale-invariant prediction interface concentrates readout-relevant structure, while the vast prediction-distal complement expands with model scale. Because the prediction direction sits nearly orthogonal to the principal variance axes, variance-based analyses recover this organization only partly, and the shortfall grows with prompt heterogeneity. Anchoring reveals a steep geometric gradient: prediction-proximal regions are highly structured and cluster related prompts, while the complement is flatter and anti-discriminates among prompt groups. The interface is a narrow slice but functionally decisive. Disrupting the variance directions closest to the prediction causes immediate divergence and frequent task-frame shifts; disrupting the next level down delays divergence and preserves framing. The complement is weakly readout-aligned per direction yet causally and temporally load-bearing, and behavior is driven by direction rather than magnitude. These results establish the prediction direction as a privileged anchor distinct from previously described coordinate axes, and give a geometric account of how high-dimensional computation coexists with linear readout.

Sleep stage classification is important for the diagnosis and management of sleep disorders, yet most automatic staging studies evaluate models against a single reference hypnogram despite known inter-scorer variability. This study investigates whether multi-scored datasets can be used to construct more reliable reference labels from the collective behavior of multiple experts. We use the publicly available DOD-H and DOD-O datasets. EEG (C3-M2) and chin EMG signals were segmented into 30-s epochs, and 30 features were extracted from each modality, yielding 60 features for EEG+EMG. We propose a learning-based hypnogram (LBH) that models the stage-specific behavior of each scorer using confusion matrices derived from machine-learning models. After column normalization, these matrices estimate the probability of each true sleep stage given each scorer's label; probabilities are aggregated across scorers to assign the final label for each epoch. LBH was evaluated with random forest, support vector machine, and multilayer perceptron classifiers under EEG-only and EEG+EMG settings, and compared with the dataset hypnogram (DH) and best-scorer hypnogram (BSH). LBH consistently improved overall performance. The best results were obtained with random forest and EEG+EMG, reaching 86.07% accuracy, 85.46% precision, and 85.29% F1-score on DOD-H, and 86.04% accuracy, 85.21% precision, and 84.70% F1-score on DOD-O. These findings suggest that personalized scorer modeling can improve reference hypnogram construction without discarding information from individual experts.

An unconditional risk bound on automated decisions can be satisfied without automating anything, since a selector that never acts drives the bound to zero. We show this is structural: any risk certificate is defined over a decision contract, the inputs a system acts on plus the semantic relation under which an output counts correct, and weakening either hides base-classifier error. We develop a decision-contract theory: an error-conservation law showing error is only reassigned among harmful automation, human deferral, and semantic masking; a label-free singleton capacity certifying structural incapacity, with a risk-feasible refinement separating recoverable threshold misalignment from risk-constrained incapacity; and a non-degenerate actionability certificate excluding all-abstain solutions by construction. We instantiate this on ATT\&CK-aligned alert triage for LLM-based intrusion detection, the setting that exposed the vacuity failure. Across 3 IDS datasets, 6 LLMs, and 4 error-rate thresholds, empirical false-attribution risk stays at or below target in 90.3% of configurations, with 83.4% mean correct automation. The capacity diagnostic explains every low-utility configuration; its refinement separates genuine misalignment from risk-constrained incapacity, confirmed by an exhibited alternative threshold; a training-stability re-run finds no confirmed structural-incapacity instance; and real fine-grained attack-subtype labels confirm the coarsening-transfer identity under a genuine many-to-one map, with small but non-zero masking mass.

Large language models are increasingly trained and deployed with long contexts that span documents, code repositories, and interaction histories. This scaling reflects the implicit assumption that training on longer contexts will only help the model by exposing it to richer evidence. We challenge this view by studying how the context window shapes a model's mode of learning, shifting it between parametric internalization and contextualization. We propose the Information Abundance Paradox, which hypothesizes that abundant relevant information in the training context can reduce the incentive to encode that information parametrically, thereby increasing reliance on context. In pretraining with long documents, increasing the context window improves language modeling, natural language understanding, and closed-book MCQA only up to an intermediate optimum, after which performance consistently declines. In supervised fine-tuning, more task-relevant train-time context improves performance with supporting context, but reduces robustness when context is absent or misleading at test time. Our analysis suggests that this behavior arises when longer context provides a lower complexity solution. Mechanistically, training with informative context shifts gradient pressure from feed-forward networks, often linked to parametric knowledge, toward attention modules, and causal interventions show that this shift increases reliance on context during inference. Overall, these findings support the Information Abundance Paradox and suggest that scaling toward near-infinite context is not simply a matter of supplying more data, even when high-quality long-context data is abundant.

Neural combinatorial optimization (NCO) relies on parallel solution sampling for training, yet existing methods fail to fully exploit the rich information latent in a co-sampled solution group. Preference-optimization methods anchor on the single best solution and discard fine-grained quality and structural signal from all other peers-a failure we term gradient signal polarization. Mean-based baselines instead weight peers uniformly, so structurally near-identical peers flood the baseline with redundant information and keep gradient variance high-a failure we term baseline redundancy. We propose SSPO (Structure-Aware Similarity-Weighted Preference Optimization), which scores all $B$ sampled solutions jointly through a dissimilarity-weighted leave-one-out baseline: structurally distinct peers receive higher weight, resolving both failures in a single mechanism. The baseline uses zero-parameter, problem-adaptive solution embeddings built from the encoder's existing node representations. Experiments on TSP, EFL, and JSP benchmarks show consistent gains over prior best-anchor and uniform-weight baselines. A direct comparison against uniform RLOO on TSP and EFL confirms that structure-aware weighting is the primary driver of improvement. The SSPO-trained EFL policy has been deployed in a production facility-location system at JD$\mathord{.}$com, confirming practical viability at scale.

Deep learning detectors for anomalies in dynamic graphs have reached strong accuracy, yet they remain opaque: when an edge is flagged, the analyst receives a score but no reason. This opacity is untenable in the cooperative, regulated information systems where such detectors are deployed, where automated decisions must be auditable and trustworthy. We address this gap for AddGraph, the foundational GCN+GRU framework for edge-level anomaly detection in dynamic graphs, which to our knowledge has never been equipped with any form of explainability. We present a strictly post-hoc explainability framework, X-AddGraph, built on a Dual Spatial-Temporal Attribution (DSTA) mechanism whose three components are each aligned with one of AddGraph's architectural modules: a gradient-based relevance attribution over the current adjacency structure (spatial), a direct reading of the contextual attention weights already computed during inference (short-term temporal, at zero additional cost), and a gradient rollback through the recurrent hidden states (long-term temporal). Because the detector is frozen, detection performance is preserved exactly (Delta AUC = 0, verified empirically to ten decimal places). On the UCI Message benchmark, our trained AddGraph baseline reaches an average per-snapshot AUC of 0.8705, exceeding the originally published result; X-AddGraph reproduces every score identically while adding explanations where none existed. Evaluated across four edge populations - confident true positives, low-confidence true positives, false positives, and random samples - the long-term attribution identifies historical snapshots carrying significantly more counterfactual signal than random selection (0.127 vs. 0.074), a capability that no spatially-blind explainer can provide. We release our implementation for full reproducibility.

This paper reports a single, fully instrumented case study of a large-scale architectural refactoring by an AI coding agent under a specification-first protocol, with no human review of the generated code and no pre-existing oracle to validate the target behaviour. The task, dismantling a central invariant across a large interdependent codebase, was assessed by the author as effectively infeasible through incremental refactoring, the kind of change that conventionally calls for a rewrite instead. Under the protocol described here, the agent completed it successfully. The system is a 717,725-line production TypeScript application across 3,648 files. The task required dismantling a core lifetime invariant: the guarantee that a UI panel remains open for the duration of an AI request. The target behaviour was that a streaming generation survives the closing of its panel and can be reattached, on reopening, to the same live stream with no loss or duplication. The protocol: formal specification by the agent, 14 refinement cycles auditing that specification against the source code, atomic implementation, a compile/test feedback loop, then 17 verification cycles auditing the code against the frozen specification. Across 31 audit passes, 201 defects were corrected before any human executed the program. The convergence criterion was empirical: two consecutive verification passes returning zero findings. The change touched 189 files (31 new); with the extraction phase, the two commits total 288 files, 34,770 insertions, 16,422 deletions. Across the first and roughly thirty later sessions, the software behaved as specified, no bug observed. Elapsed: three days; cost: USD 2,430. The full specification and raw session logs, 1,500+ pages in French, are published as evidence, allowing inspection of the process and submission to a language model for consistency checking.

Training Mixture-of-Experts (MoE) models for reinforcement learning (RL) couples two load-balancing problems: sequence composition determines dense attention work in each data-parallel microbatch, while token routing determines sparse expert work on expert-parallel ranks. Optimizing either alone can shift the bottleneck to the other. In MoE RL, rollout-time routing replay exposes every sample's sequence length and layer-wise expert demand before its training step. We present RoutePack, a hierarchical planner that coordinates state-consistent, layer-wise expert rerouting with joint attention- and expert-aware data packing over an optimizer-step window. RoutePack first places experts independently at each MoE layer using aggregate routing demand. It then packs samples into the smallest certified, or best-known feasible, number of token-capped execution rows and optimizes their DP layout with a projected EDP-shard-aware objective. The objective combines a window-normalized linear-quadratic attention proxy with per-layer physical EP-rank peaks and minimizes the accumulated cost of the slowest EDP shard. Parallel population annealing searches fixed-row feasible layouts while preserving sample coverage, capacity, nonempty cells, equal microbatch counts, and communicator topology. State-consistent materialization preserves logical top-k routing and existing MoE kernels without microbatch-level expert replication. Across Ling-3.0-Tiny and Ling-3.0-Flash, expert rerouting improves mean trainer-measured token throughput by 3.80% and 10.50%, while routing-aware packing adds another 4.86% and 3.98%, respectively. Overall, RoutePack improves throughput by 8.85% and 14.89% over the baseline.

We formulate generative modeling as a path integral in which flow-based, diffusion-based, variational, and adversarial models arise as different evaluation principles for a single master action. Its Martin-Siggia-Rose-Janssen-de~Dominicis (MSRJD) form separates free from interacting probability flows and opens them to diagrammatic perturbation theory. The expansion yields a one-loop correction to deterministic samplers at no stochastic-sampling cost, which we validate on solvable and nonlinear drifts, where it reduces a 53 % tree-level error to 1.6 %. Imperfect learned scores enter as insertions and yield a response-weighted score-matching objective, and symmetry-equivariant drift design becomes an operator expansion with EFT power counting.

Federated learning (FL) enables collaborative model training across distributed clients while keeping data local. A central challenge is determining which client updates are beneficial for aggregation with respect to each client's target domain. Existing methods typically address this problem in parameter space by comparing model parameters or gradients. However, parameter-space similarity can be a poor proxy for predictive behavior, especially under heterogeneous, non-IID data. Consequently, updates that are misaligned with a client's target domain, including those caused by heterogeneous data or malfunctioning clients, may degrade local model performance. We propose Local Inference Guided Aggregation for Heterogeneous Training Environments to Yield Enhancement Through Agreement and Regularization (LIGHTYEAR), a federated learning framework that performs update selection in function space. LIGHTYEAR uses an NTK-based agreement score to characterize predictive behavior and determine a personalized aggregation set for each client. By relating model parameters to local predictive responses, the Neural Tangent Kernel (NTK) provides a more expressive criterion for update selection than parameter-space similarity alone. Because function-space information is not available before aggregation in conventional centralized FL, LIGHTYEAR uses a peer-to-peer (P2P) topology in which clients exchange updates directly and evaluate incoming models on private validation data. Each client selects only updates that are beneficial for its own target domain and aggregates them using a regularized rule that improves stability under heterogeneity. Across five datasets and nine baseline methods, LIGHTYEAR consistently outperforms centralized FL baselines and existing P2P approaches.

Multi-AUV ad-hoc network-based target tracking requires networked autonomous underwater vehicles (AUVs) to cooperatively track maneuvering targets under constrained acoustic communication, dynamic topology, and uncertain ocean disturbances. Although multi-agent reinforcement learning (MARL) enables decentralized coordination through centralized training, existing methods suffer from high-dimensional joint state-action modeling, noise-sensitive policy generation, leading to unstable training and degraded tracking. To address these issues, we propose VGG-MADiffRL, a value-gradient-guided multi-agent diffusion RL algorithm, and MDCA, a diffusion?based hierarchical control architecture. Leveraging underwater mission characteristics, we model sonar detection mechanisms and ocean current disturbances, formulating cooperative tracking for multi-AUV ad-hoc networks as an MDP. The proposed MDCA constitutes a three-tier closed-loop control framework: a global intelligent control layer, a local online training layer, and a physical action execution layer. This structure enables synergistic optimization across task allocation, local decision processes, and execution feedback. Within MDCA, the local online training layer is the policy learning framework; VGG-MADiffRL builds on diffusion policies and incorporates value gradients to guide action generation in the reverse denoising process, steering the generated actions towards higher expected returns. It employs twin value networks with joint optimization and soft target updates to mitigate overestimation and training oscillations, promoting more stable convergence. Experimental results show that VGG-MADiffRL consistently achieves faster convergence, higher tracking accuracy, and smoother training dynamics in cooperative tracking scenarios, validating its effectiveness and practical engineering value in dynamic underwater settings.

Protein language models (PLMs) have transferred the latest advances from natural language processing to computational biology. These models, trained on large corpora of protein sequence data, are widely used to translate amino acid sequences into latent-space embeddings, ready for use in diverse downstream tasks (DTs). By a common consensus, embeddings from the model's last layer are used, and the model's internal behavior remains poorly understood. We analyzed 13 PLMs across 15 DTs from 11 datasets to investigate the informativeness of embeddings created in intermediate PLM layers. We trained probe models on embeddings from each layer, compared their performance, and computed characteristics of the latent spaces they span to estimate the information they contain, and found that the last layers of PLMs rarely contained embeddings that led to the best results on downstream tasks. Furthermore, we identified a connection between DTs and the distribution across PLMs' layers of the relevant information to predict that task. For example, similarity between the pre-training objective and the objective of predicting properties of individual residues leads to a steady increase in understanding of such tasks across the layers of PLMs. On the other hand, for whole-protein tasks, we observe that the dataset, rather than the task itself, defines PLMs' ability to perform well on a DT. Embeddings from shallow layers of PLMs perform better for datasets that contain deep mutational scan (DMS) data, while datasets containing diverse natural proteins find most useful embeddings in the models' deeper layers. Additionally, we discover that the performance of PLMs drops significantly when tasks are introduced for artificial proteins.

Transformers owe much of their strong long-context retrieval capability to a token-level memory that grows with context length. This flexibility, however, incurs a quadratic computation complexity during training and a key--value cache that grows linearly during autoregressive inference. Recurrent alternatives offer efficient decoding by compressing the entire history into a fixed-size state, but often underperform on recall-intensive tasks since earlier associations usually get overwritten by subsequent updates, and only the most recent contextual information is retained. In this paper, we introduce Memory-Anchor Routing across Context History (MARCH), a network architecture that effectively scales state-space models beyond a fixed-size dimension, while maintaining computational efficiency over long-sequences. MARCH periodically caches cumulative recurrent-state checkpoints as state anchors and associates each anchor with a compact, content-conditioned anchor key. This lets MARCH maintain a memory bank, which can grow as context length increases, providing a controllable trade-off between historical resolution and memory cost. At each token, MARCH produces an anchor query to attend all causally available state anchors, and the output is calculated as an attention-style aggregation over all historical anchors along the current state. We show that after standard pretraining, MARCH consistently outperforms multiple linear attention variants across commonsense reasoning, LongBench, and in-context retrieval. These results demonstrate that content-routed state caching substantially strengthens recurrent long-range memory while preserving its native computation path.

Web agents often struggle to generalize to unseen websites because they lack website-specific supervision. Recent exploration-based data synthesis methods reduce manual annotation, but they still face two key limitations: they often fail to cover the full functionality of a website, and without sufficient website prior knowledge, they tend to propose hallucinated tasks, which in turn limits the diversity and efficiency of downstream trajectory synthesis. We present \textbf{SynWeaver}, a website-prior task-trajectory co-synthesis framework designed to address these challenges. SynWeaver first performs structured website exploration and constructs a website map that covers a broad set of functionally distinct page states and executable interactions on the target website. It then derives page-level and transition-level supervision from this map to train a UI-aware model with website-specific priors, enabling more grounded task proposals. Finally, SynWeaver performs collaborative task-trajectory synthesis, jointly updating the task and execution trajectory when they become inconsistent, and then verifies and repairs the collected results to produce executable, semantically aligned supervision. Experiments on WebArena and WebVoyager demonstrate that SynWeaver consistently outperforms strong synthesis baselines and yields more effective supervision for both in-domain and out-of-domain generalization.

Memory is a core component of AI agents, enabling them to accumulate experience, maintain personalization, and adapt over long-term interactions. However, existing memory systems often remain fixed after development, limiting their ability to adapt their memory models, organization strategies, and procedural knowledge through continued use. We present MindMemOS, a portable and self-evolving memory operating layer that organizes open-world information using a unified entity property timestructure. MindMemOS supports scenario-adaptive memory modeling, higher-order pattern discovery, autonomous memory refinement, and continuous skill evolution. Its MindMemEvolve algorithm employs validation-driven evolutionary search to optimize memory schemas for target scenarios, whiledreaming consolidates accumulated memories by merging redundant records and resolving conflicts. In addition, implicit corrective feedback serves as a human-in-the-loop signal for identifying and revising potentially inaccurate or misaligned memories. Its MindSkillEvolve algorithm further transforms agent execution trajectories into reusable and progressively refined skills. MindMemOS achieves 94.03% accuracy on LOCOMO and 70.63% on PersonaMem. MindSkillEvolve improves SpreadsheetBench success by 9.2 percentage points over the initial-skill baseline.

We construct OEIS Open, a benchmark based on 492 open mathematical conjectures from the OEIS, formalized in Lean by Tsoukalas et al. Whereas these conjectures had previously been attempted only with a bespoke agent, our open-source evaluation code runs any generic language model (LM) against them, and is secure against LM cheating attempts. We find that LMs equipped with a minimal set of tools resolve 147 of these conjectures with a budget of \$50 per attempt, scoring 30% on OEIS Open. OEIS Open Lite is a random subset of 100 conjectures for cheaper evaluation. When evaluated with a budget of \$200 per attempt, the best current LM scores 44% on OEIS Open Lite. Giving LMs access to the mathematics literature via 476,000 papers from arXiv did not increase performance on OEIS Open Lite, and nor did using more sophisticated agent loops. The conjectures covered in this work are of uncertain mathematical significance, and most have likely received little previous attention. Nevertheless, our results show that LMs can resolve open research conjectures autonomously and at modest cost.

Large language models are increasingly deployed in settings that require simultaneous adherence to multiple explicit constraints - reasoning structure, safety boundaries, output schemas. Individual constraints are handled proficiently, but the compositional regime, where many must hold jointly, remains poorly characterized: how rapidly does performance degrade, what governs the degradation, and can the collapse be mitigated? We introduce Constraint Saturation Evaluation (CSE), a procedurally generated benchmark that systematically varies the number of simultaneous constraints (k), with every constraint scored by a deterministic, rule-based verifier and zero LLM-judge involvement: 15 models, 36 constraint types, 369,753 checks at k=1-12. Three findings emerge. First, per-constraint pass rate decays gradually and predictably, while the chance of satisfying all k constraints collapses - a model passing individual constraints at ~41% at k=8 succeeds on all eight just 5.7% of the time. Second, constraints do not degrade equally: structural constraints lose 2x more baseline capability per added constraint than lexical ones, ordered by a comprehension-maintenance gap that separates constraints requiring sustained tracking from binary decisions immune to composition. Third, failures are nearly independent, which is what makes the accumulation multiplicative; the residual coupling that does exist tracks shared output features rather than pairwise interference - a wrong sentence count fails every constraint that reads it. Reliable instruction following breaks down beyond 5-6 simultaneous constraints: probe-level success falls below 50% at 7 constraints for the strongest model, and at 3 or fewer for 12 of 15.

A central promise of useful quantum advantage is the ability to compute ground states of Hamiltonian systems beyond the reach of classical simulation methods. Here we demonstrate that this problem can be effectively amortized across an arbitrary and universal set of Hamiltonians by a foundation model with $\sim0.5$B variational parameters, trained with contemporary techniques from large language models and deep reinforcement learning. To do this, we formulate $\text{spin-}1/2$ quantum ground-state learning as manifold variational optimisation over centrally odd scalar functions on $\mathrm{SU}(2)^N$. This replaces explicit Hilbert-space vector amplitudes with manifold functions on which the Hamiltonian acts through Lie derivatives, evaluated by custom automatic differentiation primitives. We prove that the resulting variational principle on this manifold preserves the $\text{spin-}1/2$ sector's ground-state upper bound using the Peter-Weyl theorem, then pre-train our foundation model on a dataset of hundreds of thousands of different Hamiltonian systems, varying the connection topology, system size, interaction types and strengths, bringing together a century of many-body literature. Using a novel $\mathrm{SU}(2)$ replica-exchange Langevin sampler and sharded natural-gradient optimisation, we train our model with our own extension of the Kronecker-Factored Approximate Curvature (KFAC) optimiser on system sizes up to 64 qubits. On a held-out generalisation dataset, we fine-tune our model on system sizes of up to 1024 qubits, and evaluate on systems up to 8100 qubits.

This study focuses on developing an AI-supported prototype for multiperspective interest rate forecasting that combines classical econometric models with modern artificial intel-ligence methods. Tested in a major European bank, the system enables more precise and flexible prediction of interest rate developments, supporting strategic decision-making in Asset-Liability Management (ALM). It integrates topic modeling, sentiment analysis, econometric forecasting, and market-based analyses within an interactive platform. Leveraging AI to analyze large volumes of financial documents and market data enables the identification of monetary policy trends and sentiment signals at an early stage. The core econometric model is a Bayesian vector autoregression (BVAR) that enables simulation-based scenario analyses to evaluate economic developments from multiple perspectives. The system's innovation lies in its integration of several forecasting approaches that consolidate previously separate information sources and present them transparently and interpretably. Financial analysts and risk managers thus gain a better basis for making decisions, allowing them to assess interest rate risks more accurately and manage market movements more proactively. While the prototype demonstrates how AI can transform interest rate management in banking, further development is required to optimize real-time data integration and regulatory compliance. Even at this stage, the study shows that multi-perspective, AI-driven forecasting provides substantial added value for banks by increasing transparency, strengthening evidence-based decision-making, and improving risk management.

Two free satellite signals carry real information about glacial-lake outburst risk in the Nepal Himalaya: radar interferometry sees a moraine dam slowly sagging, and satellite weather marks the weeks when a primed lake is under stress. A companion feasibility study found that deformation indicates which lake is destabilizing and weather indicates when it is at risk, but proposed no predictive model. To address this gap, we propose and evaluate models that predict which site is susceptible and when a trigger arrives. We test three related hazards on free data alone: large moraine- and ice-dammed bursts, rainfall-triggered landslides, and smaller floods from ponds on and around a glacier. Each hazard gets two questions, never blended. Using 589 dated outbursts from HMAGLOFDB and several thousand catalogued landslides, we match each event against similar but unfailed sites, and hold every model to a strong simple baseline under spatial cross-validation that withholds whole map tiles, so no model succeeds by recognising a trained-on neighbourhood. Antecedent weather times the trigger at ROC 0.73 for big bursts, 0.83 for landslides, and 0.82 for small floods. Terrain ranks susceptibility only in part: scored naively it appears near 0.9, largely because catalogued failures cluster in wetter ranges; matched against comparable nearby sites the honest figures are 0.76, 0.71, and 0.54 (no better than chance). The burst signal holds within single regions, reaching 0.89 in Nepal alone. Five deep-learning models do not decisively beat a simple gradient-boosted baseline. Three score marginally higher on landslides, a hint too small to confirm. For the lake hazards the baseline wins outright, reproduced by a three-rule decision tree on ruggedness and monsoon rainfall. We close with a ranked Nepal watchlist, a prioritisation aid, not a prediction, and note where free data reaches its limits.

We revisit the Sinkhorn-Knopp (SK) algorithm for the matrix scaling problem. Despite extensive literature on the global convergence of SK and its variants, its local linear convergence behavior remains less understood. We address this gap by providing the first nonasymptotic local analysis of SK that matches the rate obtained from existing asymptotic Jacobian-based arguments. We show that under certain connectivity conditions, SK is a polynomial-time algorithm for doubly stochastic matrix scaling. With the developed tools, we showcase the local suboptimality of SK and provide accelerated variants. Finally, for dense matrices, we improve the complexity of existing first-order matrix scaling algorithms from $O(\tfrac{n^{7/3}}{\varepsilon^{2/3}})$ to $O(\tfrac{n^{9/4}}{\sqrt{\varepsilon}})$.

Large language models (LLMs) have achieved remarkable breakthroughs across various applications. However, their architectures remain inefficient in pretraining due to two main limitations: (i) self-attention lacks an explicit inductive bias for locality, leading to redundant modeling of sequence-internal local information; (ii) mixture-of-experts (MoE) implicitly couples knowledge storage with computational pathways, hindering flexible access to sequence-external global knowledge. To overcome these limitations, we propose LoKiFormer, a novel LLM architecture that augments the standard decoder with two dedicated modules: 1) Local Fusion Attention (LFA), which incorporates a convolutional fusion to attention, explicitly capturing local patterns and allowing the attention to operate on more informative representations; 2) Knowledge Memory Module (KMM), which introduces a parametric key-value memory that explicitly stores global knowledge in addressable slots, decoupling storage from computation and enabling direct knowledge retrieval. Together, these modules enable LoKiFormer to achieve more efficient and effective integration of information at both levels. Experimental results show that LoKiFormer converges 1.33x faster in pre-training than baseline models, underscoring its superiority over existing LLM architectures.

On-policy distillation (OPD) trains a student on its own trajectories under dense token-level supervision from a teacher. Reward-extrapolation methods such as ExOPD amplify the teacher-reference log-likelihood ratio to move beyond direct imitation, but apply a single global coefficient $λ$ to every token. This can drive the student to fit extreme peaks in the implicit reward, causing reward hacking and unstable training, and the optimal $λ$ varies across domains, requiring costly sweeps. We propose REOPD, a reliability-adaptive reward extrapolation framework for OPD. REOPD combines a token-level compatibility weight with a batch-level adaptive budget, yielding a token-wise coefficient $λ_{b,t}=1+γ_b q_t$ that preserves teacher alignment while selectively extrapolating along reliable teacher-reference directions. It requires no verifier, reward model, value model, or extra rollout beyond standard OPD. REOPD outperforms G-OPD on single-teacher mathematics and on both domains in the multi-teacher setting, while matching G-OPD on single-teacher code, demonstrating effective fine-grained reliability adaptation across domains and teacher configurations.