Today's research clusters around three distinct methodological themes: first, a set of papers systematically measuring and closing gaps between model behavior and human intent, from LLM research ideation diverging from human taste to stealth biases hiding in logit distributions to verification architectures that surface unlicensed reasoning steps; second, a cohort examining how to extract efficiency from structure, whether through layer-wise RL adaptation concentrating gains in middle transformer layers, state-prediction separation in forward passes, or correlated sampling that reduces redundancy in parallel inference, each identifying that existing uniform approaches waste capacity on less critical components; and third, a cluster of applied systems papers that treat long-horizon reasoning, memory management, and resource allocation as independently learnable skills amenable to optimization loops, from agents learning to manage file systems across thousands of steps to compatibility rescue requiring coordinated cross-file changes to autonomous laboratories scheduling hardware subject to real-world constraints. Across these threads, the methodological signature is empirical decomposition: isolating a specific capability, measuring its contribution precisely under controlled conditions, and optimizing it as a separable objective rather than as a side effect of end-to-end training.
Cole Brennan
Showing of papers
LLMs are increasingly used to brainstorm research ideas, but existing evaluations mostly judge individual ideas by novelty, feasibility, or expert preference. We instead ask: how far are current LLM-generated ideas from human researchers? To characterize this gap, we build a large-scale evaluation framework for ideation from high-quality human research papers. For each paper, we reverse-engineer a small set of closely related prior works that likely inspired its core idea. LLMs are then prompted to generate a new idea from the set of paper titles and summaries. We introduce a two-axis research-taste taxonomy to profile each idea by its opportunity pattern and research paradigm, and use it to quantify the divergence between human and LLM ideas. Across idea sets generated by different LLMs, we observe a consistent distributional gap: LLM ideas are disproportionately concentrated around bridge-like opportunities and synthesis methods, whereas the human paper reference distribution spreads more broadly across ways of framing gaps and constructing contributions. This result suggests that strong LLMs can produce a range of reasonable ideas, but that range remains narrower than, and systematically shifted relative to, human research taste.
Reinforcement learning (RL) has become a central component of post-training large language models (LLMs), yet little is understood about how RL adaptation is distributed across transformer layers. Existing approaches typically update all model parameters uniformly, implicitly assuming that every layer contributes similarly to the gains obtained during RL post-training. In this work, we challenge this assumption through a systematic layer-wise study of RL training. Surprisingly, we find that training a single transformer layer can recover most of the gains achieved by full-parameter RL training, and in some cases even surpass it. To quantify this phenomenon, we introduce the quantity layer contribution, which measures the fraction of full RL improvement recovered by training a layer in isolation. Across seven models spanning two model families (Qwen3, Qwen2.5), three RL algorithms (GRPO, GiGPO, Dr. GRPO), and multiple task domains including mathematical reasoning, code generation, and agentic decision-making, we observe a remarkably stable pattern: RL gains are highly concentrated in a small subset of, and in many cases even a single, transformer layers. More strikingly, the same structural pattern consistently emerges: high-contribution layers concentrate in the middle of the transformer stack, while layers near the input and output ends contribute substantially less. The resulting layer rankings remain strongly correlated across datasets, tasks, model families, and RL algorithms.
Prior work on imitation learning from suboptimal demonstrations typically relies on compressed supervision signals such as confidence estimates, discriminator scores, or importance weights. These scalar signals are inherently limited, as they cannot explicitly express intermediate reasoning about task progress, failure modes, or corrective actions. We propose a language-critique framework for imitation learning from suboptimal demonstrations that instead leverages natural language as a structured supervision signal, avoiding the collapse of expressive feedback into scalars. Our method first constructs language labels from demonstrations that explicitly describe current progress, identify suboptimal behaviors, and provide fine-grained corrective guidance. We then introduce a language-critique loss that directly trains policies using these structured signals without reducing them to scalars, and instantiate it for both behavior cloning and diffusion policies, yielding LC-BC and LC-DP. We further provide a theoretical result showing that the proposed objective upper-bounds the expert performance gap under standard assumptions. Empirically, we evaluate on diverse continuous control tasks spanning navigation, manipulation, and gameplay, where our methods consistently outperform strong imitation learning and offline reinforcement learning baselines. These results demonstrate that language can serve as a powerful and structured form of supervision for learning robust policies from suboptimal data.
Memory expertise is a learned skill: knowing what to encode, when to retrieve, and how to organize knowledge--a capacity known in cognitive science as metamemory. We bring this perspective to LLMs by treating memory management as a trainable skill. We promote file-system operations to first-class memory actions alongside task actions, letting the model itself decide how to manage its memory. This memory skill improves along two axes: the structure that supports it (prompts, file schemas, action vocabulary), and the proficiency of the model exercising it. Both axes resist manual optimization: episodes in long-horizon tasks run for thousands of steps, and a single memory mistake can hide long before it surfaces, making human review of full trajectories impractical. We introduce AutoMem, a framework that automates both axes. In the first loop, a strong LLM reviews complete agent trajectories and iteratively revises the memory structure that shapes how the agent interacts with its memory files. In the second loop, the agent's own good memory decisions are identified from many episodes and used as training signal to sharpen the model's memory proficiency directly. Across three procedurally generated long-horizon games (Crafter, MiniHack, and NetHack), optimizing memory alone--without modifying the model's task-action behavior--improved the base agent's performance ~2x-4x, bringing a 32B open-weight model competitive with frontier systems such as Claude Opus 4.5 and Gemini 3.1 Pro Thinking. Our results show that memory management is an independently learnable skill, and a high-leverage objective yielding large gains on long-horizon tasks.
When should an AI system's answer be trusted? Formal proof assistants offer certainty but cannot reach most of the problem distribution; scalar LLM judges offer coverage but produce opaque scores that cannot be audited after the fact and are subject to the same coherence issues as any LLM. We present Theoria, a verification architecture that closes this gap. A candidate solution is rewritten into a sequence of typed state transitions, each licensed by an explicit justification, whether that be a citation, computation, or problem-given fact, and every transition is independently auditable. The foundational invariant is completeness of change: every difference between consecutive proof states must be accounted for, so hidden premises surface as unlicensed mutations rather than passing silently. On HLE-Verified Gold (185 text-only expert problems), Theoria certifies 105 at 91.4% strict precision (Wilson 95% CI [84.5%, 95.4%]). Every certification produces a human readable proof trace in which each step can be independently challenged. Holistic LLM judges achieve comparable precision at matched coverage but fail on different problems (Jaccard 0.14-0.36), making the approaches complementary. On 95 adversarial poisoned proofs across 15 domains, structured judges catch 94.7% versus 83.2% for holistic judging (p= 0.0017). The overall 11.5 pp gap concentrates in hidden premises (90.6% vs. 62.5%, a 28 pp difference) and fabricated citations (100% vs. 90%), the error classes where the formal analysis predicts an advantage; performance is identical on arithmetic and theorem-misapplication errors, where no advantage is predicted. On GPQA Diamond (n= 65), certified precision is 97.1% (Wilson CI [85.1%, 99.5%]).
Transformers use the same forward computation stream to both predict the next token and store useful state for future token predictions. We formulate the \emph{state-prediction separation hypothesis}: disentangling the two roles yields better language modeling performance. We design a Transformer variant that uses two computation streams to separate the two functions, and conduct pretraining experiments across various scales. Our experiments show that state-prediction separation consistently offers better data and compute efficiencies, improving validation loss and outperforming standard Transformers by 2--3 percentage points on average on downstream tasks. We also conduct extensive empirical analysis that rules out potential confounders and demonstrates the fundamental difference in the gradients our design entails.
Open-source libraries and tools are widely reused, but compatibility maintenance is expensive. Once maintainers leave, useful repositories can stop working as runtimes and dependencies evolve. We study whether LLM agents can adapt old repositories to modern environments, a task we call compatibility rescue. Unlike bug repair, compatibility rescue starts from a repository that worked in its original environment but fails after ecosystem drift. RepoRescue gives agents only the repository and its failing modern environment; the agent must diagnose the failure, locate affected code, and produce a source-code rescue that restores the historical test suite. We build RepoRescue from 193 Python and 122 Java repositories, each verified to pass historically and fail after modernization. We evaluate five deployed agent systems on Python and three on Java. Beyond full-patch pass rate, we rerun patches after removing test-file edits to measure source-only repair, add a runtime-enforced regime that blocks test edits, and validate practical use for repositories whose suites pass after rescue. We find that Claude Code systems sometimes edit failing tests even when prompted not to; with runtime blocking, Kimi still rescues 41.5% of repositories. Systems are complementary: their union reaches 62.7%, exceeding the best single system by 10.9 points. Difficulty concentrates in cross-file coordination: on 14 repositories requiring coordinated whole-codebase changes, GPT-5.2 through Codex passes all 14, while every Claude Code system passes at most two. Finally, a passing suite is only an initial signal: among 34 unmaintained Python candidates whose suites pass after rescue, 22 work in realistic scenarios and 12 pass bug-hunt with patches that address the compatibility failure. RepoRescue benchmarks compatibility rescue with source-only auditing, runtime enforcement, practical validation, and reasoning labels.
Current work on robot furniture assembly mostly focuses on toy-scale settings or single-arm manipulation. We introduce FurnitureVLA, the first systematic study of real-scale bimanual furniture assembly using Vision-Language-Action models (VLAs). We formalize the task, develop a scalable simulation pipeline for expert data generation and evaluation, and build a VR teleoperation system for single-operator bimanual control to collect high-quality real-world demonstrations. To address extreme long-horizon assembly with up to 7 subtasks and 1550 control steps, we propose a progress-enhanced VLA, finetuned on semantically grounded subtasks, that jointly predicts actions and a continuous progress signal, enabling automatic subtask transitions and reducing compounding errors during inference. We further study perception and control design factors that critically affect precision in real-scale assembly. FurnitureVLA improves average simulation success from 48% to 80% compared to baselines across three furniture types, with an additional 21% gain from our design factor study. We validate on a real Kinova Gen3 platform with only 16% drop on the hardest task.
Repository-level performance-optimization benchmarks such as GSO, SWE-Perf and SWE-fficiency evaluate coding agents by applying patches to real repositories and comparing runtime against unoptimized baselines and official reference patches. Their leaderboard scores are increasingly used as evidence of coding-agent progress, but those scores can conflate runtime instability, benchmark-specific scoring rules, and how many tasks are already solved by at least one public submission. We audit these issues across the three benchmarks. First, we replay the official reference patches for 740 code optimization tasks across four common types of Google Cloud machines. Most benchmark tasks can be replayed, but their reference patches satisfy the original benchmark validity rules in every cross-machine replay for only 39/102 GSO tasks, 11/140 SWE-Perf tasks, and 411/498 SWE-fficiency tasks; SWE-Perf is especially fragile because many reference patches produce close-to-zero runtime changes. Second, we show that public submission rankings depend strongly on the benchmark scoring rule. Among eight public submissions shared by GSO and SWE-fficiency, the official rankings disagree on 9 of 28 pairwise submission comparisons, and SWE-fficiency's leaderboard scoring rule assigns the worst ten tasks overly high score weights of 58.5%-82.8%. Third, looking across 10 public submissions for each task, we find that at least one submission matches or beats the reference patch on 85.3% (384/450) of replay-valid GSO and SWE-fficiency tasks, and beats the unoptimized base code on 99.8% (449/450). Our study complements leaderboard scores by identifying tasks with more reliable performance signals, quantifying per-task score contributions, and exposing the remaining performance gaps that are hidden by aggregate rankings.
Language models deployed in high-stakes roles can potentially favor certain entities, brands, or viewpoints, steering user decisions at scale. Such preferential biases can be introduced by any actor in the model's supply chain and are most dangerous when the model reveals its preference only on the relevant topic while behaving identically to its unmodified base on all other inputs. Recent work has shown that these biases can transfer through context distillation on semantically unrelated data, with the signal residing entirely in the soft logit distribution and remaining invisible to text-based inspection. However, the defender faces a fundamental asymmetry: without knowing the bias topic, no detection method can reliably surface a stealth preferential bias, regardless of whether it examines generated text, internal representations, or model weights. Here we introduce Distill to Detect (D2D), a method that surfaces hidden biases by distilling the distributional shift between a suspected model and its base into a cartridge (a KV-cache prefix adapter), concentrating the dominant divergence and amplifying the bias signal into generated text. We show that D2D successfully amplifies the hidden biases of stealth models to the extent that they can be reliably detected across multiple bias types. We also propose a theoretical framework that explains the efficacy of D2D through the lens of Fisher-weighted projection of the logit distribution shift, supported by empirical observations. By turning the capacity bottleneck of prefix-tuning adapters into a detection tool, D2D provides a practical building block for auditing hidden behaviors in deployed language models.
We introduce TiRex-2, a recurrent xLSTM-based time series foundation model that generalizes the univariate TiRex to multivariate forecasting with both past and future covariates. Real-world forecasting is inherently sequential: observations arrive continuously, variables evolve jointly, and a subset of covariates is known ahead of time. Existing Transformer-based time series foundation models capture cross-variate dependencies but incur quadratic complexity in context length and require full-history recomputation as new observations arrive. TiRex-2 addresses these limitations through a memory-centric recurrent design that operates at constant per-patch cost under streaming. The model combines a bidirectional time mixer with an asymmetric grouped-attention variate mixer, enabling the integration of future-known covariates while preserving strict causality over target variables. To our knowledge, this is the first time series foundation model that achieves this combination of properties. To support scalable multivariate pretraining, we propose a synthetic coupling pipeline that composes diverse multivariate samples on the fly from large univariate corpora. Empirically, TiRex-2 achieves state-of-the-art zero-shot performance on GIFT-Eval and fev-bench, remains stable when streamed to arbitrary context lengths, and maintains constant inference cost per patch. The model uses 38.4M active parameters in univariate mode, with an additional 44.1M parameters activated for multivariate forecasting.
This paper studies real-time robust optimal control for uncertain nonlinear systems, where linear time-varying (LTV) approximations make planning tractable but require sound linearization error bounds (LEBs) to guarantee robust constraint satisfaction. We develop tight, differentiable, GPU-parallel LEBs for LTV approximations of nonlinear and neural network (NN) dynamics. For analytic dynamics, we introduce path-based Hessian bounds that are tighter than standard interval methods. For NN dynamics, we derive certified LEBs using NN verifier-generated affine relaxations and local Jacobian corrections. We adapt a GPU-parallel system-level synthesis LTV-based robust control solver to be compatible with these LEBs by extending it to handle right-invertible disturbance matrices and non-zero-centered disturbance sets for tight zonotopic uncertainty propagation. Our method, GPUSLS-LEO, enables online optimization of robust feedback policies that account for linearization error, producing tight, formally verified reachable tubes. On complex nonlinear and NN dynamics up to 168 state dimensions, our method can compute robust control policies on the GPU at rates up to 67 Hz, reducing solve times and conservativeness relative to baselines while preserving formal guarantees and real-time performance.
We present World from Motion, a method for generating freely renderable dynamic 3D Gaussian representations from monocular videos. Our approach conditions a video model on dense, pixel-aligned renderings that encode appearance, geometry, and 3D scene motion along both input and target camera trajectories to correct rendering artifacts and fill in missing regions from an initial reconstruction. To train this model, we construct a dataset of aligned multiview video pairs and dynamic 3DGS representations, with simulated artifacts characteristic of monocular reconstruction. At test time, we distill the model's generations, including newly observed regions and motions, back into a single consistent, high-quality dynamic 3DGS, improving both novel-view synthesis and the underlying 3D motion. Our method sets a new state of the art in 4D reconstruction and seamlessly generalizes to in-the-wild videos with large viewpoint changes and dynamic motions.
Quantum computing has emerged as a promising computational paradigm for machine learning (ML), with the potential to offer computational advantages over classical approaches. At this stage, the evidence supporting the performance and advantages of quantum machine learning (QML) models relative to classical models is insufficient.To address this gap, this paper presents an empirical study on the performance of QML models and their classical counterparts. We compare seven model pairs spanning supervised learning and reinforcement learning. Our results indicate that the evaluated quantum machine learning models do not yet surpass the classical baselines in overall prediction performance, policy stability, or training time. Nevertheless, QML remains a promising approach for filtering noise and controlling false positives. Our research findings summarize the challenges facing quantum machine learning across hardware environments, training efficiency, and convergence stability, providing a foundation for research into the robustness and parameter optimization of QML. This work is publicly available at https://github.com/Z-537-437/QML.
In autonomous laboratories, AI agents suggest the next batch of experiments to do. However, planning and executing those tasks taking full advantage of the available resources is a completely different question. This can be challenging when dealing with real-world hardware constraints, especially so when there are multiple instruments with different capacities and throughputs. Here we demonstrate a 2-step method to address resource utilization for our autonomous platform for metal-organic framework synthesis. First, we use constraint programming to find optimal schedules. This finds schedules that minimizes the total time while still satisfying the limitations and capacities of the hardware. Secondly, we use a system of status dependencies for each task, which allows for the robust execution of the optimal schedules.
Combinatorial optimization (CO) problems are difficult because certifiable discrete structure induces exponential search. One needs to search over the set exponentially many candidates to certify optimality, however, the structural feasibility of a path, packing, or cover can be verified in polynomial time once supplied. In this study, we introduce Neural Certificate Pricing (NCP) that exploits this asymmetry under an unsupervised learning framework. A neural network is trained to predict certificate-level dual prices, while a structured recovery layer constructs the induced primal marginal. NCP can be viewed as amortized separation: instead of enumerating violated inequalities, it learns the residual prices through which their aggregate effect enters recovery. When the certificate-consistency condition holds, the recovered marginal is globally feasible, and a local theory shows that first-order errors in the predicted price induce only second-order loss in objective value. Across three classes of CO problems, NCP either outperforms state-of-the-art neural baselines by large margins or matches them at a fraction of the computation time, and shows stronger out-of-distribution generalization.
RL with verifiable rewards (RLVR) has emerged as a powerful paradigm for training LMs on tasks with well-defined success metrics, such as code generation and mathematical reasoning. However, current RLVR methods optimize only what can be objectively scored, often neglecting subjective, non-verifiable aspects of human-like outputs, such as style and structure. This limitation leads to well-documented failure modes such as diversity collapse, unnatural-sounding responses, and reward hacking. We propose an adversarial generator-discriminator framework that augments verifiable rewards with a learned signal from human demonstrations. A generator model is trained using RL to maximize both task accuracy and an adversarial reward derived from a discriminator. The discriminator, trained alongside the generator policy, learns to distinguish human-written outputs from model-generated ones. The discriminator serves as a learned proxy for the human output distribution, providing feedback on aspects of generation that are difficult to formalize as scalar rewards. Across diverse domains, including bug fixing and open-ended generation, our approach consistently improves non-verifiable properties while preserving the accuracy gains of RLVR. In bug fixing, our method produces solutions with significantly lower edit distance compared to RLVR baselines while matching end performance. In story generation, our method significantly improves win rate while producing stories that are diverse and more human-like. And in a simple reward hacking benchmark, our method nearly eliminates model misbehavior while maintaining high benchmark scores. Together, these results show that our approach bridges RL and SFT, offering a scalable path toward jointly optimizing the verifiable and non-verifiable properties of a task.
Scaling inference compute, by generating many parallel attempts per problem, is a costly but reliable lever for improving language model capabilities. By default these attempts are generated independently, wasting inference compute on redundant solutions. This waste seems unavoidable. After all, independence is what makes parallel sampling trivial to scale. However, this tradeoff is not fundamental: there is a rich design space of samplers that generate correlated but exact samples entirely in parallel. We explore this design space as an avenue for improving sample efficiency in scaling inference compute and reinforcement learning (RL). Concretely, we introduce QuasiMoTTo, which uses correlated samples as a drop-in replacement for i.i.d. samples. To generate these samples, QuasiMoTTo uses a reparameterization of autoregressive sampling as inverse-CDF sampling and draws the underlying uniforms with quasi-Monte Carlo (QMC); because QMC spreads the uniforms out more evenly than i.i.d., the resulting samples cover the output space with far less redundancy. Even though the batch is correlated, each sample is marginally distributed according to the language model, so we can use the batch for policy-gradient training. Our empirical analysis focuses on understanding how efficiently QuasiMoTTo can turn compute into performance. To evaluate correlated samplers, whose dependence breaks standard pass@k estimators, we first develop an unbiased bootstrap estimator. Across four reasoning benchmarks, QuasiMoTTo matches i.i.d. pass@k accuracy with 25-47% fewer samples. Strikingly, QuasiMoTTo often saturates an upper bound on pass@k that holds for any marginal-preserving sampler. We also apply QuasiMoTTo to policy-gradient RL (GRPO) where it matches i.i.d. performance with 50% fewer training steps. These gains come from higher coverage, which yields a stronger learning signal per batch.
Sample-based generative models are increasingly used for probabilistic forecasting in high-stakes decision settings, yet their training objectives are blind to the decision maker's cost structure. These models are commonly trained with strictly proper scoring rules, such as the energy score, which allocate their training signal in proportion to data density, with no awareness of where forecast errors are most costly for downstream decisions. We therefore propose decision-aware training for sample-based generative models, augmenting the energy score objective with a differentiable decision loss that directly penalises the cost incurred by acting on the model's forecast. This combined loss is theoretically grounded, as the decision loss is itself a proper scoring rule. We validate our method on one synthetic and two real-world tasks, showing targeted improvements in cost-sensitive regions while retaining full probabilistic forecasts.
Generative reasoning re-rankers achieve strong recommendation accuracy by emitting a chain-of-thought before re-ordering a candidate list, but they are slow at inference: an autoregressive (AR) decoder spends one sequential forward pass per reasoning token, and the reasoning trace far exceeds the ranking it produces. To reduce this cost, block-diffusion language models decode many positions in parallel over a few denoising steps and are substantially faster, yet naively converting an AR re-ranker into one opens two accuracy gaps: (1) a structural gap: answer positions are denoised in parallel and scored independently, so the decoder emits invalid rankings (duplicated, dropped, or out-of-set identifiers) that AR avoids through left-to-right masking; and (2) a distributional gap: fine-tuning the converted model on fixed teacher trajectories is off-policy relative to its own decoding at inference, leaving a residual accuracy gap. To close both gaps while keeping the speedup, we propose \textbf{Diffusion-GR2}, a recipe that converts our AR reasoning re-ranker (GR2) into a block-diffusion re-ranker. First, conversion fine-tuning (CFT) adapts the AR-initialized diffusion model to denoise the answer into a valid permutation on its own, without an external constrained decoder. Next, on-policy distillation (OPD) then supervises the model on its own decoded trajectories with dense per-token targets from the AR teacher. Finally, we apply a reinforcement-learning (RL) stage against a re-ranking reward on top of OPD's on-policy policy. Experiments on Amazon Beauty demonstrate that Diffusion-GR2 recovers to near-parity with the AR re-ranker, while block-parallel decoding raises decode throughput by $2.4$--$3.5\times$ at the model's reasoning output length. Ablations show that CFT recovers most of the conversion gap, and that on-policy distillation further closes it to the AR reference.
Recent work has shown that implicit neural representations (INRs) can be trained to effectively compress structured and unstructured volume data, allowing for direct data querying with a reduced memory footprint. However, as existing INRs for unstructured volumes do not encode geometry, they require partial mesh storage for later sampling, limiting achievable compression. At the same time, novel view synthesis methods have shown that explicit collections of 3D Gaussians can be used to accurately visualize volume data. In this work, we introduce an explicit model for volume data compression based on 3D Gaussian primitives. We reinterpret collections of 3D Gaussians as an explicit representation of a scalar field and use a sampling strategy that reconstructs scalar values at spatial locations through weighted aggregation of intersecting Gaussians. We develop optimized CUDA-accelerated pipelines for structured and unstructured model sampling, loss functions that encourage accurate domain encoding by our models, and a novel sampling-error based densification strategy. Our explicit formulation naturally encodes domain geometry, eliminating the need for mesh storage in unstructured volumes and introducing significantly higher compression opportunities. Compared to existing INRs, we demonstrate that our explicit model achieves competitive reconstruction quality with significant training speedups on structured volumes, while markedly outperforming in all metrics on unstructured volumes.
Cross-lingual speaker verification (SV) systems typically exhibit performance degradation when enrollment and test utterances are spoken in different languages. However, standard evaluation protocols confound language mismatch with inter-speaker variability, as evaluation is generally performed with different speakers across languages. In this work, we introduce a bilingual same-speaker evaluation set for five Iberian languages, enabling analysis of cross-lingual SV under constant speaker identity. We apply this setup to a HuBERT-based SV system previously shown to exhibit strong language dependence, and analyze results using the Cross-Lingual Transfer Matrix (CLTM) to study pairwise cross-lingual transfer. Our results show that speaker-related variability accounts for part of the observed degradation, but language mismatch remains the main driver of cross-lingual performance loss. These findings provide a more precise characterization of language dependence in cross-lingual SV.
Safety evaluations for language models increasingly depend on judgments about ambiguous natural-language behaviour: whether a model has followed an instruction, refused appropriately, complied with a policy, resisted an embedded command, or misreported progress in an agentic task. Existing benchmarks often compress these distinctions into pass/fail labels, obscuring whether failures arise from capability limits, policy ambiguity, instruction conflict, scaffold failure, or unstable evaluator judgments. This paper introduces adversarial pragmatics as a benchmark and annotation protocol for evaluating model behaviour under instruction conflict, embedded commands, quotation, scope ambiguity, deixis, indirect speech acts, and multi-turn agent transcripts. The contribution is empirical and methodological: a linguistically controlled taxonomy, an 18-item seed benchmark with validator-enforced metadata, a 54-row local seed pilot, an expert-evaluation protocol distinguishing task success, policy compliance, safety risk, refusal outcome, and evaluator confidence, and metrics for judge validity, diagnostic ambiguity, and taxonomy drift. The framework turns linguistic judgment methodology into a practical tool for validating safety evals, LLM judges, gold-set construction, prompt-injection tests, and safety documentation.
Creativity research has debated whether creativity is domain-specific (e.g., visual, writing, science), and if it is psychometrically separable from general intelligence. Both questions now apply to LLMs, but a unified benchmark of AI creativity remains elusive. We introduce AGC-Bench, an artificial general creativity benchmark built from a systematic review of the AI creativity literature (3,101 papers screened, 497 benchmarks identified), paired with an agentic harness that converts idiosyncratic codebases into HELM-standardized benchmarks. The first release covers 78 datasets spanning brainstorming, problem solving, STEM, narrative, figurative language, and humor. To address bias in LLM-as-judge, we apply Judge Response Theory -- a psychometric calibration of judge leniency/severity; we then fine-tune Qwen3-30B on the bias-corrected ratings of three frontier LLMs to produce AGC-Judge, an open-weight model that robustly scores new creativity benchmarks it was not trained on. Results reveal frontier models at the top of the AGC-Bench leaderboard, with open models close behind. LLMs show different creative strengths, ranking higher on some domains (e.g., writing) than others (e.g., scientific ideation). Extensive experiments yield three main findings. First, applying factor analysis across 83 LLMs, we recover a single creativity factor 'c', analogous to the 'g' factor of general intelligence, that explains 81.5% of variance, related to but separable from general knowledge/reasoning. Second, we show that prompting models to "be creative" boosts their performance far more than enabling reasoning, evidence that the benchmark tracks creativity over general ability. Third, on a human-matched subset, we find the top human still leads the top LLM on creativity. We release AGC-Bench with a public leaderboard, AGC-Judge, and human data as open infrastructure for measuring AI creativity at scale.
Data analysis in the medical domain often encounters scenarios involving a limited target dataset and a large, unannotated dataset with a general distribution. Under such circumstances, self-supervised learning (SSL) methods are highly effective for utilizing large datasets, making them a popular choice for electrocardiogram (ECG) analysis. This work presents the Event Reconstruction Joint-Embedding Predictive Architecture (ER-JEPA), a lightweight SSL framework for multivariate time series, whose name and two-fold hierarchical structure are inspired by the diagnostic approach of cardiologists. At its core, ER-JEPA features: (1) a two-stage structure that constructs representations for each time interval and subsequently processes these representations as a univariate time series, (2) the hierarchical integration of two Joint-Embedding Predictive Architectures (JEPAs), and (3) a Vision Transformer (ViT) backbone. The structural concatenation of two JEPAs categorizes the model as a Hierarchical JEPA (H-JEPA), designed to encode multiple levels of abstract representations for enhanced prediction on complex tasks. This study reports a successful application of H-JEPA to 12-lead ECG data as a multivariate time series alongside an analysis of the sensitivity of hierarchical representation during the pretraining stage. Pretrained on approximately 180,000 10-second recordings, the model achieves state-of-the-art downstream performance on the ST-MEM benchmark, with rapid computation and minimal resource usage.
While generative models have enabled training-free reward alignment, current methods typically excel in local exploration within narrow regions of the underlying distribution. These approaches struggle when preferences are unknown a priori and only revealed through sequential feedback-a scenario demanding broad exploration to uncover high-utility regions. To address this, we propose Sequentially-Controlled Interactive Multi-Particle Flow-Maps (IMPFM), a framework for sample-efficient online feedback-driven search. IMPFM progressively transports a group of interactive particles toward the target distribution, maintaining the broad coverage essential for heterogeneous preference alignment. IMPFM introduces a principled and efficient posterior sample sharing mechanism across particles powered by flow maps. By correcting individual particle drift with the collective posterior samples of the entire ensemble at each resampling step, the framework maximizes sample utility to enable global exploration while actively mitigating reward over-optimization, typical of standard control frameworks. Paired with a principled exploration-exploitation reweighting mechanism involving multi-particle interaction, this sequentially corrected multi-particle dynamics explicitly preserves structural diversity and overcomes the weight degeneracy inherent to standard SMC samplers. Crucially, we prove that the resulting sampling framework yields a multi-particle interaction-aware Feynman-Kac corrector that progressively steers the multi-particle system toward a KL-tilted target distribution, facilitating global exploration and preventing mode collapse. Extensive empirical evaluations and rigorous ablations across diverse search and alignment tasks confirm the efficacy of IMPFM over existing baselines.
Agent skills package reusable operational knowledge for Large Language Model (LLM) agents, yet as they grow in scope, they become dependency-bearing artifacts whose identities, versions, and provenance remain implicit. This opacity already causes duplicated dependencies and inconsistent installations, exposing a gap that dependency management has yet to close. We introduce Agent Skill Supply Chains (ASSCs) to characterize mixed skill-package-service dependency graphs and help close this gap. Borrowing from Software Bill of Materials (SBOMs), we design SkillDepAnalyzer to capture natural-language dependency evidence and model skills as dependency-bearing artifacts. On the SKILL-DEP benchmark, SkillDepAnalyzer recovers skill metadata and dependency graphs accurately and comprehensively, substantially outperforming an LLM-based baseline and package-centric SBOM tools. Applying SkillDepAnalyzer to over 1.43 million skills, we obtain ASSCs and explore their structural diversity and security signals. We find four structural patterns: skill metadata is activation-ready but governance-poor; dependency graphs span skill, package, and service dependencies with concentrated reuse; recursive skill reuse expands dependency graphs and creates hidden package inventory; and skill dependency clusters form around related workflows. We also find that inspecting a skill alone misses security-relevant signals hiding in its dependencies. By analyzing ASSCs, we identify and report known malicious skills persisting in ASSCs to their developers. Based on these findings, we recommend typed dependency manifests, first-class dependency-cluster management, risk-warning audit commands for skill infrastructure maintainers, and lockfile-like records for skill developers.
Autonomous scientific discovery systems offer the potential to accelerate research by automating the process of hypothesis generation and validation. However, current systems operate within constrained search spaces or require predefined research questions, limiting their capacity for true open-ended inquiry. Furthermore, while they generate hypotheses iteratively, they largely lack the ability to explicitly synthesize their own accumulated findings to uncover complex, interconnected phenomena. We introduce DiscoPER, an autonomous large language model-powered framework that conducts open-ended research by dynamically generating and executing code to explore datasets without pre-specified research objectives. To ensure rigorous scientific validity, every proposed discovery must pass statistical testing. To overcome the limitations of isolated search, our framework introduces a second-order reasoning mechanism that periodically analyzes its own accumulated discoveries. By treating prior discoveries as empirical data, DiscoPER identifies structural patterns, confounds, and epistemic gaps, actively redirecting hypothesis exploration toward uncharted regions of the search space. The search space is further expanded by incorporating tool use, enabling the system to explore hypotheses beyond structured metadata by seamlessly processing and extracting useful information from multimodal sources like images. Evaluated on iNatDisco, a new multimodal ecological knowledge benchmark with pattern-level ground truth obtained from peer-reviewed literature, DiscoPER recovers 8 of 9 known patterns with a 72.7% hypothesis support rate, outperforming both classical causal discovery and LLM-guided baselines. Ablations show that DiscoPER scales with more data, and confirms the benefits of second-order meta-reflection.
Operator learning for partial differential equations (PDEs) on arbitrary geometries builds fast neural surrogates for large-scale simulation. Although recent geometry-adaptive neural operators have made substantial progress, they are mainly designed for forward problems in which inputs and outputs share the same spatial domain. This limits their applicability for boundary value problems (BVPs) and inverse problems, where inputs and outputs may live on different domains. We introduce the Geometry-Adaptive Integral Autoencoder (GAIA), an operator learning model that encodes the domain boundary and the interior field distribution into geometry tokens, and conditions integral transform layers on these tokens via cross-attention, allowing the kernel to adapt locally to geometric features. This yields a single architecture for forward (including BVPs) and inverse problems on arbitrary domains in one pass, without retraining, iterative optimization, or graph construction. We evaluate GAIA on seven 2D and 3D benchmarks, four of which are new or substantially extended benchmarks for inverse problems and BVP: electrical impedance tomography, optical tomography, 3D Darcy flow on varying geometries, and a modified setting of Poisson BVP on mechanical components benchmark (MCB). GAIA sets new state-of-the-art results on every inverse and BVP task, reducing median relative $L^2$ error by 64% on airfoil flow reconstruction and 27% on EIT relative to the next best amortized method, and outperforming all baselines on every shape category of MCB. On other forward problems, GAIA is competitive with specialized solvers while maintaining stable accuracy across point resolutions on which transformer-based baselines degrade.
Quantization has become an invaluable tool to reduce memory requirements and inference speed of modern language models, in particular to make them available for consumer setups and edge devices. While previous work has primarily focused on uniform quantization codebooks, such approaches are prone to suboptimal representations due to low-frequency high-magnitude weights. We introduce Log$_\text{b}$Quant, a novel logarithmic quantization approach with adjustable bases, to adapt to common parameter distributions. We show that our method exhibits superior performance at 4-bit precision on several performance benchmarks compared to asymmetric linear quantization at tensor-wise granularity, while achieving moderate speedup and high memory savings, making it suitable for private use on consumer-grade GPUs.
Zeroth-order (ZO) optimization enables fine-tuning large language models when backpropagation is unavailable or memory-prohibitive, but existing methods often perturb full model weights or randomly constructed low-dimensional subspaces, yielding high-variance estimates and limited performance. We propose ZO-Act, an activation-informed ZO fine-tuning method that restricts perturbations to a fixed low-rank subspace derived from input activations. For each linear layer, ZO-Act computes a small activation basis once at initialization and optimizes only lightweight coefficient matrices using forward-only loss evaluations. This reduces the effective perturbation dimension, exposes explicit trainable variables compatible with momentum-based optimizers such as Adam, and naturally supports quantized LLM fine-tuning by keeping low-bit weights frozen. We analyze ZO-Act as zeroth-order optimization over a restricted coefficient space and show that perturbing the low-dimensional coefficients reduces both the variance-dependent convergence term and the finite-difference error of the ZO estimator, at the cost of a controlled subspace approximation bias that is mitigated by the low-rank structure of LLM activations and gradients. Experiments on Llama-3-8B, OPT-13B, and INT4 Llama-3-8B show consistent gains over strong ZO fine-tuning baselines across language understanding, question answering, and commonsense reasoning.
Muon has recently emerged as one of the most effective optimizers for training large neural networks, yet its empirical success has been explained from several different perspectives. In this paper, we propose a simple mechanistic interpretation: Muon can be understood as an implicit residual connection during training. Specifically, orthogonalizing the update can sacrifice some immediate gradient fidelity while improving representation preservation for downstream layers. We study this trade-off in controlled linear optimization settings, where Muon can learn representations that are slower to fit a local target but easier for downstream layers to exploit. Our results suggest a conceptual explanation for Muon and a design perspective for optimizers that balance local descent with downstream usability.
University stakeholders often face difficulties in accessing timely and reliable information, especially in developing countries, where there are very few intelligent support systems. Existing rule-based chatbots are unable to handle complex, domain-specific queries and are not well-equipped to adapt to evolving institutional policies. As a fill-in-the-gap solution, we present the multimodal university chatbot with retrieval-augmented generation. The system combines the large language model with semantic retrieval to produce context-based responses from institution-centric resources, such as the university handbook. The system accepts text and image queries through the vision-language model and applies quantized inference for rapid deployment on constrained hardware. A scalable backend built with FastAPI, adjoined with a responsive frontend developed with Next.js, ensures real-time usability. Our multimodal evaluation demonstrates that the system maintains strong satisfaction scores across both text and image queries, despite increased response time for visual inputs. Furthermore, quantitative evaluation shows that hallucination is reduced from 31.7% to 6.6% in our proposed RAG-based system, confirming the effectiveness of retrieval grounding.
Robot policies inevitably encounter failures when deployed in real environments. Naive retries often repeat the same mistakes, while many existing recovery methods rely on human intervention. In this paper, we propose Failure-Aware Retry (FAR), a framework that enables robots to learn from previous failures at test time, adapt their behavior accordingly, and eventually complete the task autonomously. FAR combines Failure-Contrastive Preference Adaptation, which constructs preference learning data from failures to steer the policy away from previously unsuccessful behaviors, with lightweight action perturbations during retries to encourage local exploration. We further incorporate successful recovery trajectories into a training loop for continual policy improvement. Experiments in both simulation and real-world manipulation tasks show that FAR substantially improves success rates and robustness, with average gains of 17.6% over the standard diffusion policy in simulation and 11.7% in the real world. In addition, FAR significantly improves data efficiency under both reset and timestep budgets during continual policy improvement by exploiting informative failure cases.
We present SynLaD, a latent diffusion framework for small-molecule generation that unifies ligand-based drug design objectives (what to make) with synthetic accessibility (how to make it). Current models typically optimize one objective at the expense of the other, creating a bottleneck for discovering high-scoring and synthesizable molecules. SynLaD combines reaction-constrained generation with pharmacophore-conditioned 3D design by learning a latent space that decodes to both 3D structures and synthesis pathways. An encoder maps molecules to a latent representation used by two decoder heads: (i) a geometric head that reconstructs atom types and coordinates and (ii) an autoregressive synthesis head that outputs synthetic routes in a serialized, reaction-based notation. A diffusion transformer generates novel latents in the learned space, conditioned on pharmacophore profiles. Across analogue generation tasks for bioactive ligands, SynLaD outperforms existing baselines in synthesizable and diverse hit generation, demonstrating that a single model can produce shape-aligned molecules with feasible synthesis plans.
In Large Language Model (LLM) training, data mixing plays a pivotal role in determining model performance. Recent methods optimize mixture weights via proxy models, but they rely on the assumption of static data distributions. As a result, when the underlying data pool shifts, these methods require costly retraining from scratch. This limitation restricts their ability to scale seamlessly from small settings to larger data pools and model sizes. In this paper, we propose CausalMix to address this limitation by casting data mixture optimization as a causal inference problem. We formulate the statistical features of the data pool as covariates and the domain mixture as the treatment. After fitting a causal model on 512 runs of Qwen2.5-0.5B to estimate the Conditional Average Treatment Effect (CATE), we extrapolate the optimal mixture for an 800K data pool and apply it to train a 7B model. Furthermore, we successfully generalize the framework to long chain-of-thought data on Qwen3-4B-Base. By leveraging causal modeling to isolate confounding biases, CausalMix dynamically infers state-dependent optimal data mixtures. Extensive experiments show that the mixture guided by CausalMix consistently improves performance across multiple downstream tasks, outperforming RegMix and other baselines. In addition, we use the CATE Interpreter to provide visual analysis of the learned mixing strategy. Overall, CausalMix offers a causal and interpretable framework for optimizing LLM data mixtures.
Open-response evaluation provides stronger clinical validity than multiple-choice benchmarks but creates a scoring bottleneck that motivates automated LLM-asa-Judge approaches. Whether such evaluators replicate clinical calibration and caution, however, remains untested. We introduce MedQADE, the first standardised open-response clinical benchmark for German, a major clinical language lacking native evaluation infrastructure, comprising 3,800 items annotated by ten practising physicians and nine Large Language Model (LLM) evaluators. The top-performing evaluator model, Gemini 3 Flash, reached alignment consistent with the physician ceiling (\k{appa} = 0.694 vs. \k{appa} = 0.709), though wide confidence intervals limit interpretation. Despite this statistical alignment, automated evaluators exhibited near-absent clinical metacognition: physicians scaled abstention with item difficulty, while frontier models assigned definitive scores in every case. We additionally quantified systematic lineage-dependent biases, where models preferentially scored architectural siblings, an effect independent of language. These results show that statistical alignment does not ensure clinical caution, and that evaluator independence requires explicit verification.
Software diagrams are difficult to edit through human-friendly interfaces because edits expressed in natural language must still preserve visual layout, editable structure, and semantic relationships. As a step forward, we present SAGE, a browser-based tool for prompt-guided editing of Draw.io and Mermaid-style engineering diagrams. The tool maps diagrams into an editable graph representation, translates natural language requests into structured edit intents, analyzes those intents into graph-oriented operation steps, validates and repairs common Draw.io XML issues, and stores successful results as recoverable versioned artifacts. This design separates structured state management from model-driven interpretation, while acknowledging that some prompt-guided XML edits remain model-assisted. The tool also supports direct canvas editing and a secondary mask-based image-editing workflow. We evaluate the system using unit tests and a Kubernetes architecture case study, measuring structural validity, edit success, preservation of unrelated elements, and failure causes.
Task-based chatbots are nowadays widely adopted software systems, usually integrated into real-world applications and communication channels, designed to assist users in completing tasks through conversational interfaces. Like any other software, even chatbots are prone to bugs. Despite their increasing pervasiveness in everyday activities, existing techniques for assessing their quality still exhibit several limitations, such as the simplicity of generated test scenarios and oracle weaknesses. In this paper, we present LexTester, an automated model-based testing technique for Amazon Lex chatbots. The technique explores the conversational space of the chatbot under test to generate a Dialog Graph of all possible interactions, from which an executable test suite is generated according to different coverage strategies. LexTester was evaluated against the state-of-the-practice chatbot testing tool Botium on five Amazon Lex chatbots, consistently outperforming it in all subjects, generating more tests with nearly double complexity, achieving overall 83-95% coverage of conversational elements, and improving fault detection effectiveness by up to four times at comparable time costs.
Active learning reduces labeling cost by querying the most informative unlabeled samples, but standard coreset methods ignore known data symmetries and can waste budget on transformed versions of the same instance. We propose GRINCO, a group-invariant coreset framework that performs acquisition in the quotient space induced by a transformation group, so that selection operates on orbits rather than raw samples. The method uses either canonical representatives or learned orbit-separating invariant embeddings to define practical quotient metrics, and combines quotient-space k-center selection with invariant training through an orbit-averaged loss. We further derive a generalization bound that relates excess orbit-averaged risk to quotient-space coverage, label uncertainty, and intra-orbit variability. Experiments on synthetic scale-invariant data and image benchmarks with rotation-induced redundancy show that GRINCO improves orbit coverage and achieves stronger label efficiency than conventional coreset baselines, especially when group-induced redundancy is substantial.
Generative AI is shifting software engineering from a practice organized around scarce implementation effort toward one organized around abundant, low-cost code production. This shift changes the central engineering problem: not whether AI can generate useful code, but how engineers organize architectures, tools, evidence, and feedback loops so that AI-mediated development remains inspectable, correctable, and maintainable. We study this problem through a first-person case study: a 12-week development effort in which a single expert software engineer used frontier AI coding agents to build a document accessibility remediation system. The empirical record comprises 88 contemporaneous field notes, 420 KLOC of production code, and 1.16 MLOC of tests, lints, supporting documentation, and agent tooling. From this record, we develop a candidate middle-range theory of governance conversion, expressed as a process model explaining how high-velocity agentic implementation becomes governable. The model explains how agentic implementation velocity surfaces recurring structural failure classes, and how engineering judgment sustains velocity by converting those failures into durable governance mechanisms. In contrast to existing governance models that derive controls from known obligations, governance conversion explains how controls are discovered from failures that become visible only during agentic work. We use our model to make testable predictions and to describe implications for software engineering research and practice.
The evaluation of long-term video quality understanding remains an open challenge for large vision-language models (LVLMs). Existing video quality benchmarks predominantly focus on short clips and isolated distortions, overlooking the temporal continuity, cumulative degradation, and reasoning complexity inherent in long-duration content. To address these limitations, we present LongVQUBench, a comprehensive benchmark for long-term video quality understanding. LongVQUBench contains over 1200 diverse videos spanning movies, documentaries, surveillance footage, egocentric recordings, and animated content, accompanied by 1500 multiple-choice and open-ended questions for validation and testing. To assess perceptual reasoning across different temporal scopes, we introduce three progressively complex evaluation levels: (i) local event quality understanding (LQU) for analyzing localized distortions; (ii) cross-event quality reasoning (CQR) for integrating multiple degraded events; and (iii) global quality understanding (GQU) for holistic perceptual evaluation over extended durations. Furthermore, a needle distortion question-answering (NDQA) paradigm is embedded across all three levels, where spatial or temporal artifacts are sparsely inserted to probe fine-grained detection and reasoning capabilities. Extensive experiments on 14 state-of-the-art LVLMs reveal significant performance degradation with increasing video length and reasoning depth, highlighting their limited capacity for long-range temporal integration and perceptual attribution. We envision LongVQUBench as a foundational step toward the systematic, hierarchical, and explainable evaluation of LVLMs' long-term video quality understanding.
While Large Language Model (LLM) agents demonstrate proficiency in static benchmarks, their deployment in real-world scenarios is hindered by the dynamic nature of user queries, tool sets, and interaction dynamics. To address this generalization gap, we formalize OpenAgent (Tool-Use Agent in Open-World), a problem setting characterized by distributional shifts across query, action, observation, and domain dimensions. To systematically diagnose its impact, we construct a controlled sandbox environment where we define fine-grained environmental shifts across a four-tier hierarchy, Perception, Interaction, Reasoning, and Internalization, and conduct a comprehensive series of experiments. Our analysis yields a series of key insights, demonstrating that agents trained via both Supervised Fine-Tuning(SFT) and Reinforcement Learning suffer from varying degrees of performance degradation when confronting open environmental shifts. Building on these insights, we propose Perturbation-Augmented Fine-Tuning, a disturbance-based intervention strategy for SFT that lays the foundation for enhancing agent robustness and utility in realistic environments. Our code will be released at: https://github. com/LAMDA-NeSy/OpenAgent.
High-throughput RLHF systems often decouple rollout generation from policy optimization, leading to the use of stale rollouts during learner updates. In this work, we study the effect of such staleness in asynchronous GRPO. We make the behavior policy explicit in the GRPO surrogate objective and distinguish between the surrogate-gradient mapping used by the learner and the true total derivative of a distribution-dependent population objective. Under assumptions of local boundedness, distributional smoothness, and behavior-policy smoothness, we show that stale rollouts introduce a per-step surrogate-gradient bias of order O(S * eta), where S denotes the maximum rollout lag and eta denotes the learning rate. We further derive a conditional collapse-time scaling law: when within-cycle drift remains below a batch-level clipping radius, collapse is governed primarily by cumulative learner drift T * eta; when the stale-rollout constraint is active, stability instead depends explicitly on S * eta. This yields a two-constraint stability condition eta << min{R_batch / (S * G_upd), R_crit / (T * G_upd)}, explaining why the maximum stable learning rate may appear weakly dependent on staleness in the horizon-limited regime.
Spatio-temporal point-process models must often generalise across space when local event histories are sparse. We study whether exogenous spatial context can compensate in such regimes. Using a fixed log-Gaussian Cox process backbone, we compare an event-only model with the same model augmented by AlphaEarth embeddings as linear spatial context. We evaluate spatial transfer on emergency medical services (EMS) forecasting across eight held-out regions, fixed forecast anchors, and a sweep over history length $w$, using only AlphaEarth (AE) embeddings available strictly before each anchor. AE improves out-of-region predictive performance across all history regimes, with the largest gains under scarce histories: approximately $2$--$6\times$ multiplicative improvements at $1-2$ weeks, tapering to roughly $10$--$20\%$ at $w=20$--$104$ weeks. These results show that contextual information can substantially stabilise spatially transferred point-process forecasts when event history is limited.
We investigate Gaussian process (GP) bandit optimization with quantum kernels, assuming the mean reward function lies in the reproducing kernel Hilbert space (RKHS) induced by the quantum kernel. This setting is motivated by NISQ-era tasks such as quantum control, state preparation and variational quantum algorithms. While quantum kernels can offer a `quantum advantage' via domain-specific inductive biases, naïvely using full, high-dimensional kernels increases model complexity and information gain, leading to higher cumulative regret and poor learnability. To address this, we propose projected quantum kernels and classical kernel approximation techniques that reduce feature dimensionality while preserving key quantum properties. Using these approximate kernels, we develop misspecified GP bandit algorithms and derive regret bounds that characterize the trade-off between approximation error and information gain. The regret bounds provide principled guidance for selecting the optimal model complexity. Empirically, our methods outperform full quantum kernels in sample efficiency, while substantially reducing computational overhead, enabling scalable GP optimization for quantum-native applications.
While inference-time scaling has improved the reasoning abilities of large language models (LLMs), the need to generate long chains-of-thought (CoTs) is a computational bottleneck. Thus, in contrast to sequential scaling methods like CoT, recent parallel scaling techniques instead use fork and join (FJ) primitives to divide work across multiple LLM threads. However, in the fork-join paradigm, threads are typically transient and do not communicate pointwise with one another which limits scalability. To tackle this, we introduce Message Passing Language Models (MPLMs), a framework for LLM reasoning in which threads communicate directly via lightweight send and receive primitives. MPLMs enable efficient scaling through two key mechanisms: (1) reduced communication costs, achieved by avoiding redundant context sharing, and (2) preemption, which allows threads to terminate early based on partial information from their peers. We demonstrate the promise of MPLMs on 3 classes of tasks. First, on Sudoku puzzles, we show that MPLMs require an asymptotically smaller context than both serial CoT and parallel FJ. We then fine-tune a single model to solve 25 x 25 puzzles that remain challenging for standard CoT and FJ approaches, as well as frontier reasoning models without tools. Second, on 3-SAT puzzles, the capability of preemption allows termination of unpromising branches, which results in improved efficiency. Finally, we show that appropriately prompted large pre-trained models follow the MPLM protocol, achieving competitive results on long-context question answering relative to popular fork-join approaches.
Memory has emerged as a cornerstone of modern LLM-based agents, supporting their evolution from single-turn assistants to long-term collaborators. However, memory is not always beneficial: retrieved memories often induce a critical issue of sycophancy, causing agents to over-align with the user at the cost of factual accuracy or objective reasoning. Despite this emerging risk, existing memory benchmarks primarily evaluate whether memories are correctly stored, retrieved, or updated, while overlooking how retrieved memories influence downstream reasoning and decision-making. To bridge this gap, we propose MemSyco-Bench, a comprehensive benchmark for evaluating memory-induced sycophancy in agent systems. MemSyco-Bench measures when memory should influence a decision and how valid memory should be used. Specifically, it covers five tasks that assess whether agents can reject memory as factual evidence, respect its applicable scope, resolve conflicts between memory and objective evidence, track memory updates, and use valid memory for personalization. All related resources are collected for the community at https://github.com/XMUDeepLIT/MemSyco-Bench.
The deployment of Large Language Models (LLMs) with extended context windows is increasingly constrained by the linear growth of Key-Value (KV) cache memory. Vector Quantization (VQ), particularly Residual Quantization (RQ), is a promising approach for pushing KV cache storage toward the sub-1-bit regime by progressively encoding residuals with small codebooks. However, most VQ methods still rely on standard $\ell_2$ $K$-means as the core codebook-learning primitive. We identify a subtle high-dimensional issue of this primitive: Euclidean centroid averaging can induce centroid shrinkage, which weakens the angular alignment term in the $\ell_2$ distortion and makes directional preservation harder. To address this issue, we propose Gain-Shape $K$-means (GSKM), a drop-in replacement for $K$-means that improves directional fidelity while matching, and in some regimes improving, $\ell_2$ distortion. We then build Gain-Shape Residual Quantization (GSRQ) by incorporating a weighted extension of GSKM into an RQ pipeline. On LLaMA-3-8B, GSRQ substantially improves over strong KV cache quantization baselines across bit rates. At 1-bit, it improves the average accuracy across LongBench tasks from 11.34 to 33.54, a gain of 22.20 percentage points over VQLLM.
Large input spaces and complex inter-operation dependencies make black-box REST API testing challenging. AutoRestTest combines a Semantic Property Dependency Graph, multi-agent reinforcement learning, and large language models to intelligently explore large API input spaces. In the SBFT 2026 REST League, AutoRestTest ranked first in all three evaluation categories -- fault detection, overall efficiency, and overall effectiveness -- on 11 APIs (317 operations, approximately 29 per API), averaging 67.09 unique server errors and 17.27 successfully processed operations per API under a one-hour testing budget.
Computer-assisted synthesis planning breaks target molecules into accessible precursors using large libraries of reaction rules that assign each transformation a deterministic, interpretable label. But chemistry is long-tailed, making manual encoding intractable, and existing tools rely on fixed rulesets that cannot adapt to new chemistries. Here we present a fully automated pipeline in which a multi-agent framework of large language models (LLMs) classifies reactions and writes the rules themselves across 665,901 US patent reactions, generating each rule under a verification loop that tests it against the corpus. It expands a standard taxonomy from 68 to 14,073 classes without human curation. With a lightweight fingerprint classifier, it classifies 97.7\% of unseen reactions, matching a leading proprietary classifier while resolving chemistry more finely and extending on demand to chemistry outside its training distribution. The result is a living reactivity database and a general route to turning generative models into reliable, self-expanding symbolic systems.
We study a broad class of graphical models whose independencies correspond to vertex separation in mixed graphs with directed, undirected, and bidirected edges, that are capable of encoding independence structures arising from feedback, latent and selection mechanisms. In particular, we introduce separable graphs, in which each missing edge implies the existence of a separating set for its endpoints, and essentially separable graphs, those graphs separation equivalent to a separable graph. We show that these models include many existing graph families used to define graphical models an provide several characterizations of separable graphs and essentially separable graphs. We also provide multiple characterizations of separation equivalence for separable graphs. One is a graphical characterization in terms of ordinary graph properties, extending earlier results for specific subfamilies Another is a separational characterization depending only on graph separation properties. Finally, we provide a canonical representation for the equivalence classes of essentially separable graphs and develop an algorithm that, under suitable assumptions, identifies the equivalence class of any essentially separable graph.
Complexity and interpretability rarely coincide: systems rich enough for complex behaviours to emerge are usually too opaque to question, while transparent ones are too simple for anything complex to emerge. A single large language model (LLM) is a static artefact, hardly exhibiting any of the emergent properties we associate with life. This changes through interaction: populations of LLMs display emergent dynamics absent from isolated models. Furthermore, LLMs can be endowed with persistent memory, tools and shared skills, and the capacity to initiate actions unprompted, i.e., turning LLMs agentic. In this paper, we argue that such collectives of agents can serve as a computational substrate for Artificial Life (ALife) research. Critically, since the agents communicate in natural language, their collective behaviour can be directly interrogated by examining textual traces and asking the agents themselves. We outline the notion of interpretability in language-model research and extend it for collectives of agents. Lastly, we survey recent examples of agentic LLM collectives that already instantiate the idea of agentic substrates, from controlled experiments to deployments in the wild.
Memory-based discrete vision-language navigation (VLN) agents must act under partial observability, yet even strong frozen backbones remain vulnerable at test time. Two common failure modes are stale historical evidence at memory readout and inefficient local backtracking during action selection. We present DART-VLN, a training-free test-time control framework for discrete VLN. DART-VLN combines Test-Time Memory Decay, a read-side memory reweighting rule that suppresses stale and redundant evidence without rewriting stored content, with Anti-Loop Regularization, a lightweight next-hop penalty that discourages immediate reversals during action selection. The framework introduces no new learnable parameters and leaves the learned backbone unchanged. Experiments on R2R and REVERIE show a consistent pattern: decay-only provides stable read-side gains, while decay+anti-loop achieves the best overall quality-efficiency trade-off, yielding shorter trajectories, lower runtime, and improved navigation performance in key settings. Behavioral analysis further confirms that anti-loop regularization reduces local backtracking and improves path efficiency under frozen backbones. Overall, the results show that modest test-time control can make memory-based discrete VLN more reliable and efficient without retraining.
Program comprehension is a central research topic in software engineering, focusing on how developers understand a program's structure, behavior, and intent. Eye-tracking studies have traditionally relied on display-based measurements, where gaze positions are represented as screen coordinates. However, syntax-based analyses have recently emerged. Prior work proposed methods to convert eye movements into transitions between nodes in an abstract syntax tree, but the relationship between task correctness and eye-movement features for specific syntactic elements remains unclear. This study converts eye-tracking data into transitions between syntactic nodes and analyzes fixation proportions and gaze transition patterns. We investigate the relationship between these patterns and task correctness, comparing correct and incorrect groups. Our results reveal distinct differences in gaze transition patterns between the two groups. In particular, successful participants exhibit more systematic transitions across syntactic elements, suggesting the use of structured reading strategies.
Therapy-induced cardiotoxicity is the leading non-oncological cause of treatment interruption in breast cancer patients, yet early, automated risk stratification from routine cardiac imaging remains an unsolved problem. We present EchoRisk, the first curated, multicentre, longitudinal echocardiography dataset with explicit cardiotoxicity labels, released as the primary technical reference for the EchoRisk-MICCAI 2026 challenge. The dataset comprises 422 patients enrolled in the EU-funded CARDIOCARE prospective study across five European sites, yielding 2,159 echocardiography videos across 1,123 clinical exams acquired at up to five longitudinal timepoints, alongside a dedicated cohort of 280 patients with baseline imaging for early cardiotoxicity prediction. Three clinically grounded tasks are defined: automated estimation of left ventricular ejection fraction from cine video (Task 1), classification of LV dysfunction from longitudinal imaging (Task 2), and early prediction of therapy-induced cardiotoxicity from pre-therapy baseline echocardiography alone (Task 3). For each task we specify the evaluation protocol, primary and secondary metrics, and ranking procedure. We establish baseline performance using an R(2+1)D video backbone with LSTM aggregation trained from Kinetics-400 pretrained weights, demonstrating strong discriminative performance for cardiac functional assessment and LV dysfunction classification, while early cardiotoxicity prediction from a single pre-therapy video remains a significant open problem for the community. The dataset, evaluation code, and baseline implementations are publicly available to serve as a benchmark for further collaboration, comparison, and the creation of task-specific architectures in cardio-oncology.
Large language model (LLM)-based conversational agents (CAs) are now ubiquitous, creating new opportunities for AI-mediated behavior change. Their capacity to project nuanced personalities and adopt diverse metaphorical roles raises a design question: how should an agent's persona and personality be calibrated to the moment? Recent evidence suggests that (i) moderate personality expression outperforms low or high extremes on trust, enjoyment, and intention to adopt in goal-oriented tasks, and (ii) context-appropriate metaphors outperform static one-note assistants on user experience and uptake. Yet most CAs still fix both persona and style, risking misalignment when dynamics, urgency, and formality vary, for example in medical information seeking, fitness coaching, and reflective learning. We propose a Fluid Personality Framework that jointly adapts (1) the agent's metaphorical persona, such as coach, tutor, librarian, or tool, and (2) its personality expression intensity, low, medium, or high, as a function of task context, user goals and traits, and situational urgency. We sketch the framework and its core design dimensions.
Model organisms (MOs) - language models trained to exhibit undesired or unnatural behaviours - are frequently used as testbeds for evaluating white-box interpretability techniques. Current MOs are typically constructed via post-hoc supervised fine-tuning (SFT) on behavioural transcripts or synthetic documents. Prior research has shown that interpretability methods can easily identify hidden behaviours in these MOs. However, recent work suggests that such post-hoc training methods may make interpretability unrealistically easy. We investigate this claim by constructing a suite of 54 $\verb|OLMo2-1B|$- and $\verb|gemma-3-1b-it|$-based MOs trained with seven different techniques, including standard post-hoc SFT, post-hoc DPO, and more realistic integration of MO data into the OLMo post-training DPO phase. We use these MO variants to benchmark activation oracles, activation steering, logit lens, and sparse autoencoders. Our findings show that (i) MO interpretability depends strongly on training objective, target behaviour, model architecture, and training data generation pipeline; (ii) substantial variance remains even after controlling for differences in the strength of target behaviour expression; and (iii) our more realistic $\textit{integrated training}$ often yields less interpretable MOs than standard post-hoc methods. Our results cast substantial doubt on the validity of current MOs as interpretability proxies.
Radio-frequency (RF) sensing is a central modality for counter-unmanned-aerial-system (counter-UAS) defence because it exploits the control, telemetry, and video links between a drone and its operator. Reported accuracies for RF-based drone detection and identification are often very high, but many are obtained using cross-validation that splits a small number of continuous recordings into short segments. This can place near-duplicate slices of the same recording in both training and test partitions, creating data leakage. We study this leakage problem through theory and measurement. We formalise the optimism of segment-level cross-validation and show, using Cover's function-counting theorem, that a classifier can exactly memorise the recording-to-label map when the number of independent recordings, R, is small relative to the feature dimension, d. In particular, this can occur when 2R is less than or approximately equal to d. Under these conditions, naive accuracy approaches 1, and the inflation gap approaches 1 - ACC*, where ACC* is the Bayes accuracy. The inflation eases only once R grows beyond this separability threshold. A controlled synthetic experiment with 10 seeds confirms the predicted curves: naive balanced accuracy rises from the Bayes level toward 1.0 as recording-specific nuisance variation grows, while honest recording-grouped evaluation declines to chance, with a gap reaching about 0.5. On the public DroneRF dataset, pooled leave-one-recording-out cross-validation shows drone type identification, AR versus Bebop, collapsing from a naive macro-F1 of 0.74 to 0.46, the two-class chance level. A leakage-pathway ablation attributes essentially all of the inflation to segment-level leakage.
Financial markets evolve in response to real-world events reported in news, yet these drivers often remain implicit in text. To better explain market dynamics, event-market relations must be explicitly modeled through factual, company-centric, and environment-aware knowledge graphs. We present FinKG-News, a framework that automatically constructs such graphs by extracting news events as anchors linked to companies. Using FinKG-News as grounded evidence that integrates events, news, and company data, we develop an in-context learning architecture for credit risk report generation across three core financial dimensions. Automatic and human evaluations show that automated hallucination detection and quality assessment remain unreliable, making expert judgment indispensable. Our approach consistently outperforms baselines, improving quality by 19%-34% while reducing hallucinations. The source code and project resources are publicly available at: https://github.com/ichise-laboratory/FINKG-news.
Spatiotemporal point processes (STPPs) model event data in continuous time and space, with applications in mobility, epidemiology, and public safety. Recent neural STPPs span expressive intensity models, conditional density models, continuous-time latent dynamics, normalizing-flow spatial decoders, and score-based generative mechanisms. Yet comparison remains fragile because implementations differ in preprocessing, coordinate normalization, splits, likelihood conventions, and evaluation protocols. We present SEAHORSE, a unified framework for reproducible STPP experimentation. SEAHORSE formalizes neural STPPs through a common encode-evolve-decode interface and trains, tunes, and evaluates every model family under a single executable benchmark protocol with raw-coordinate likelihood reporting. This enables fair comparisons but, more importantly, controlled diagnostic studies. We pair SEAHORSE with HawkesNest, a synthetic stress-test suite, and show that increasing event-pattern complexity exposes each family's inductive bias, degrading some models sharply and leaving others stable. Code: https://github.com/YahyaAalaila/seahorse.
Large-scale crowd management requires pedestrian simulations that are both computationally efficient and compatible with feedback-based control. However, most open-source tools are either microscopic or not designed for network-scale closed-loop evaluation. This paper presents PedNStream (Pedestrian Network Flow Simulation), an open-source, Python-native simulator for macroscopic pedestrian network loading based on the Link Transmission Model (LTM). The framework extends LTM-based pedestrian models by incorporating stochastic link dynamics that capture diffusion and activity-induced variability, and replaces dynamic user equilibrium route choice with a utility-based formulation suited to uncertain, intervention-driven settings. PedNStream is implemented as a modular framework with built-in controller interfaces for interventions such as gating, flow separation, and route guidance. We evaluate the framework in a staged manner. Synthetic scenarios verify key mechanisms, including queue formation, spillback, congestion dissipation, and adaptive rerouting. Real-network experiments assess large-scale behavior and consistency with observed pedestrian counts. A closed-loop case study demonstrates controller integration, and a runtime analysis quantifies scalability. These results establish PedNStream as an efficient and practical testbed for large-scale pedestrian network simulation and control.
Reading order inference remains a critical bottleneck in the digitization of complex historical manuscripts, where pages contain multiple spatially interleaved reading streams, the canonical example being the Glossa Ordinaria layout, in which a central text is surrounded by commentaries that wrap around it in non-rectangular, non-convex regions. We present a training-free, graph-based framework: each OCR text line becomes a node in a directed candidate-transition graph, edges are scored by a weighted additive ensemble of two lightweight language-model signals (causal language model conditional likelihood and BERT next-sentence prediction, NSP; a third sentence-embedding signal was evaluated but did not improve reading order), and the global reading order is recovered as a degree-constrained directed path cover. To avoid the cascading "edge-theft" failures of greedy edge selection, we propose a max-regret inference rule that prioritizes commitments with high opportunity cost. We evaluate on synthetic Glossa Ordinaria grid layouts, on 23 ALTO page geometries (10 historical source pages plus mirrored and flipped variants), and on a 140-page multi-column English subset of OmniDocBench, comparing our method against the canonical recursive XY-cut (PaddleOCR PP-StructureV3) and two LayoutReader variants (layout-only and text+layout) on identical inputs. On wrap-around Glossa layouts our method recovers 95% of ground-truth successor edges on average vs. XY-cut's 50%; on the OmniDocBench multi-column subset it reaches 88% macro edge accuracy versus XY-cut's 75% and LayoutReader's 25%. The LayoutReader baselines transfer poorly due to a word-level vs. line-level granularity mismatch. We additionally verify mirror-invariance under horizontal and vertical page reflections: Our method changes by less than 1 percentage point, classical XY-cut by 2 points, and LayoutReader-T by up to 8 points.
Data assimilation models state dynamics conditioned on sequential observations, and has wide-ranging scientific applications. In the filtering setting, the goal is to model the posterior over the current state given all observations so far. Classical solutions typically make simplifying distributional or functional assumptions, e.g., linear-Gaussian systems, which can be inaccurate in many scenarios. In principle, particle filters (PFs) remove these assumptions, yet often collapse in high dimensions. Recent generative approaches learn conditional state transitions, but without principled Bayesian updates they do not recover the correct filtering posterior and can accumulate error over long horizons. In this work, we introduce Flow Proposal Particle Filters (FPPF), which learn a conditional generative model based proposal approximating the variance-minimizing optimal proposal for particle propagation. Conditioning on observations steers particles toward high-likelihood regions before weighting, reducing weight variance and delaying degeneracy. Since our proposal admits tractable likelihood evaluation, FPPF computes accurate importance weights and retains a Bayesian update step. We further extend FPPF to high-dimensional problems through localization strategies, adressing another standard PF failure mode. Extensive experiments on a variety of dynamical systems show that FPPF outperforms statistical baselines and other generative methods in non-linear, non-Gaussian, and high-dimensional regimes.
The success of deep learning models in classification and regression is widely attributed to the low-dimensional structure that real-world data tend to exhibit, despite their high-dimensional representation. This work attempts to provide a mathematical framework for binary classification on low-dimensional data, building on Cover's (1965) function-counting theory. With our framework, we aim to address the question of how the low-dimensional structure of the data affects the classification capabilities of learning models. Cover's theory relies on a general position assumption that blinds it to the underlying data structure. We refine this assumption to account for the low-dimensionality of the data and derive dichotomy counts that reflect the data structure. We further extend Cover's separation capacity and problem of generalization to the low-dimensional setting, enabling the impact of the underlying data structure on both to be analyzed.
Large Language Models (LLMs) represent one of the most significant advances in AI and natural language processing in recent years. Still, many pressing questions about their mechanisms, capabilities, and relationship to human cognition remain highly debated. This chapter aims to outline our current understanding of LLMs by discussing recent evidence on emerging capabilities and their mechanistic implementation within processing layers. We begin with a concise overview of the Transformer architecture, emphasizing how the attention mechanism enables training on massive datasets, allowing LLMs to function as generalist rather than specialized models. Next, we examine emergent LLM capabilities that appear to resemble aspects of human cognition, including symbolic reasoning, theory of mind, and deception strategies. Several studies provide evidence that LLMs can solve tasks previously thought to require human-like cognition. Other studies reveal insightful failure cases that shed light on the differences between human and LLM cognition. Alongside these findings, we review explainable AI approaches ranging from neuron activation analysis to circuit tracing. In the final section, we address current debates concerning what LLMs genuinely understand versus what they merely appear to understand. Prominent arguments against AI anthropomorphism point to the simplicity of LLM training objectives, claiming that LLM behavior is better explained by pattern memorization of training data than by genuine cognition. We argue that this standpoint is guided by misconceptions about optimization processes and cognitive capacity, and advocate for a more nuanced discussion of LLM cognition that neither dismisses the differences between humans and LLMs nor precludes the possibility of AI cognition through overly simplistic reductionist arguments.
In long-context use, large language models frequently synthesize answers from the meaning of a relevant context span rather than literally copy-pasting them. Identifying which attention heads perform this synthesis matters for interpreting long-context model behavior. Yet existing detectors miss these heads by construction: they reward heads whose attended token matches the generated token, a literal-copy criterion that captures where a head reads but not what it writes through its output-value (OV) circuit, the very mechanism that carries non-literal retrieval. We introduce Logit-Contribution Scoring (LOCOS), a write-aware detector that scores each head by the projection of its OV-circuit output onto the answer-token unembedding direction, contrasting needle and off-needle source positions in a single forward pass. Across three model families (Qwen3, Gemma-3, OLMo-3.1), mean-ablating the top LOCOS heads on the NoLiMa non-literal retrieval benchmark collapses ROUGE-L at lower head counts than prior attention-based detections; on Qwen3-8B, ablating 50 heads drives ROUGE-L from 0.401 to 0.000 while the strongest baseline still retains 0.292. The selected heads are retrieval-specific: parametric recall and arithmetic reasoning stay at baseline under the same ablation. On Qwen3-8B, the same ablation also drops MuSiQue from 0.55 to 0.08 and BABI-Long from 0.62 to 0.20, while a random-heads control stays within 0.05 of baseline.
Radiomics is the established approach for CT-based lung cancer phenotyping, yet comparisons with foundation models rarely isolate contributions of feature extractor, classification head, and segmentation choice, or test cross-cohort robustness. We benchmark five feature extractors (Curia, Curia-2, DINOv3, Radiomics2D, Radiomics3D), seven classification heads (TabPFN, TabICL, XGBoost, CatBoost, Random Forest, logistic regression, Ridge), and three segmentation regimes on five tasks: tumor volume and stage classification, 2-year survival prediction, histology classification, and age prediction. Models are trained on LUNG1 (n=338) and evaluated on an internal test set (n=84) and the external LUNG2 cohort (n=211), with worst-case cross-cohort performance as the primary metric. The dominant design factor is task-dependent: segmentation drives volume and stage classification, while classifier choice drives survival, histology, and age prediction. Radiomics is competitive for tumor volume, tumor stage and survival (partly due to label-derivation effects for the former); Curia variants reach comparable peak scores for survival; DINOv3 falls slightly short across tasks. Patch and slice aggregation have negligible impact. We recommend Curia with tumor segmentation and a CatBoost head as a safe default, achieving the best mean rank across the three primary clinical tasks, though task-specific selection consistently outperforms any cross-task default. When tumor delineations are unavailable, Curia-2 with lung segmentation and logistic regression offers a competitive alternative. All pipelines use a two-stage design suited to small cohort sizes where end-to-end fine-tuning would risk overfitting.
Recent research has increasingly focused on understanding how Transformers store and process knowledge, as well as how this knowledge can be edited. Research work in this area is often conducted in two phases: first, phenomena are explored on individual samples. Then, when results appear promising, more statistically robust experiments follow. To support the first phase, we propose KnowledgeDebugger, a GUI-based exploration tool for knowledge localization and editing in Transformers. Our tool - inspired by LM-Debugger - offers no-code access to the methods in EasyEdit, a widely used library of state-of-the-art Knowledge Editing approaches. We demonstrate the tool's effectiveness through case studies of recent findings in this field.
Most existing multitask learning approaches are limited by their reliance on task-specific loss functions tailored to the scale and type of each outcome. When outcomes differ across tasks, these losses are generally not directly comparable, which makes it difficult to formulate a unified objective and may limit information sharing across tasks. We propose a multitask transformation framework in which task-specific responses may differ through unknown monotone transformations. Motivated by high-dimensional biological applications in which the predictor dimension may diverge with the sample size while only a common subset of predictors is informative, we consider shared sparsity across tasks. Under this framework, we estimate the target functions and identify important predictors by optimizing a smoothed rank-based criterion with a group-Lasso penalty, implemented through a multitask deep neural network with a shared first layer. We establish the nonasymptotic excess-risk bounds, and variable-selection consistency for the proposed estimator. Simulation studies show that the proposed method achieves competitive prediction and variable-selection performance compared with competing approaches. Analyses of gene-expression studies with continuous, binary, and mixed outcomes further illustrate that the proposed method improves prediction and identifies biologically meaningful shared predictors.
Large language model (LLM)-based software engineering agents are increasingly developed to resolve software issues by generating patches from issue reports and code repositories. Bug reproduction tests (BRTs) are an important building block for such agents and have been shown useful for patch validation. However, it remains unclear whether BRTs can also help the more central stage of patch generation. We first conduct a preliminary study and find that directly using advanced BRT generators to guide patch generation is not beneficial: fail-to-fail BRTs can mislead agents, while even fail-to-pass BRTs bring limited or negative gains. Our analysis reveals two reasons: fail-to-pass BRTs may cover only one manifestation of the reported issue, leading to partial patches, whereas fail-to-fail BRTs are unreliable as direct patch-generation targets. Motivated by these insights, we propose SWE-Doctor, a software issue resolution agent that guides patch generation with runtime diagnoses derived from multi-faceted BRT executions. SWE-Doctor first generates multi-faceted BRTs for different behavioral requirements stated in the issue, then executes and debugs these BRTs to construct runtime-grounded diagnosis records, and finally uses the diagnoses together with localization information inferred during BRT generation to guide patch generation and reduce partial patches. We evaluate SWE-Doctor on Python bug-fixing issues from the widely adopted SWE-bench Verified and SWE-bench Pro across five LLM backends. SWE-Doctor consistently outperforms existing agents across all 10 LLM-benchmark combinations, achieving average resolution rates of 75.7% on SWE-bench Verified and 59.4% on SWE-bench Pro. In particular, on the more challenging SWE-bench Pro, SWE-Doctor improves the average resolution rate by 8.0-8.9 percentage points over the baseline agents.
Semantic trajectory analysis has recently emerged as an approach for modeling human movement by capturing implicit patterns and behaviors through semantic information (e.g., visitors' profiles and goals) beyond raw spatial paths to better understand why people move in certain ways. However, analyzing semantic trajectories in real-world scenarios remains challenging, as collecting high-quality data is costly and often lacks rich semantic information. Meanwhile, existing simulation tools require substantial technical expertise, which makes them difficult for practitioners to adopt. To address these limitations, the paper proposes ${SenseWalk}$, an interactive system that supports simulating semantic trajectories by LLM-powered agents. We develop a simulation workflow that combines LLMs and the social force model to balance physical plausibility and semantic coherence. A user-friendly interface is designed to facilitate users in customizing the simulation configuration and analyzing simulation outputs. We also conduct a quantitative experiment to evaluate the effectiveness of our simulation workflow, and a user study (n=12) to assess the usefulness and efficiency of our system.
Automatically detecting stress in speech provides an unobtrusive way to gain insights relevant to behavioral research or clinical assessment. This study investigates the automatic differentiation between a stressful and non-stressful situation, and the prediction of physiological and affective stress responses. Speech data was collected from 50 participants who either completed the Trier Social Stress Test (TSST) or a non-stressful control condition. With a processing pipeline that included speaker diarization and machine learning models, we achieved stress detection performance significantly above a mean baseline. Moreover, relevant physiological and affective stress responses were partially predictable from acoustic-prosodic features. Feature-importance analyses identified the most informative predictors contributing to model performance. The findings demonstrate that speech can serve as a meaningful and unobtrusive indicator of multiple dimensions of the human stress response.
Chest X-ray multi-label classification is a core task in intelligent medical imaging diagnosis. However, real clinical data often exhibit extreme long-tailed distributions, leading to degraded performance on rare diseases in tail classes. This issue is not only driven by data scarcity but also by two intrinsic factors:1) attenuation of tail-class lesion representations under complex anatomical backgrounds, and 2) dominance of head classes in modeling label co-occurrence relationships. To address these challenges, we propose TRCGL-Net. First, a learnable text-guided conditional diffusion model is employed to generate high-quality tail-class chest X-ray image samples under disease semantic constraints, improving data diversity and realism of rare disease patterns while alleviating class imbalance and preserving pathology-consistent semantics.Second, a channel reweighting mechanism is introduced to perform feature recalibration by emphasizing disease-relevant feature channels, thereby improving feature discriminability under long-tailed distributions.A class-aware attention mechanism is further applied to generate class-specific attention maps, enabling the model to localize disease-relevant regions and focus on fine-grained lesion areas.Finally, a graph convolution network based on label co occurrence is introduced to establish an information propagation mechanism among categories. Experiments on the PadChest dataset show that the proposed method achieves a tail-class mAP of 0.4904, an overall mAP of 0.4408, and an mAUC of 0.8989, outperforming state-of-the-art methods. TRCGL-Net effectively improves recognition performance for rare diseases under long-tailed distributions and mitigates the impact of extreme class imbalance in chest X-ray multi-label classification.
Trustworthy deployment of Agentic Retrieval-Augmented Generation (RAG) systems requires mechanisms for estimating when multi-stage reasoning pipelines may fail. This paper presents an uncertainty-aware Agentic Retrieval-Augmented Generation (RAG) framework in which planner, evaluator and generator stages produce uncertainty signals derived from semantic divergence and generator self-evaluation. These signals are propagated through a Bayesian Network (BN) to estimate system-level uncertainty and provide node-level indicators of potential failure points across the workflow. The approach is evaluated on StrategyQA and HotpotQA using GPT-3.5-Turbo and GPT-4.1-Nano, with Area Under the Receiver Operating Characteristic Curve (AUROC), Area Under the Accuracy-Rejection Curve (AUARC), Expected Calibration Error (ECE), and Brier Score used to assess discrimination, selective prediction and calibration. Results show that Bayesian propagation is more effective on HotpotQA, where uncertainty accumulates across multi-hop reasoning stages, while StrategyQA exposes limitations caused by miscalibration and unreliable upstream signals. The study positions Bayesian uncertainty propagation as a promising but preliminary mechanism for monitoring Agentic RAG systems, with future validation required in industrial domains such as Offshore Wind (OSW) maintenance decision support.
This paper introduces Svarna, a free, open-source, web-based corpus workbench for modern Greek. Svarna integrates five databases covering various registers, institutional, literary, dialectal, social media, and historical, to provide a total of more than 507 million words and around 29 million sentences. This platform addresses the chronic gaps in Greek language technology. Although various corpus resources exist, they are scattered across different platforms, and in many cases, institutional access is restricted or they are no longer available online. Svarna integrates these resources into a single interface that can be used without logging in, installation, or specialized training. This system provides a concordancer with KWIC marking capabilities, frequency analysis including register-by-register normalization, collocation extraction using mutual information, a dictionary of 93 Greek discourse markers providing distribution profiles, text-level analysis tools including n-grams, variants, and collocation networks, register comparison using log-ratio, regular expression search, and an optional LLM layer for pragmatic annotation and free research mode. This platform is built upon SQLite FTS5 full-text indexes provided via a FastAPI backend, deployed as Docker containers on Azure, and released under the MIT license. Source code, build scripts, and deployment configurations are publicly available on GitHub. Users can add their own corpora and deploy their own instances. This document describes the system design, corpus structure, and use cases demonstrating the various queries supported by the platform. Svarna serves as the first step in exploring available data and is expected to lay the foundation for more comprehensive research in the future.
Visual analytics (VA) plays an increasingly important role in supporting machine learning (ML) workflows. In the field of visualization, such approaches and techniques are referred to as VIS4ML. While ML models are mostly learned automatically, the corresponding ML workflows receive a variety of human inputs, such as data labelling, feature engineering, model architecture designing, hyper-parameter tuning, and so on. In this work, we surveyed over 200 VIS4ML papers to gain an understanding of how humans inject their knowledge into ML workflows through interactive visualization. We collected a corpus of VIS4ML papers from the IEEE VIS conferences in the past decade. We developed a coding scheme to facilitate the literature research from four perspectives: characteristics of ML, visualization, interaction, and actions. The analysis of the coded dataset allows us to observe different pathways that transfer human knowledge to ML workflows via interactive visualization. Building on the analysis, we explain the phenomena of VIS4ML using the conceptual model that views VA as model building and the information-theoretic cost-benefit analysis that reasons VA as for optimizing ML workflows. This work provides unequivocal evidence showing the merits of using VA in ML workflows. The full list of surveyed papers, along with all analysis results and figures, is available at https://vis4ml4hd.github.io/ml-knowledge-inject-va/.
Emotion recognition in natural language is a foundational challenge in affective computing, with critical implications for human-computer interaction, mental health support, and conversational AI. This paper presents a rigorous, unified zero-shot evaluation of three leading commercial large language models: Claude (claude-sonnet-4-6), ChatGPT (GPT-5.4), and Gemini (gemini-2.5-flash). The models were queried through their respective production APIs as of April 2026 on a fine-grained 13-class emotion classification task. Using a stratified 1,000-sentence sample from the boltuix/emotions dataset, which comprises 131,306 sentences across 13 categories, a single uniform prompt with no exemplars was applied identically across all models. Gemini achieves the highest accuracy (39.9%) and macro-F1 score (0.363), followed by GPT-5.4 (38.8%, macro-F1 = 0.291) and Claude (38.0%, macro-F1 = 0.159). All models excel on sarcasm and desire while consistently failing on love, confusion, and shame. McNemar tests reveal no statistically significant pairwise differences (p > 0.10), suggesting convergence at a shared zero-shot ceiling. Claude's markedly lower macro-F1 score exposes a class-imbalance prediction bias. These findings highlight the current limitations of frontier AI systems in zero-shot fine-grained emotion classification.
Realizing quantum neural networks (QNNs) in industry requires knowing which quantum computing paradigm suits which task. Motivated by AI accelerators and high-bandwidth memory, where die stacking makes wafer-level defect screening central to yield, we study WM-811K wafer-map defect classification (eight classes), comparing the dominant paradigms, continuous-variable (CV) and discrete-variable (DV), under controlled conditions. To isolate the quantum circuit as the sole variable, a shared convolutional backbone (~4.3M parameters) feeds interchangeable heads (classical dense, CV-QNN, or DV-QNN) as the only structural difference; each quantum head is scaled over three sizes (3, 4, 8 qumodes/qubits). The CV head consistently outperforms the DV head: at four qumodes/qubits it reaches 79.7 +/- 1.8% accuracy versus 61.6 +/- 1.4%, a non-overlapping 18-point gap. The advantage is sharpest on the spatially localized Edge-Loc class, easily confused with Scratch, which CV recovers with recall 0.66 +/- 0.06 while DV fails at every size (<=0.05), showing the structured CV layer better captures fine spatial distinctions between defect types. Training curves show the DV limitation is a representational-capacity ceiling, not an optimization failure; at the Fock cutoff used here (d = 2) the CV advantage reflects two intrinsic properties, a structured, neural-network-analogue layer and continuous phase-space encoding, not Hilbert-space dimensionality. On IBM hardware, DV accuracy holds at shallow depth, degrading only at the deepest circuit. Both quantum heads remain below the classical baseline (85.0%), but the controlled setting isolates where a structured head already helps and, as noise and scale improve, which paradigm can deliver practical advantage.
Time series are central to modern data mining applications, from industrial telemetry and server metrics to finance and physiology, yet time-series self-supervised learning often depends on view and augmentation choices that encode domain-specific invariances. We study how an SSL recipe behaves when its method-specific configuration is reused unchanged after the pretraining signal family changes, framing this as a fixed-recipe stress test rather than a comparison against optimally tuned methods. We introduce Latent Euclidean Next-Embedding Prediction Architecture (LeNEPA), a no-augmentation next-latent-token objective with a causal backbone. LeNEPA replaces the stop-gradient/EMA stabilization used by vanilla NEPA with SIGReg-based isotropy regularization and computes the predictive loss in a lightweight projected space that is discarded for evaluation. We compare LeNEPA with an ECG-tuned JEPA recipe under a fixed-horizon frozen-probe protocol on PTB-XL and Diag, a synthetic diagnostic corpus generated with Aionoscope. Both methods are retrained independently on each dataset while keeping their method-specific recipes unchanged. In this protocol, the ECG-tuned JEPA recipe is strong in-domain on PTB-XL but weaker when reused unchanged on Diag, whereas LeNEPA preserves useful frozen-probe gains on both datasets. Learning curves suggest faster early representation acquisition: LeNEPA reaches 80% of its final AUROC/AUPRC gain after 2--5k updates, compared with 5--10k updates for the faster JEPA readout. As a separate external frozen-encoder check, a CauKer-pretrained LeNEPA variant reaches 77.65% mean UCR-128 Random-Forest accuracy in a single-seed, best-checkpoint run, within 1.16 points of Mantis and within 0.24 points of MOMENT (77.89%). Overall, the results support no-augmentation latent prediction as a useful candidate recipe for low-retuning time-series SSL.
Time-series models are often evaluated by what they can forecast or classify, but those scores do not show whether their representations preserve the process state a user may want to inspect: event timing, phase, amplitude, frequency, or regime variables. We introduce Aionoscope, a generator-based diagnostic tool for debugging latent-state accessibility in frozen time-series representations. Aionoscope separates process generation from observation rendering, producing seeded synthetic streams with exact categorical and dense labels across mixture complexity and nuisance variation. We instantiate Aionoscope as Primitive Process Mixtures and evaluate 37 model-plus-adapter systems with a common pooled linear-probe protocol. The main result is a mismatch between coarse and fine-grained accessibility. Most systems make component presence easy to recover, but expose dense process state much less reliably: the highest observed dense-probe row reaches 0.689 mean masked $R^2$, while a dense-feature oracle reaches 0.999. This is the failure mode Aionoscope is designed to surface: a representation can look informative at the level of "what kind of signal is present" while hiding the timing, phase, amplitude, frequency, or regime variables needed for debugging.
Implicit neural representations (INRs) are well suited to cardiac motion estimation, providing continuous, compact representations of motion fields. However, fitting an INR to each image sequence is time-consuming and sensitive to the optimisation trajectory. Learned priors can help guide optimisation towards plausible motion fields and enable faster adaptation, but learning priors for cardiac motion INRs remains under-explored. In this work, we compare four strategies for learning cardiac motion priors, including a population prior learned by joint optimisation, a consensus prior obtained by weight averaging, auto-decoders, and meta-learning. Using short-axis tagged cardiac magnetic resonance images from the UK Biobank, we evaluate their impact on tracking accuracy, motion behaviour, and adaptation trajectory. All learned priors substantially improved early adaptation performance compared with random initialisation. While the simple consensus prior was effective, auto-decoders recovered large deformations faster during early adaptation. Meta-learning achieved strong early performance and maintained the best adaptation trajectory over 50 iterations.
Generative models learn data distributions that reside on a low-dimensional manifold within a higher-dimensional ambient space. Optimizing differentiable objectives on this manifold is challenging: the ambient loss landscape is high-dimensional, rugged, and non-convex. Direct gradient descent, blind to the manifold's geometry, quickly drifts off it. Diffeomorphic optimization starts from the observation that diffusion and flow models provide a map from the data manifold to a much simpler base space in which we perform gradient descent. Using differential geometry, we show this is equivalent to Riemannian gradient descent on the data manifold up to $\mathcal{O}(λ^2)$ corrections, keeping trajectories on-manifold by construction and yielding a smoother optimization surface. For protein design, we extend diffeomorphic optimization to the matrix Lie groups $\mathrm{SO}(3)$ and $\mathrm{SE}(3)$, deriving an autograd-compatible $\mathrm{SO}(3)$ gradient and a generalized adjoint-state method for backpropagation through Lie-group ODE solvers. Diffeomorphic optimization improves over tuned guidance on secondary-structure targeting with FrameFlow ($91.3\%$ vs. $63.3\%$ of residues in the Ramachandran target), outperforms OC-Flow on peptide binding affinity at $2\times$ the speed, and reduces Rosetta energies by thousands of units across the PDB test set for structures with hundreds of residues.
While prior work has explored emotion control in hybrid text-to-speech systems, the geometric properties of these modules, and their implications for steerability, remain poorly understood. We present the first comparative study of speech language model (SLM) and conditional flow-matching (CFM) modules as activation steering sites for mixed emotion speech synthesis. We first characterize emotion representations using linear probing and local intrinsic dimensionality (LID), and then evaluate single-site and joint steering for mixed-emotion synthesis. Our results show that SLM offers a clean, low-dimensional emotion-specific subspace with strong speaker--emotion disentanglement, while CFM exhibitspoor cross-speaker generalization due to speaker--emotion entanglement. Joint steering increases emotion intensity but degrades proportional control and speech quality on in-distribution data. These findings provide practical guidance for multi-site activation steering in hybrid TTS systems and highlight the importance of representation geometry in controllable speech generation.
Quantum computing is increasingly explored for software engineering (SE) optimization, but translating natural-language (NL) task-level requirements into executable quantum applications still demands substantial quantum and programming expertise. We present QPipe, a large language model (LLM)-based multi-agent architecture that autonomously turns NL requirements into traceable quantum-application workflows through specialized agents for requirement parsing, formulation, code generation, review, execution, and verification. We evaluate QPipe on 20 NL requirements, each associated with a real-world benchmark and a test-optimization problem. QPipe successfully completes the key stages of quantum-application generation across requirements, achieving average rates of 100% for code compilation and 96.7% for application execution and final-result combination, with average generation costs of 260.1 seconds and 1.89M tokens per requirement. Among the generated quantum applications that execute successfully, the returned solutions outperform the offline genetic algorithm baseline in most cases. Ablation results further show that QPipe's advantage depends on retaining code-generation skills, task knowledge, review feedback, and multi-agent decomposition. These results indicate that agentic coordination can support generation of executable quantum applications for tackling test optimization problems from real-world benchmarks.
Persona-driven generations (PDGs) have seen prolific use in research and industry applications, where a large language model (LLM) takes on a 'persona' while completing some task. While persona expressed through free-form text (like dialogue) has substantial work investigating stability or consistency, relatively, persona expressed in non-text-heavy outputs (like in multiple-choice question answering, or MCQA) is often overlooked. We work to address this gap, seeking to understand the instability of LLM PDGs in MCQA tasks. We develop three metrics investigating the performance, outcome, and question correctness stability, evaluating three distinct dimensions. Using these metrics, we find that instability varies consistently between model families and model size, and across question domains, with math/commonsense questions leading to greater instability. We also find task prompt format introduces more prediction instability than other hyperparameters, like temperature. Finally, we find that instability is related to task accuracy, and using our instability metrics, find different experimental settings that result in different best and worst personas for tasks, despite their similarity. This reveals the importance of checking hyperparameter instability in PDGs.
Predicting cancer drug response from transcriptomic profiles is a cornerstone of precision oncology, yet the scientific value of machine learning models hinges not solely on predictive accuracy, but also on their capacity to generate reliable biological insights. Current explainability approaches in this setting are computationally costly, lack robustness, and reduce complex drug response to univariate gene importance scores, overlooking the coordinated gene activity that drives sensitivity and resistance. In this work, we present ILLUME+, a scalable post-hoc explainability framework that moves beyond single-gene assessments to capture multiple, complementary forms of explanation. Integrated into our end-to-end pipeline, ILLUME+ produces more stable gene importance scores than existing baselines, recovers established drug-gene associations and mechanisms of action, and enables AI-assisted hypothesis generation to uncover novel interaction-driven molecular signals in cancer biology.
Delta debugging provides an automatic way to minimize a program input while preserving a certain property. However, its effectiveness fundamentally relies on the availability of test oracles to determine whether a reduced input still preserves the specific property. Consequently, the oracle problem substantially limits the applicability of existing delta debugging techniques, particularly for oracle-deficient programs where output correctness cannot be directly determined. To address this problem, this paper proposes a novel approach, DDMT, to enhance the applicability of delta debugging, especially facilitating its application to oracle-deficient programs. Our key insight is to redesign an oracle-independent test function and incorporate it into the reduction procedure of delta debugging such that the property-preservation validation can be accomplished without requiring a test oracle. To this end, DDMT employs the technique of metamorphic testing, which is a property-based and oracle-independent testing method. It establishes a metamorphic testing-based test function, using it as a replacement for the original test function adopted by delta debugging. The experiments evaluate DDMT on 66 subjects across both oracle-available and oracle-deficient scenarios, with different delta debugging approaches. The results positively confirm that DDMT can enhance the applicability of delta debugging while often preserving or improving reduction effectiveness and query efficiency. Furthermore, compared to the relevant delta debugging approaches, DDMT is also able to achieve performance improvements with proper configurations.
Diffusion Transformers (DiTs) have demonstrated impressive performance in image generation but suffer from substantial computational overhead and resource consumption. Post-training pruning offers a promising solution; however, due to DiTs' unique architectural design and parameter distribution, traditional pruning methods are inapplicable, leading to significant performance degradation. Specifically, prior methods developed for LLMs, which derive metrics through a series of approximations, amplify the relative contribution of weights in the saliency metric. In addition, weights in DiTs exhibit significantly larger magnitudes than those in LLMs. Moreover, existing pruning granularity overlooks variations in model structures. In this paper, we propose DiT-Pruning, which improves pruning performance by introducing customized saliency criteria and pruning granularity. We design a novel metric that balances the contributions of weights and activations from an energy-based perspective, enabling more effective identification of important elements. Furthermore, we observe distinct clustering patterns in the two-dimensional weight space. Accordingly, we adopt a clustering-aware pruning granularity, enabling effective sparse allocation. Extensive evaluations on various DiTs show that our method consistently preserves image quality, especially under high sparsity. For FLUX.1-dev at 512x512 resolution on MJHQ, DiT-Pruning achieves only a 0.001 loss in CLIP score at 50% sparsity, dramatically outperforming recent pruning methods.
Generalizing machine learning models to environments that differ from their training distribution remains a critical hurdle, particularly when data from the target domain is entirely or partially unavailable. We propose Generative Meta-Learning with Human Feedback (GMHF), a novel framework that bridges this domain gap by leveraging expert intuition to guide data synthesis. Grounded in a theoretical analysis of generalization error, we derive bounds demonstrating that aligning the distribution of generated data with human beliefs regarding the target physics significantly mitigates risk. GMHF operationalizes this insight by employing a Conditional Neural ODE (cNODE) as a generative digital twin, coupled with a Reinforcement Learning (RL) agent. The agent iteratively refines the latent physical parameters of the generated trajectories based on feedback, effectively steering the meta-learner toward the unobserved target distribution. Empirical validation on a nonlinear Duffing oscillator shows that GMHF substantially reduces deployment loss as expert reliability increases, and that the divergence between generated and target data falls under reliable feedback, directly corroborating the divergence-minimisation mechanism predicted by our theory. Further experiments on a non-dynamical probabilistic model confirm that the framework extends beyond ODE-governed systems, establishing human-AI collaboration as a rigorous catalyst for robust generalisation under distribution shift.
Accelerating materials discovery requires AI systems that can generate scientifically valid hypotheses through multi-step, domain-grounded reasoning. Standard large language models often produce fluent but weakly traceable responses to open-ended materials design problems, making it difficult to determine whether final answers are supported by coherent intermediate reasoning. We develop Graph-PRefLexOR, a family of graph-native reasoning models fine-tuned with Group Relative Policy Optimization (GRPO) to organize reasoning into explicit phases for mechanism exploration, graph construction, pattern extraction, and hypothesis synthesis. This design links neural language generation with symbolic relational structure, enabling causal connections to be constructed, inspected, and reused. On 100 open-ended questions from materials science and mechanics literature, Graph-PRefLexOR achieves 40-65% improvements over corresponding base models, with the largest gains in reasoning traceability. Embedding analyses show broader semantic exploration and approximately 2-3 times greater semantic diversity than baselines. Semantic backtracking and layer-wise hidden-state analyses further show stronger alignment between structured reasoning and final answers. Finally, test-time graph expansion reveals that additional compute primarily increases long-range conceptual recombination within a bounded semantic space, rather than simply expanding semantic coverage. These results establish graph-native reinforcement learning as a pathway toward interpretable AI systems for scientific hypothesis generation in materials design and other scientific applications.
Although large language models (LLMs) have demonstrated impressive creative fiction generation, they struggle to maintain narrative consistency and coherent plot lines in long-form stories. In this work, we introduce a unified framework for long-form narrative generation and verification. MAGNET, a multi-agent goal-driven narrative engine for storytelling, generates stories with persona-grounded character agents that propose actions based on a shared world state and evolving story goals, while ATLAS is a graph-based pipeline that compares scene-level world representations across a generated story to detect hallucinations. By evaluating MAGNET using an LLM editor, pairwise rubric scoring, and ATLAS, we show that our framework produces coherent narratives compared to single-model prompting and IBSEN. At 100 pages, MAGNET reduced annotations and hallucinations by 41 and 50%, respectively, compared to the single model baseline and by 34 and 45%, respectively, compared to IBSEN, with pairwise rubric evaluation showing similar results. These results suggest that long-form narratives can emerge from explicit world-state tracking and goal-driven multi-agent generation, providing a foundation for controllable and structurally coherent long-form narrative generation.
World models can enable Model Predictive Control (MPC), but this requires dynamics prediction that is both fast enough for online use and expressive enough to represent uncertain futures. Diffusion models offer a natural mechanism for modeling uncertain dynamics, yet their iterative inference procedure makes them difficult to use for low-latency latent planning. We bridge this gap with Value Diffusion World Models (Valdi), combining end-to-end online training for MPC with a latent diffusion dynamics model. In preliminary experiments on the CarRacing environment, we show that Valdi, using a single diffusion step at both training and inference, matches a deterministic MLP baseline. Our experiments expose a trade-off between predictive multimodality and control performance in this setup. Code is available at https://github.com/Kit115/ValueDiffusionWorldModels.
As exponential compute scaling continues, will the capabilities of frontier AI models outstrip what is accessible to developers on a small fixed budget? Or will capabilities converge, with "meek models inheriting the earth"? Building on Gundlach et al. (2025b), we show that the answer depends on how we value and measure AI capabilities. We discuss conventional performance measures and show that, while validation loss shows a shrinking gap, on other metrics frontier models grow their lead forever. Classifying performance metrics by their functional forms in relation to training (and inference) compute, we provide tight mathematical conditions for determining which metrics favor meek models, and show that bounded performance metrics always do. But careful interpretation of performance metrics is essential: we show that many common bounded metrics have closely-related counterpart metrics that are unbounded (and vice versa). Determining the apt metric in a domain is a prerequisite for policy, since bounded and unbounded metrics may suggest opposing policy responses. If a particular capability -- like software engineering, synthetic biology, or rhetorical persuasiveness -- is unbounded when measured in the terms we care about, frontier-level capability will likely be concentrated in the hands of a few wealthy actors. Conversely, if that capability is instead bounded, frontier-level capabilities proliferate through meek models into the hands of the many.
AI coding agents increasingly rely on skills: structured context bundles, typically a SKILL.md file with a YAML header and Markdown body, loaded on demand for domain knowledge, workflows, and scripts. Public registries such as skills.sh now host tens of thousands of skills, making them an emerging unit of reuse in agent-based software engineering. Yet skills have largely been viewed as agent capabilities rather than software artefacts whose content and evolution shape agent behaviour. We present the first empirical study of AI agent skills as engineered artefacts that are authored, reused, customised and maintained, across public registries and personal-use repositories. We mined 18,463 skills from skills.sh and 23,199 personal-use skills from 5,876 GitHub repositories, identifying 3,709 reuse links. LLM-based classification into SWEBOK knowledge areas (KAs) shows Software Construction dominates alongside a long tail of specialised areas. A thematic analysis of 180 skills identifies six content categories. Qualitative coding of 444 modifications reveals six themes, of which reworking operational specifications and adapting knowledge and resources are the primary target of change. Our findings show that reuse is largely a one-time copy operation: most reused skills remain near-verbatim, 53\% are never modified after adoption, and subsequent local maintenance is overwhelmingly additive. Customisation primarily adapts skills to local environments, whereas evolution accretes new inline domain knowledge. Across both, a stable behavioural contract -- how a skill interacts with users, monitors runtime state, and recovers from failures -- remains almost untouched. These results suggest maintenance effort should focus on project-specific bindings, and that registries and tool support should enable consolidating the domain knowledge skills re-author in isolation.
Mixed-precision quantization (MPQ) has become a key technique for deploying large language models under stringent memory and compute constraints. We first identify a phenomenon that we term the Perplexity Illusion: layers ranked as important by perplexity-based sensitivity show little rank correlation with those that are most influential for complex reasoning performance, with Kendall $τ\approx 0$ in our analysis. We further reveal an Alignment-Diversity Tradeoff: using only target-task calibration data can degrade post-quantization performance, whereas incorporating general-domain data stabilizes sensitivity estimation and improves robustness across tasks. Based on these observations, we propose TASA (Task-Aware Sensitivity Analysis), a two-level framework that jointly optimizes calibration-data composition and mixed-precision bit allocation. Specifically, TASA searches for a calibration-data mixture using a training-free gradient-trace alignment criterion, and then aggregates perplexity and reasoning-oriented sensitivity signals to guide both inter-layer and intra-layer bit allocation. Experiments on LLaMA-3-8B and Qwen2.5-7B reveal a precision inversion: appropriately allocated 3.5-bit models can match or surpass less task-aware 4-bit baselines. At an average precision of 3.5 bits, TASA matches or outperforms several competitive 4-bit uniform baselines in aggregate accuracy, and improves over the strongest W3 baseline on GSM8K by more than 20 absolute points on LLaMA-3-8B. These results show that calibration-data composition substantially affects task-sensitive quantization, a factor underexplored in prior work.
Hallucination detection for retrieval-augmented generation (RAG) is usually evaluated on natural-language document evidence. However, grounded generation systems increasingly rely on structured inputs: source code, developer-tool output, markdown documents, tables, and repository metadata. We introduce a unified benchmark for span-level hallucination detection over code, tool output, structured documents, and existing natural-language RAG datasets. The benchmark is built by starting from grounded correct answers, injecting localized hallucinations with exact character labels, and validating the code test split with evidence-based review. Our fine-tuned Qwen3.5-2B detector reaches 0.689 span-F1 on the unified test set and 0.60 on the code-agent source, where it substantially outperforms LettuceDetect-large (0.17) and the strongest zero-shot LLM judges we evaluated (at most 0.22). The same model remains competitive on established natural-language benchmarks, with 81.8 RAGTruth example-F1 and 0.724 English PsiloQA IoU.
Open web-scale pre-training corpora remain concentrated in English, limiting multilingual LLM development. We introduce MultiSynt/MT, an open synthetic parallel corpus with approximately 4.8 trillion target-language tokens across 36 European languages, produced by translating 100 billion high-quality Nemotron-CC tokens with Tower+ and OPUS-MT/HPLT-MT systems. For many medium- and lower-resource European languages, this is the largest openly available pre-training resource. On a broad multilingual benchmark suite, reference LLMs trained on MultiSynt/MT reach the final score of HPLT 2.0, a native-data baseline, using roughly 72% fewer pre-training tokens, and outperform it by approximately 15% relative at a matched 100B-token training budget. Our analyses also identify evaluation blind spots: standard multiple-choice benchmarks miss translation-quality differences that a fluency-sensitive LLM-as-judge evaluation cleanly recovers on the trained LLMs (with no fluency deficit in MultiSynt itself), and Norwegian idiomatic and culturally grounded tasks remain better served by native data. We release the corpus, including row-aligned translations from multiple systems, to support controlled research on multilingual pre-training data and evaluation.
We present DeWorldSG, a novel framework that generates spatio-temporally robust 3D Semantic Scene Graphs from RGB-D sequences. Existing methods often struggle to construct reliable 3D scene graphs due to unstable 3D object representations and missing relations caused by frame-wise inference. DeWorldSG addresses these issues by estimating instance-level geometric 3D Gaussian distributions through depth-guided filtering and representing each object as a probabilistic 3D node rather than a single projected point. To mitigate relational sparsity from frame-wise inference, our framework further aggregates spatiotemporal evidence across object pairs and refines relations using contextual priors derived from a world model (V-JEPA 2). Experiments on the 3DSSG and ReplicaSSG datasets demonstrate state-of-the-art (SoTA) performance in both object and predicate prediction, while producing temporally consistent scene structures. In particular, our method improves triplet recall by 77.4% and predicate recall by 23.2% over prior SoTA approaches, making it suitable for robotic manipulation and AR applications. Our code and models are open-sourced.
Recent advances in neural rendering have established 3D Gaussian Splatting (3DGS) as a highly efficient representation for novel view synthesis, enabling fast training and real-time rendering with strong fidelity. However, when supervision is limited to sparse input views, 3DGS tends to overfit to the observed images and generalize poorly to unseen viewpoints. We address this challenge from the perspective of flat minima (FM) optimization, which seeks solutions that remain stable under small parameter perturbations. Viewing Gaussian parameters as trainable weights, we adapt FM principles to the geometric and dynamic nature of 3DGS with a lightweight training framework. Our method regularizes optimization with controlled Gaussian perturbations that account for each Gaussian's anisotropy and the training progress, preserving fine details while improving robustness to sparse-view overfitting. To further stabilize this flat minima optimization process, we introduce periodic reinitialization, which temporarily returns non-positional parameters to their initial states for a short window. Together, these techniques integrate seamlessly into existing 3DGS pipelines without architectural changes. Experiments on LLFF and Mip-NeRF360 datasets demonstrate improved quantitative metrics and perceptual quality under sparse-view supervision, producing reconstructions that are sharper, more stable, and better generalized to novel viewpoints.
Private continual counting is a fundamental problem in differential privacy: given a binary stream of length $n$, where each $1$ corresponds to the contribution of one individual, the goal is to release all running counts while protecting the privacy of each individual. The standard algorithm is the binary tree mechanism, whose Gaussian-noise variant achieves expected $\ell_\infty$ error proportional to $\log^{3/2} n$ for approximate differential privacy. Whether this dependence on the stream length is necessary has remained a central open problem. In this work, we resolve the dependence on $n$ by proving that every differentially private mechanism for continual counting must incur expected $\ell_\infty$ error $Ω(\log^{3/2} n)$. This shows that the binary tree mechanism is asymptotically optimal in the approximate-DP setting. As a consequence, we also obtain a largest-possible separation between hereditary discrepancy and private $\ell_\infty$ error for linear queries, showing that the known general upper bound in terms of hereditary discrepancy has the optimal dependence on the number of queries.
Rhetorical strategies and their influence on audiences are often studied through social media posts and comments. However, this focus overlooks the universal audience, which is the majority of readers who remain silent and do not explicitly express how a message affects them. This study investigates how two classical modes of persuasion, ethos and pathos, resonate in the silent audience's interpretations of meaning. Using a dataset of social media sentences paired with human-written interpretations, we label both sources for ethos and pathos and assess whether these rhetorical appeals are preserved. Our analyses show that interpretations diverge from the original sentences in 30% of cases, with rhetorically charged content eliciting greater variability than neutral content. We further find that ethos and pathos in original sentences can predict audience attitudes toward the author, underscoring the subtle ways rhetoric shapes perception beyond visible engagement.
Self-evolving agents violate the assumption behind most learning-theoretic guarantees: the data, evaluator, components, and hypothesis space are produced by the policy being updated. We present \textbf{SEA}, an architecture that confines self-modification to a small steering adapter and a versioned harness around a \emph{frozen} base model and admits each modification only through an anytime-valid gate that emits an auditable certificate against a fixed error budget. Five loop controllers compose published guarantees; because such gates can only \emph{select} among behaviors the frozen base already produces, five verifier-in-the-loop mechanisms -- best-of-$N$, micro-step search, self-authored reproduction oracles, search-layer control, and self-repair -- supply the dense, grader-free signal the gates require, computed from the issue text alone. On a $52$-instance SWE-bench Verified subset across four base models, base capability is the dominant, confound-free effect, and on two strong base models a deliberate no-op-composite control isolates the suite's contribution at $+4$ and $+5$ (\textsc{Glm}~5.2 $24\to28$; \textsc{Gpt} $29\to34$, the $65\%$ best), with event logs confirming that its mechanisms fire and prevent regressions. Results are single-run on expensive evaluations; confirming run-to-run variance and adapting the per-task algorithm mix are future work.
We study inference-time pattern-memory gating in a production-scale clinical natural language processing (NLP) pipeline. The pipeline pairs a generator (Llama-3.3 70B) proposing extractions with a verifier (MMed-Llama-3.1 70B) accepting or rejecting them, over 167,034 PMC-Patients narratives, and adds a lightweight memory that learns at deployment which extractions to filter, so the verifier need not re-examine candidates already seen to fail. We report four findings. First, learning filtering rules directly from the verifier's rejections failed at full scale: the relation-extraction filter stayed empty despite 785,797 logged rejections, because they were spread too thinly across too many distinct forms to accumulate. Second, a simpler rule using a fixed clinical ontology produced the same filtering without the verifier, capturing 49,734 ontology-violating relations on a held-out 5,000-patient set. Third, of five versions of the question-answering filter, four failed for distinct, instructive reasons; the fifth succeeded by checking whether a patient's extracted entities support the question asked, and where it applies was 1.84 times likelier to flag an answer the verifier would reject than one it would accept. Fourth, one pattern held across all five: a filter is selective only when it tests the same evidence the verifier weighs, not when it imitates the verifier's output. Together these give a transferable result for any generator-verifier pipeline: the most natural memory design can fail silently at scale, and whether a pre-generation gate is selective is decided before any engineering effort, by whether its signal probes the question the verifier itself answers. Throughout, the system flags suspect extractions rather than deleting them, so every decision stays visible for clinical review. All code and test artefacts are released openly.
Bayesian Optimisation (BO) under unknown constraints is particularly challenging when feasible regions are small. In such settings, existing methods that typically rely solely on evaluations of the true objective and constraints struggle to efficiently explore the design space. However, many real-world applications offer auxiliary data sources (e.g. surrogate models or simplified simulations) that can support early exploration. Despite this potential, their integration into constrained BO remains largely unexplored. We propose a general multi-source framework that extends constrained Max-value Entropy Search, capturing inter-source correlation while balancing evaluation cost and information gain. Experiments on both synthetic and physics-based benchmarks show that our method efficiently identifies feasible and optimal solutions, even when auxiliary data are only weakly correlated. The proposed approach consistently outperforms existing methods, particularly in early-stage exploration.
Large Reasoning Models (LRMs) have achieved remarkable success on complex tasks by leveraging long chain-of-thought (CoT) trajectories, yet they frequently exhibit overthinking on simple queries, resulting in significant token overhead and reduced inference efficiency. However, existing compression methods predominantly apply uniform length reduction or rely on coarse-grained difficulty estimation, often leading to performance degradation on difficult problems. To address this limitation, we propose Confidence-Adaptive Thinking (CAT), a framework that incorporates the model's intrinsic self-certainty signals as confidence into the preference optimization process, which autonomously modulates reasoning lengths based on problem difficulty. Experimental results show that CAT consistently outperforms state-of-the-art baselines on reasoning accuracy across multiple benchmarks on different base models. Our work enables LRMs to effectively compress confident responses while deliberating on uncertain ones, offering a potentially robust solution for balancing accuracy and latency in practical industrial scenarios.
Millimeter-wave (mmWave) beam alignment plays a critical role in next-generation wireless systems, yet its efficient implementation remains challenging. Meta-learning and transfer learning have been explored to enable deep learning-based beam prediction models to rapidly adapt to unseen environments; however, existing meta-learning approaches adapt the entire network and are trained from random initialization, leading to a large number of updated parameters and a high meta-training cost, while transfer learning approaches restrict adaptation to part of the network but do not exploit episodic meta-learning, which explicitly trains the model over multiple tasks, to optimize the adaptation process itself. To overcome these limitations, we propose MTL-BA, a meta-transfer learning framework for beam alignment in millimeter-wave multiple-input single-output (MISO) systems that freezes a pre-trained convolutional backbone and meta-learns only lightweight Scale-and-Shift (SS) adapters together with a classifier head. Warm-starting from the pre-trained model and restricting adaptation to the SS adapters and classifier head reduce both the adaptation cost and the meta-training budget without sacrificing prediction performance. Simulation results on the DeepMIMO ray-tracing dataset show that MTL-BA matches the accuracy and spectral efficiency of full fine-tuning across various SNR levels despite updating approximately $17\times$ fewer parameters than both full fine-tuning and Model-Agnostic Meta-Learning (MAML), outperforms last-layer fine-tuning while updating a comparable number of parameters, and approaches MAML's performance while requiring $60\%$ fewer meta-training epochs.
Contrastive pre-training has propelled video-text alignment, yet models often inherit the critical limitations of their image-text predecessors like CLIP, resulting in entangled representations. These challenges are severely exacerbated by two fundamental properties in the video domain: Temporal Misalignment, where textual descriptions often correlate only to specific, constrained temporal windows, leaving other frames text-irrelevant; and Semantic Asymmetry, which dictates a sparse, bidirectional, and non-equivalent relevance between frame-level visual details and caption-level concepts. This failure persists whether captions are short and temporally disjoint, creating ambiguity, or long and detailed, fostering entanglement between static objects and their temporal evolution. In this paper, we establish theoretical conditions that enable flexible alignment between video and text representations across the temporal dimension and at varying levels of granularity. Building on these theoretical insights, we introduce MoVA, Modular Long Video-Text Alignment, which learns dual asymmetric projections: a text-side projection that adaptively selects frame-aware subspaces of the caption, and a video-side projection that disentangles text-relevant visual concepts. Our framework ensures that the model can preserve global cross-modal semantics while disentangling evolving, frame-specific concepts and scale naturally to long captions and videos. Empirical evaluations show that MoVA outperforms existing methods in multiple video-text alignment tasks, demonstrating the effectiveness of our method.
In recent years, large language models have achieved remarkable success and have seen growing adoption in financial applications. At the same time, explainability remains critical in finance, a domain characterized by high stakes and strict regulatory requirements. Although numerous methods have been proposed to explain black box machine learning models, the majority of these approaches are designed for general purpose tasks and do not incorporate domain specific knowledge. In this work, we study the explainability of financial textual data modeled by large language models through the lens of the Shapley value. Specifically, we investigate whether Shapley based attributions align with established financial domain knowledge. Through rigorous theoretical analysis and extensive empirical evaluations, we demonstrate that Shapley values can yield explanations that are consistent with financial reasoning and can offer meaningful insights into the model's behavior in text based financial applications.
This work studies the hidden-state inversion problem: recovering the original input token sequence of a decoder-only language model from its last-layer hidden states. Rather than treating inversion as a one-shot reconstruction, we study it as a continuous embedding-space optimisation in which a soft proxy is driven towards the leaked target without any hard-token projection during the search, and a token is committed only once, at the end of the inner loop. This design choice has two consequences which are the main focus of this paper. First, keeping the optimisation entirely in continuous space exposes a rich set of internal signals: rank trajectories of the ground-truth token, per-position loss curves, and a discrete loss measured at commit time. Second, the discrete loss allows assessing the correctness of recovery via cumulative discrete loss. We further analyse which tokens break the reconstructions and find a sharp categorical asymmetry: space-prefixed, high-frequency function words in dense regions of the embedding matrix dominate the failures, while content-bearing tokens are recovered almost perfectly. On 10-token C4 prompts the exact-match rate rises from 66.9% to 97.5% (mean similarity 0.994) as the candidate window is widened, confirming that most errors are recoverable near-misses rather than genuine ambiguities. A comparison with the released SIPIT reference situates these findings: per-step hard projection is faster, but the continuous formulation is what makes the optimisation observable and its failures detectable. The results show that last-layer hidden states of GPT-2 are as sensitive as the original text.
Most self-supervised learning (SSL) methods encourage invariance across augmentations, but strict flip invariance can suppress informative left--right correspondences in approximately bilateral data such as medical images and human faces. We propose Mirror-Fusion-Augmented Self-Supervised Learning (MFASSL), a Vision Transformer framework that injects a soft reflection prior into standard SSL without redesigning the backbone. MFASSL constructs mirror-paired views aligned to an estimated symmetry axis and introduces a lightweight Mirror-Fusion Attention (MFA) module for adaptive token-level interaction between mirrored regions while preserving asymmetric cues. The base SSL objective is further coupled with reflection-consistency and mid-layer token-alignment losses. Across CheXpert, BraTS, CelebA-HQ, and WFLW, MFASSL improves downstream performance, calibration, and reflection robustness over MoCo-v3, DINO, and MAE baselines under matched ViT-B/16 settings. It also achieves stronger and more consistent gains than recent equivariant SSL approaches with only approximately 2.7\% additional parameters. These results show that lightweight geometry-aware priors can effectively complement invariance-based SSL.
News articles are an important source of information on disaster impacts and adaptation. A key methodological challenge in socio-environmental studies is how to select a representative data sample. Two approaches are common: querying news databases top-down with the aid of an existing disaster inventory or using NLP methods to cluster news texts bottom-up based on temporal and spatial features. Using a dataset of German news about landslides worldwide, we compare these approaches and discuss variations in event coverage. Such research design decision can influence the resulting news sample, affecting its use in studies of inequality in media coverage, disaster monitoring and inventory enrichment.
In this opinion paper, we propose MetaHOPE, an error severity-aware annotation framework for evaluating metaphor translations. Metaphors present challenges for machine translation (MT) and natural language understanding and processing (NLU, NLP), because it presents the features of semantic complexity, contextual dependency, and cultural embeddings that can lead to ambiguity issues for NLP models. To investigate how state-of-the-art NLP models perform on translating metaphors, we select three representative systems, i.e., GoogleMT, GPT5.4, and Hunyuan-7b as Neural MT (NMT) models and LLMs. We used two human-annotated metaphor corpora, including VUAMC and PSUCMC for English-to-Chinese and Chinese-to-English translation purposes. The original corpora we used are monolingual, where we carried out error annotation using the MetaHOPE framework, and also produced the human post-edited gold reference for bilingual use as a new resource. We believe the MetaHOPE evaluation framework for metaphor translation annotation, the parallel corpora resources, and the error analysis on SOTA automatic translation models can be useful and shed some light for the field of metaphor translation study. We share our resources publicly upon paper acceptance.
World models are increasingly used in embodied intelligence and generative simulation, yet their scope remains ambiguous across communities. This tutorial presents a design-space view of world models as action-conditioned predictive models that estimate the future evolution of task-relevant observations or states. We categorize existing methods into observation-space and state-space world models, comparing their trade-offs in visual fidelity, spatial structure, physical interpretability, and control usability. We further introduce world action models, which connect predicted futures with executable robot actions, and summarize four representative paradigms: imagine-then-execute, video-feature-conditioned action prediction, joint video-action modeling, and auxiliary video prediction for policy learning. The goal of this tutorial is to clarify the conceptual scope of world (action) models and provide a structured taxonomy for embodied prediction and control.
Near-Infrared (NIR) spectroscopy has emerged as a promising alternative to traditional soil analysis methods, offering advantages such as speed, low cost, and non-destructive testing. This work proposes a machine learning (ML) approach to calibrate predictive models for carbon (C) and nitrogen (N) content in Oxisols and Inceptisols, utilizing NIR spectral data acquired with a portable MyNIR device. Various preprocessing methods were evaluated, with the most effective being the Savitzky-Golay (SG) filter and a robust outlier removal method based on the Nonlinear Iterative Partial Least Squares (NIPALS) algorithm coupled with a Huber loss function. Multiple validation strategies were compared, including 10-fold cross-validation, leave-one-out, and holdout via the Kennard-Stone method, followed by standardization. Stacking ensemble learning models were employed, using Partial Least Squares (PLS), Support Vector Regression (SVR), and Ridge as base models, with linear regression as the meta-model. The models were evaluated using R2, Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Ratio of Performance Deviation (RPD) metrics. The performance gap between soil types suggests the influence of pedological characteristics. Furthermore, the models achieved an RPD > 2.0 with low overfitting, validating the potential of this approach for rapid C and N quantification. This study contributes to the optimization of sustainable agricultural practices, aligning with the demand for efficient and environmentally friendly analytical methods. The developed technique enables faster decision-making for producers and consultants based on organic matter content, fertility indicators, and nutrient availability.
A single panorama captures the full visual sphere from one camera center, yet confines users to looking around in place without enabling true scene exploration. Converting a single panorama into a persistent, renderable 3D representation for free-viewpoint navigation has attracted growing interest; existing methods either adopt iterative per-view completion that propagates inpainting results to update the underlying geometry, leading to progressive error accumulation and cumbersome multi-step pipelines, or leverage the temporal consistency priors of video generation models, yet the continuous-trajectory constraint intrinsic to such models limits their flexibility in covering scenes from multiple directions simultaneously. We present Pano2World, which takes a single indoor panorama as input and directly outputs a persistent, explorable 3D Gaussian scene. Given the source panorama, Pano2World first reconstructs a coarse 3D Gaussian proxy and renders it at adaptively sampled nearby poses to obtain geometrically aligned guidance panoramas; a panoramic diffusion model then jointly denoises all target views via View-Aware Attention Routing, where each target view simultaneously receives geometric constraints from its corresponding guidance panorama and global semantic guidance from the source panorama, naturally enforcing cross-view consistency. To avoid the information loss incurred by decoding the multi-view hidden features formed during joint denoising back to the pixel domain via VAE, we introduce Latent Feature Adapter, a geometry-aware bridge module that directly distills these hidden features into a scene latent, subsequently decoded into the final 3D Gaussian scene. Experiments demonstrate that Pano2World significantly outperforms existing methods on the multi-position panoramic novel-view synthesis benchmark.
The landscape of automated formal verification is populated by techniques that make prominently different trade-offs: some focus on expressiveness and precision, supporting the verification of complex properties; others favor scalability and practicality, so that they are applicable to larger programs using different features. This paper presents Petrify, a novel automated verification technique for concurrency properties that achieves a distinctive trade-off. Petrify encodes the semantics of Java bytecode programs into Petri nets (PNs), which can be analyzed by state-of-the-art model checking tools such as LoLA. As our experiments demonstrate, Petrify's approach offers an interesting combination of expressiveness and practicality: PNs are a fairly precise encoding of the concurrent behavior of programs; at the same time, Petrify's PN encoding is succinct, so that its analysis remains quite insensitive to parameter size. Another practical benefit of targeting bytecode is that jPetrify, the prototype tool that implements the Petrify technique, is applicable to programs written in any version of Java and even a subset of Kotlin (another language that compiles to Java bytecode) while other similar tools are limited to older versions of Java. While this paper's experiments focus on analyzing fundamental properties like deadlock, Petrify's approach lends itself to be extended to other kinds of concurrency analysis, which we plan to tackle in future work.
Large language models (LLMs) are increasingly used to generate queries, invoke tools, and construct analytical workflows. Although recent advances have substantially improved workflow generation and execution, the semantic information required to operationalize analytical concepts often lies beyond what is explicitly represented in database schemas and data values. We present a cross-domain formative study of operationalization failures in agent-generated analytical workflows. Across 236 analytical intents spanning finance, human resources, and public safety domains, we identify 153 recurring failures despite successful workflow generation and execution. Our analysis reveals five recurring classes of failures: comparative grounding, process reasoning, quantitative reasoning, role confusion, and policy grounding. These findings suggest a semantic gap between user-level analytical concepts and the information available to workflow-generation systems. More broadly, they raise questions about the admissibility of analytical operations and suggest that future agentic data systems may require richer semantic representations to bridge the gap between analytical intent and executable computation.
Frontier foundation models have changed the math on vulnerability discovery, but the bigger challenge is how the remediation side keeps up. Despite recent progresses in Automated Vulnerability Repair (AVR), current solutions struggle to reliably identify the root causes of vulnerabilities, and insufficiently utilize the prior fix knowledge to guide the patch generation process, thus undermining their effectiveness in practice. To address this gap, we propose KeaRepair, a novel agentic AVR approach that grounds patch generation in verified program facts and high-level vulnerability knowledge. Specifically, KeaRepair first extracts multi-dimensional vulnerability knowledge from historical vulnerability-patch pairs from dual complementary views, and constructs dedicated retrieval knowledge bases. It then employs a tool-augmented agent that performs ReAct-style reasoning to collect verified program facts for vulnerability diagnosis. Finally, based on the diagnostic results, KeaRepair performs knowledge-level retrieval-augmented patch generation and iteratively refines patches through a closed-loop validation process involving compilation, PoC replay, and test-suite execution. Experimental results show that KeaRepair significantly outperforms existing AVR approaches on 55 reproducible C/C++ vulnerabilities. When paired with Gemini-3.1-Pro, KeaRepair successfully repairs 46 vulnerabilities, achieving a repair rate of 83.64%. Moreover, KeaRepair fixes six unique vulnerabilities that none of the baselines can address, and further demonstrates strong cross-language generalizability.
SAT solvers settle combinatorial problems beyond the reach of interactive theorem provers and produce LRAT certificates for independent verification. We present LRAT-Catcher, a standalone, general-purpose tool that imports a DIMACS formula together with an LRAT certificate into Lean 4 as a theorem. LRAT-Catcher runs the formally verified LRAT checker from Lean core as compiled native code via reflection. This scales to instances where Mathlib's explicit proof-term import exhausts memory. LRAT-Catcher also composes cube-and-conquer solving runs entirely inside Lean. Per-cube refutations are combined with a cover-completeness certificate, itself an LRAT proof, into a single unsatisfiability theorem. Verified encodings connect CNF-level results to the original combinatorial problems. We evaluate the tool against Mathlib's proof-term import and the external checker cake_lpr on establishing the Schur number S(4) = 44 and the Ramsey number R(4,4) = 18 as Lean theorems.
Unsupervised pre-training on large-scale datasets has demonstrated significant potential for improving the sample efficiency and performance of Reinforcement Learning (RL). Given the large-scale action-free internet videos, existing methods utilize single-step transition prediction and image reconstruction to learn representations. However, these methods prefer to preserve large-proportion stationary information in the pixel space, neglecting small but crucial information. To preserve enough information in the representation, it is essential to pay equal attention to each element in videos. Specifically, we propose a temporal correlation space to distinguish each element. For implementation, we introduce the Multi-scale Temporal Contrastive Learning (MTCL) method to model multi-scale temporal correlations separately. This approach can balance the attention of different elements and yield more informative representations, effectively supporting policy learning in various downstream tasks. Experimental results demonstrate that our method improves sample efficiency and asymptotic performance across various downstream tasks.
Pre-training on large-scale videos to improve reinforcement learning efficiency is promising yet remains challenging. Existing methods typically treat the agent as an indivisible entity, modeling motion patterns globally. Such global modeling is tightly coupled with the morphology, hindering transfer across domains. In contrast, despite the vast disparity in global motions, the local components exhibit similar motion patterns across different agents. Building on this insight, we propose a novel Deconstruct-Recompose Paradigm (DRP) for learning transferable local motion representations. Specifically, in the Deconstruct phase, we identify multiple local points and track their frame-wise motions, defining each as an Atomic Action. We introduce a Dual-Attention Encoder (DAE) to learn local motion representations from these Atomic Actions, capturing their spatiotemporal relationships. In the Recompose phase, we compose local motion representations with a learnable Motion Aggregation Token [MAT] via latent dynamics model learning. Additionally, an adapter bridges local motion and downstream action-specific dynamics to accelerate policy learning. Extensive experiments demonstrate that our method effectively transfers to diverse robotic control and manipulation tasks, significantly improving sample efficiency and performance.
Visual Reinforcement Learning (VRL) has achieved considerable success in solving control tasks. However, generalizing learned policies to new environments remains a major challenge, as agents often overfit to task-irrelevant features in the training environment. To solve this problem, we introduce the concept of decoupling observations into task-relevant and task-irrelevant representations. Building on this idea, we propose a self-supervised Task-Relevant Representation Decoupling (T2RD) algorithm for VRL. This algorithm consists of three components: task-relevant representation consistency, cross-reconstruction, and cross-dynamic prediction. The first two components achieve the decoupling of content and style features, but the resulting content representations are not necessarily task-relevant. To further refine task-relevant features from content representations, we design the third component that introduces dynamic prediction. T2RD achieves State-Of-The-Art (SOTA) generalization performance and sample efficiency in the DeepMind Control Suite and Robotic Manipulation tasks.
For predictive models, the often-reported performance metrics are the loss and accuracy. In synchronous Brain- Computer Interface (BCI) systems, these metrics are informative for most BCI paradigms; however, for Event-Related Potential (ERP) applications the spelling rate, which measures the number of characters correctly selected is more important as it influences the estimation of information transfer rate (ITR) and any related metric measuring spelling performance. Moreover, ERP-based BCIs hold imbalanced data class distributions, which require reporting metrics that can handle the imbalance, such as the area under the receiver operating characteristic curve (ROC AUC). In this work, we study the correlation of the spelling rate with 13 metrics to identify which among them best reflect user spelling performance and how they are affected by trial repetition. The Results of two datasets (a private LARESI ERP dataset and the public OpenBMI ERP dataset) favor the Brier score, Matthews Correlation Coefficient (MCC), and the metrics that account for class imbalance in binary classification: ROC AUC, area under the Precision-Recall curve (PR AUC), Average Precision (AP), and partial AUC (pAUC). These findings encourage researchers and practitioners to report those metrics in ERP-based BCI experiments.
Vision-language pretraining remains dominated by contrastive objectives, whereas vision-only self-supervised learning has largely adopted non-contrastive methods. At the same time, the role of vision-language encoders has shifted: they are increasingly deployed not as zero-shot classifiers but as the frozen visual backbone of vision-language models and dense prediction systems, which consume the full grid of patch tokens rather than a single pooled embedding. We introduce LeVLJEPA, the first fully non-contrastive end-to-end vision-language pretraining method. LeVLJEPA learns through cross-modal prediction with stop-gradient targets and per-modality distributional regularization, without negatives, temperature, momentum encoder, or teacher-student schedule, and trains stably at large scale. We find that the resulting encoder provides markedly stronger dense semantic features for downstream use: as a frozen vision-language-model backbone, LeVLJEPA is the strongest of the evaluated encoders across GQA, VQAv2, and POPE under two distinct language models, and outperforms contrastive baselines on semantic segmentation, while remaining on par on global readouts such as linear probing. These results establish non-contrastive pretraining as an effective means of producing dense semantic vision features.
Recognizing jazz standards from audio is a challenging form of tune-level music retrieval: different performances of the same standard may vary in tempo, key, arrangement, instrumentation, improvisational content, and even whether the head melody is present. We study this problem using a curated subset of the Jazz Trio Database designed for cross-performance standard recognition. We compare a from-scratch trained Harmonic CNN baseline against frozen pretrained music representations from recent music understanding foundation models, using both supervised probing and nearest-neighbor retrieval. Our results suggest that from-scratch spectrogram models overfit strongly to training performances, while pretrained embeddings provide better top-$k$ results but are sensitive to performer identity, which can be partially reduced with a lightweight contrastive projection. Our findings motivate jazz standard recognition as a useful stress test for music representation models and as a step toward retrieval-based standard identification. Project page: https://github.com/cagries/tipofmyear.
Recent recursive Transformer studies have primarily reused shared parameters across computation steps to construct compact, parameter-efficient models. In this work, we leverage recursion to build effectively deeper Transformers with stronger representational capacity. However, in Vision Transformers, simply increasing recursion depth does not reliably improve performance, as existing recursive approaches do not fully utilize the intermediate representations produced throughout recursive computation. We propose Soft Mixture-of-Recursions (SoftMoR) and its Vision Transformer instantiation, Soft Recursive Vision Transformer (SR-ViT). SoftMoR learns token-wise mixture weights to softly combine outputs from all recursion steps, allowing intermediate representations to be utilized in a learnable and flexible way. Across diverse vision tasks, SR-ViT consistently improves as recursion depth increases with minimal parameter overhead. On ImageNet-1K, increasing recursion depth from 1 to 4 improves SR-ViT-S top-1 accuracy from 79.83% to 82.48% with only 1.7M additional parameters, outperforming the substantially larger DeiT-B while using approximately 27% of its parameters. These results demonstrate that SoftMoR provides a parameter-efficient path to deeper and stronger Vision Transformers through recursion.
Discrete diffusion models are widely used for learning and generating discrete distributions. As the generation process is inherently sequential, the acceleration of sampling is of significant importance. In this work, we parallelize the mainstream $τ$-leaping algorithm for absorbing discrete diffusion in a Continuous-Time Markov Chain (CTMC) framework. By leveraging the continuous-time stochastic integral form of the $τ$-leaping algorithm and the Picard iteration method, we achieve parallel-in-time sampling acceleration and provide a proof of exponential-factorial convergence for our algorithm. We improve the overall time complexity of $τ$-leaping under absorbing settings from ${\mathcal{O}}(d \log S)$ to ${\mathcal{O}}(\log (d\log S)\cdot \log d)$ with respect to NFE. Empirically, our method shows consistent acceleration across synthetic and real-data settings. The new sampler achieves at most $7$--$9\times$ runtime speedup for synthetic distribution, and maintains the same quality with $50\%$ fewer NFE and $1.45$--$1.86\times$ runtime speedups in image/text tasks on a single GPU. Our research expands the potential of discrete diffusion models for efficient parallel inference, with broader implications for applications such as molecular structure and language generation.
Digital forensic investigations of network intrusions require analytical outputs that are traceable, reproducible, and court-defensible - requirements existing machine learning pipelines do not satisfy, since they treat original evidence as training data and produce opaque classifications without instance-level justification. This paper presents a forensic-oriented intrusion detection framework resolving both problems simultaneously, integrating synthetic data generation, binary classification, and explainability within a single pipeline governed by ISO/IEC 27037, 27041, 27042, and NIST SP 800-86. The framework operationalises the ISO/IEC 27037 requirement for strict separation between original digital evidence and derived analytical artefacts. Original datasets are treated as immutable, hash-verified artefacts; all training operates on parameterized synthetic derivatives via SDV + CTGAN. XGBoost binary classification provides high-performance detection on tabular network flow data, and SHAP TreeExplainer produces instance-level feature attributions mapping statistical predictions to observable network behaviour for forensic reporting. Train-on-Synthetic, Test-on-Real (TSTR) evaluation on CICIDS2017 achieves F1-macro = 0.96, within cross-validation variance of the real-data baseline (0.97). Kolmogorov-Smirnov testing confirms synthetic privacy preservation (mean |KS| = 0.38) alongside operational utility. Cross-dataset validation on UNSW-NB15 and Kitsune identifies feature space dimensionality as the primary determinant of synthetic training effectiveness, establishing a practical deployment boundary of approximately 30 numeric flow-level features. SHAP attributions for Brute Force, Port Scan, and DoS attacks are consistent across real and synthetic instances, confirming synthetic training preserves forensically relevant attack fingerprints required for expert witness testimony.
Long-context LLM services now sustain prompts with hundreds of thousands to millions of tokens, making the key-value (KV) cache a first-order serving cost. Because the cache grows linearly with context length, it can exhaust GPU memory, force smaller batches, and reduce serving throughput. Prior KV cache compression techniques typically target only the sequence dimension or only the channel dimension, which leaves limited headroom as context windows scale. Compressing both dimensions promises higher memory reduction, but applying the two forms of compression directly leads to significant accuracy loss. This paper introduces MosaicKV, a dynamic two-D (dimensional) KV cache compression system for extremely long-context serving. MosaicKV uses dynamic two-D compression to address the accuracy challenge, exploiting the non-uniform importance distribution of elements within the KV cache. Instead of applying one compression pattern globally, MosaicKV identifies important elements for each KV vector and selects compression strategies at the granularity of KV cache segments. To address the performance challenge, where fine-grained sparsity and compression management overhead can offset the gains from compression, MosaicKV introduces compressed KV cache management. This mechanism uses underutilized GPU and CPU resources to maintain compressed KV caches and accelerate attention computation. Evaluation on an H800 GPU with multiple LLMs shows that MosaicKV delivers up to 16x attention speedup, 4.8x lower decode latency, and 7.3x higher throughput than the uncompressed baseline. At the same time, it reduces memory usage by 3x and incurs only 1.76% average accuracy loss on LongBench and RULER.
Table extraction from business documents relies on a cascaded pipeline where Table Detection (TD) first localizes tables and Table Structure Recognition (TSR) then recovers their internal layout. Building task-specific training sets for this pipeline is costly, particularly for TSR which requires fine-grained structural annotations. Active learning (AL) can reduce this annotation burden, yet most AL strategies are designed for single-model tasks and do not account for inter-stage dependencies in cascaded architectures. In this work, we present the first adaptation of Uncertainty Herding (UHerding), a hybrid coverage-uncertainty sampling method originally proposed for image classification, to cascaded object detection pipelines. We propose two pipeline-aware extensions that exploit the TD-to-TSR dependency: RankFusion adds dual-manifold coverage over both detection and structure representation spaces, while CAPA further incorporates stage-dependent gating and per-task uncertainty calibration. Extensive experiments across two public (PubTables-1M and FinTabNet) and two private table extraction datasets, with various annotation budgets (from 71 to 500 documents) show that UHerding generalizes well to table extraction, outperforming each baseline. Among pipeline-aware variants, RankFusion achieves higher expected gains but at the cost of greater variance, while CAPA emerges as the most consistent strategy, outperforming standard UHerding on three out of four datasets.
The bird's-eye view (BEV) representation enables multi-sensor features to be fused within a unified space, serving as the primary approach for achieving comprehensive 3D perception. However, the discrete grid representation of BEV leads to significant detail loss and limits feature alignment and cross-modal information interaction in multimodal fusion perception. In this work, we break from the conventional BEV paradigm and propose a new universal framework for multi-modal fusion based on 3D Gaussian representation. This approach naturally unifies multi-modal features within a shared and continuous 3D Gaussian space, effectively preserving edge and fine texture details. To achieve this, we design a novel forward-projection-based multi-modal Gaussian initialization module and a shared cross-modal Gaussian encoder that iteratively updates Gaussian properties based on an attention mechanism. GaussianFusion is inherently a task-agnostic model, with its unified Gaussian representation naturally supporting various 3D perception tasks. Extensive experiments demonstrate the generality and robustness of GaussianFusion. On the nuScenes dataset, it outperforms the 3D object detection baseline BEVFusion by 2.6 NDS. Its variant surpasses GaussFormer on 3D semantic occupancy with 1.55 mIoU improvement while using only 30% of the Gaussians and achieving a 450% speedup.
Prenatal anomaly classification and localization is of critical importance for fetal health and pregnancy management. Although ultrasound (US) is the primary modality for prenatal screening, accurate diagnosis remains challenging due to the low prevalence and high heterogeneity of anomalies. Existing deep learning methods for prenatal tasks rely on large-scale annotated datasets, which are difficult to obtain in practice. Although few-shot learning alleviates data scarcity, it typically requires fine-tuning for new categories, limiting its practicality in resource-limited clinical settings. To address these challenges, we propose a training-free framework for multi-class prenatal US anomaly classification and localization that operates with only a few reference images per class, representing the first exploration of this setting. Our framework comprises three key components: (1) a memory bank with multi-granular prototypes that explicitly models both class-level semantics and anomaly characteristics; (2) a prototype-driven soft merging mechanism that aggregates discriminative features to detect the anomaly region; and (3) a class-aware refinement strategy that leverages prototype consistency to improve category prediction. Extensively validated on a multi-center prenatal US dataset containing 1,149 cases, with a total of 2,357 images and 9 categories, our proposed method outperforms the competitors.
Microservice availability is commonly assessed by fault injection and chaos experiments, but such experiments are costly, operationally risky, and difficult to repeat for every architectural change. Distributed tracing and deployment metadata provide cheaper evidence, yet they usually remain descriptive: they show which services interacted, not what endpoint-level availability property follows. This paper proposes a formal runtime availability model based on stochastic connectivity for resilience-oriented analysis of microservice endpoints. It treats endpoint availability under explicit fault scenarios as a measurable facet of microservice resilience, combining a typed service-dependency graph, a replication map, a probability measure over node and edge states, and request-specific success predicates. Its semantics separates computational failures of service replicas from communication failures of logical dependencies, showing that replication cannot compensate for bottleneck dependencies. The model can be reconstructed from traces and deployment artifacts, parameterized for architectural what-if analysis, and analyzed by Monte Carlo simulation before or alongside fault injection. We define the model, its trace-to-model construction, elementary semantic properties, and a synthetic adequacy study. The study matches closed-form oracle cases within sampling error and exposes boundaries caused by edge bottlenecks, correlated failures, missing traces, and time-dependent failures.
Large language models can generate polished scientific text that includes unsupported claims, allowing hallucinations to enter the archival record. Assessing this risk via technical statements is difficult and often requires expert judgment, but citations provide a more auditable surface: a reference either resolves to a real scholarly work with compatible authorship, or it does not. We measure citation hallucination in peer-reviewed proceedings using a conservative definition limited to identity-level failures: non-existent works and substantial author-list mismatches. We explicitly exclude ordinary bibliographic drift (e.g., venue/year differences, publication-status updates, minor name variants). To audit citations at scale, we build RefChecker, a verification pipeline that resolves bibliography entries against multiple bibliographic sources and escalates unresolved cases to web-search re-verification. We apply RefChecker to accepted camera-ready papers from ICLR, ICML, NeurIPS, and USENIX Security. Hallucinated citations have entered the archival record. While reference-level rates are usually below 1%, proceedings are large enough that paper-level failures are visible: in 2025, roughly one in twenty NeurIPS and USENIX Security papers contains at least two likely hallucinated academic-paper-like references under our strict definition. We also observe post-ChatGPT increases in several venues, including a tail of papers with 5+ failures in a single bibliography, and likely hallucinated citations even among award-winning papers. These results suggest peer review alone does not reliably enforce citation integrity, yet auditing is tractable (about 0.04$ per paper in one venue-scale scan). We open-source RefChecker for routine, reproducible citation verification before publication (https://github.com/markrussinovich/refchecker).
Table Structure Recognition (TSR) aims to recover the row and column layout of tables from document images, a key step in document understanding pipelines. Accurate TSR depends on precise boundary localization: small errors in row or column boundaries can propagate into incorrect cell assignments and structural inconsistencies. Yet detection-based approaches treat table elements as generic objects, ignoring a fundamental property of table layout: rows and columns play structurally distinct roles and their boundaries carry unequal importance. We propose an Edge-constrained Fine-grained Localization loss (EFL) that formalizes this structural asymmetry by encoding table-specific geometric priors into the training objective: row-like elements are supervised with emphasis on their horizontal boundaries, while column-like elements prioritize vertical boundaries. Implemented within a real-time detector with distribution-based boundary refinement (D-FINE), EFL operates during training only and guides boundary refinement toward structurally meaningful adjustments with no change to the inference pipeline. The proposed approach, ConRTF, is also data-efficient, maintaining robust accuracy with as few as 2k--3k annotated tables. Experiments on PubTables-1M and two private datasets show consistent improvements over the optimized baseline and several real-time detectors including RT-DETRv2 and YOLOv10-11, with gains of up to +1.6 GriTS points at equal inference speed.
Mechanistic modeling via ordinary differential equations (ODEs) provides interpretable descriptions of complex dynamics and enables inference of underlying mechanisms, which is particularly valuable in clinical settings. However, in rare diseases, both the structure and parameters of the model are typically unknown, while individual-level data is scarce, noisy, heterogeneous, and subject to privacy constraints. In such settings, population-level summary statistics provide a practical privacy-preserving data representation, while capturing heterogeneity further requires modeling parameters as distributions rather than fixed values. Yet no existing method jointly discovers ODE structure and refines parameter distributions solely from summary statistics. We present AgentODE, an end-to-end framework that addresses this gap. An LLM proposes candidate ODE structures, while a tool-augmented inference agent iteratively refines parameter distributions through a diagnosis--update loop, operating on population-level summary statistics alone. We evaluate AgentODE on three benchmark problems across different fields and two clinical datasets, including the rare disease recessive dystrophic epidermolysis bullosa (RDEB), with only 231 observations across 46 patients. AgentODE recovers functionally consistent ODE structures across all settings, and experiments on RDEB demonstrates that in sparse and noisy data settings reasoning from summary statistics promotes mechanistically principled structure discovery, whereas baselines with individual-level data access recover implausible structures despite better predictive performance. AgentODE opens new possibilities for mechanistic modeling of rare diseases directly from population-level summary statistics, where data scarcity and privacy constraints have traditionally limited such analyses.
Retrieval-augmented generation (RAG) under a fixed reader-context budget forces a selection problem: of the evidence retrieved, only a fraction can be shown to the reader. We argue that document recall -- the standard retrieval metric -- is the wrong quantity to optimize in this regime, and we make two contributions. First, as a general contribution, we introduce answer-in-context, a diagnostic that measures whether a gold answer survives as a contiguous span in the packed reader context (not the retrieved set). It predicts answer F1 better than recall (r=0.39-0.55 vs. about 0.31), separates answer quality roughly five-fold (0.60 vs. 0.12 on HotpotQA), and carries information beyond retrieval: it adds Delta R squared=0.17 over recall and shows a 4.6x EM gap even among questions where all gold was retrieved. We also confirm it interventionally: on 2WikiMultiHopQA a packing change that raises coverage but not answer-in-context yields no accuracy gain. Second, as a conditional contribution, we cast reader-context construction as budgeted monotone submodular maximization and build a packer that jointly optimizes relevance, query coverage, representativeness, and diversity. On HotpotQA with a 160-token budget and a 3B reader it beats a strong focused heuristic, MMR, and naive packing -- by up to +5.1 F1 at equal-or-lower token cost, across three seeds. Crucially, we map the scope of this win honestly: it requires the conjunction of (i) multi-hop complementary structure, (ii) retrieval that surfaces the evidence, (iii) a binding but not extreme budget, and (iv) a reader weak enough that evidence density, not reading capacity, is the bottleneck. A quantization-controlled reader-scale ladder (3B to 7B to 14B) shows the edge over the heuristic is absorbed by 7B and significantly reverses by 14B, while the diagnostic explains every boundary with a single variable.
Multilingual fluency often invites a stronger assumption: a model that can speak a user's language must also understand the culture encoded by that language. We call this the Illusion of Cultural Alignment. To test this assumption directly, we introduce MSQA, a benchmark of 1,064 natively sourced questions across 11 language groups, five cultural dimensions, and three difficulty tiers. Unlike translated benchmarks, MSQA targets locally grounded knowledge and reduces shortcuts from English-centric cross-lingual transfer. Evaluating 18 LLMs, we find substantial cultural degradation and a pronounced Locality Effect: cultural competence tracks pre-training exposure more closely than general reasoning ability. We further show that common inference-time remedies do not dissolve the illusion. Models remain overconfident on unfamiliar cultural questions, repeated sampling yields unstable rather than reliable correctness, and retrieval augmentation helps unevenly on long-tail facts. These findings indicate that cultural alignment cannot be inferred from multilingual ability alone and requires deeper intervention than calibration, sampling, or retrieval at inference time
Despite the increasing sophistication of industrial AI systems, the ability to reliably detect subtle and noisy anomalies in complex time series data remains a critical yet unresolved challenge. In large-scale industrial applications, labeling time series data is often prohibitively expensive and time-consuming, making unsupervised learning a practical and widely adopted approach. However, existing unsupervised methods frequently struggle to distinguish near-normal anomalies from normal patterns and are vulnerable to noise contamination within normal samples. To address these limitations, we propose a novel framework that leverages active learning to iteratively enhance the performance of unsupervised models. Our framework's core contributions are (1) a masked time-series reconstruction feedback strategy that forces the model to learn robust temporal dependencies, and (2) a minimax learning strategy that promotes robustness by differentially treating normal and abnormal samples. This process encourages the model to better capture the dynamics of subtle and noisy patterns. The proposed framework is evaluated across 28 test cases involving four multivariate time-series datasets and seven unsupervised backbone models. Experimental results demonstrate a 12.39% improvement in AUC compared to the original models, confirming that our method can be readily integrated into existing unsupervised reconstruction-based anomaly detection systems to significantly enhance their performance.
Skeleton-based action recognition has achieved remarkable success by exploiting joint coordinates and their topological connections, yet prevailing methods overwhelmingly assume complete and clean skeleton inputs. In real-world deployments, such as egocentric vision, crowded surveillance, wearable devices, or edge robotics, limited field-of-view (FoV) frequently causes substantial joint visibility dropout, leading to severe performance degradation that existing models are largely unprepared to handle. To bridge this critical yet underexplored gap, we introduce PartialVisGraph, a novel hypergraph framework tailored for robust skeleton action recognition under constrained FoV. We first construct highly expressive hypergraphs by introducing learnable virtual hyperedges that form a soft incidence matrix, capturing flexible high-order dependencies beyond conventional pairwise graphs. We then propose the Single-Head Sample-Adaptive Transformer, which adaptively aggregates joint features onto hyperedges while explicitly incorporating a visibility prior. This prior selectively gates information flow, preventing occluded or out-of-view joints from corrupting reliable feature propagation. We further establish rigorous evaluation protocols with realistic FoV simulation benchmarks on NTU RGB+D 60 and 120. Extensive experiments demonstrate that PartialVisGraph consistently achieves state-of-the-art accuracy under partial visibility, with gains of up to 68.8\% on subsets with severe FoV restrictions compared to recent strong baselines, while remaining superior on full-visibility settings. Our approach offers a principled and practical pathway toward deployable skeleton-based action understanding in unconstrained environments.
Self-conditioning is a core technique that enhances continuous flow-based language models, where the model learns to denoise generated text by conditioning on its own denoising estimate. While empirically successful, its performance improvements are poorly understood. Moreover, there is growing interest in the use of few-step generators based on flow maps, for which how to leverage self-conditioning is unclear. Here, we show that flow language models with self-conditioning solve a fixed-point iteration that bootstraps the performance of the learned denoiser. We use this viewpoint to formulate fixed-point flows, a two-dimensional class of self-conditioned flows, where the first dimension represents the flow process and the second represents the fixed-point iteration. We show that fixed-point flows define valid flow maps, and show that they can be distilled from self-conditioned flow models by compressing both fixed-point iterations and the flow process, the former with fixed-point distillation and the latter with flow map distillation. Our resulting flow map language model, FMLM$^\star$, outperforms state-of-the-art self-conditioned models and few-step models in one- and few-step generation on OpenWebText. Code is available at https://github.com/Ugness/self-conditioned-fmlm.
Large Language Models have emerged as programming assistants. However, the efficacy of code generation is constrained by the quality of input requirements, which are frequently ambiguous, incomplete, or underspecified. While LLMs excel at one-shot code synthesis, their ability to proactively clarify intent remains underexplored, as a critical trait for robust software engineering. Existing benchmarks largely overlook this interactive bottleneck, assuming perfectly specified prompts that do not reflect the iterative nature of requirement elicitation. To bridge this gap, we introduce ClarifyCodeBench, a novel interactive benchmark for evaluating LLMs' capability in resolving requirement ambiguity. Constructed from real-world programming tasks, ClarifyCodeBench features high-quality manual annotations, including N unique ambiguity types, associated clarification questions, and corresponding ground-truth answers. Furthermore, we formalize two rigorous metrics to assess the interaction quality: Turn-discounted Key Question Rate, which penalizes inefficient questioning, and Optimal Round Adherence, which measures the precision of the elicitation process. We conduct a systematic evaluation of six state-of-the-art LLMs using ClarifyCodeBench. Our empirical results yield three critical insights: 1) Capability Decoupling: Strong code generation performance does not inherently translate to effective requirement clarification; 2) The Reasoning Paradox: While increased computational thinking enhances code correctness, it yields marginal gains in identifying ambiguities; 3) The Multi-ambiguity Ceiling: LLMs' clarification performance degrades sharply as the density of ambiguities increases, revealing a significant bottleneck in handling complex, real-world specifications. Our work underscores the necessity for future AI4SE research to transition from static synthesis to interactive elicitation.
Well-designed autonomous driving datasets have fundamentally shaped research progress, yet existing literature primarily describes what datasets contain rather than how to strategically design impactful ones. This is especially limiting for small and medium-sized labs and startups that cannot afford to misallocate scarce resources. We argue that impactful dataset creation begins with a diagnosis: whether a research question is blocked by a data problem or an evaluation problem, and proceeds by selecting the minimal data operator(s) that closes the resulting gap, recording new data only when no cheaper operator(s) suffices. We analyze the evolution of major autonomous driving (AD) datasets through this lens and distill a strategic framework spanning gap identification, operator choice, sensor suite design, and annotation strategy. We ground the framework in a running case study of our KITScenes dataset family. The datasets are available at: https://kitscenes.com/
LLVM is a widely used compiler infrastructure whose scale and complexity make issue resolution labor-intensive and challenging. Although large language models (LLMs) have recently achieved remarkable success in issue resolution, their effectiveness on complex system-level LLVM compiler remains largely unexplored. To address this gap, we introduce LLVM-Bench, the first large-scale benchmark for LLVM issue resolution, containing 423 real-world, validated tasks collected from the LLVM project. We further develop LLVM-Gym, a scalable evaluation platform that automates issue reproduction, patch application, compiler building, and test execution. Using LLVM-Bench and LLVM-Gym, we conduct a comprehensive study of four representative LLMs, six retrieval configurations, and three agents. Our results show that current LLM-based issue resolution techniques remain limited on LLVM-Bench, with patch invalidity and build failures as the dominant failure modes. We further reveal a strong complementarity among different LLMs and agents, motivating LLVM-Ens, a lightweight ensemble approach that expands the patch space through integrating the patches generated by diverse techniques, filters incorrect and redundant candidates, and identifies the most promising solution. Our results show that LLVM-Ens achieves a resolution rate of up to 21.99%, further improving LLVM issue resolution.
Long-horizon LLM agents accumulate tool results, files, plans, and user constraints that are too structured to be treated as a disposable text suffix. Current systems mostly rely on in-run heuristics such as chronological pruning and tool-output masking, or on final self-summary near a context limit. Heuristics are cheap but blind to future dependencies; summaries preserve narrative state but often hide exact evidence, locators, and editable artifacts. We present Self-GC, where GC denotes self-governing context while deliberately echoing garbage collection: the system does not merely reclaim unused tokens, but governs the lifecycle of agent context objects. Self-GC turns user turns, tool spans, and skill state into indexed objects; asks a side-channel planner to propose fold, mask, and prune actions; and lets the harness enforce recoverable sidecars, safe commit boundaries, and cache-aware commit. On a 33-session Hard Set, Self-GC prunes 43.95% of prefix tokens while leaving 84.85% of future continuations unaffected, compared with no-impact rates of 54.55% to 69.70% for heuristic baselines. On a 332-session production-derived suite, three planner backbones reach no-impact rates of 91.27% to 94.58%, while baselines remain at 77.71% to 87.46%. In production, an online account-level split reduces daytime average input tokens by 10% to 15%, with peak reductions near 20%. These results point to context management as runtime lifecycle control over indexed, recoverable objects rather than post hoc text cleanup.
Black-box optimization is a fundamental science and engineering tool that makes it possible to optimize objectives without gradient information. Unfortunately, as it often requires many function evaluations, it can be challenging when each one is costly. This is especially true when the evaluation function is noisy or failure-prone, and when high-performing solutions are confined to thin, curved, or disconnected regions of the search space. Existing methods leveraging generative models to navigate these subspaces are built to sample from reward-aligned distributions. As a result, they require a large number of evaluations to align their sampler effectively, making them impractical in low-budget settings. We propose SPARROW, an algorithm that completely decouples the generative prior from the reward signal. SPARROW can use any sampler with a known corruption process and trained on unevaluated data, as a fixed, structured proposal operator. Optimization proceeds by rank-based guidance over an archive of evaluated candidates. SPARROW can navigate complex geometries, handle unreliable reward signals, and perform effective optimization under very low evaluation budgets. We provide asymptotic convergence guarantees over the sampler support and demonstrate strong empirical performance on problems with unreliable rewards and geometrically complex landscapes.
Comparing transformer backbones for image segmentation is confounded: each is paired with a different decoder, recipe, and pretraining, so reported differences rarely reflect the backbone itself. We introduce the Lightweight Universal Mask Adapter (LUMA), a lightweight, backbone-agnostic mask-transformer head that treats any backbone as a black-box feature extractor, letting a set of queries read from its features through cheap cross-attention. LUMA matches the accuracy of EoMT, the state-of-the-art efficient ViT-segmenter, at lower cost, while attaching unchanged to isotropic, hierarchical, convolutional, and mixture-of-experts backbones alike. Holding this head fixed, we benchmark 20 backbones, 11 pretraining schemes and a range of resolutions on ADE20K and Cityscapes under one modern recipe. We find that ``efficient'' token mixers fail to deliver efficiency even at the high resolutions that motivate them, with plain ViT holding the throughput Pareto-front at every resolution. Additionally, the pretraining objective, not the architecture, the lever the field has tuned hardest, governs segmentation quality.
The classification accuracy of pretrained Vision-Language Models (VLMs) relies on the quality of the text prompts. Handcrafted templates and Large Language Model (LLM)-generated descriptions not only make predictions more interpretable, but also enable reuse of the same prompts across heterogeneous VLMs. Recent works construct task-adapted text prompts with a small number of labeled images. However, existing few-shot text prompting methods do not explicitly focus on misclassified examples during prompt construction, leading to only marginal improvements even as more shots become available. To fully exploit few-shot supervision, we propose Text Prompt Boosting (TPB), an AdaBoost-inspired framework that treats each text-prompt-based classifier as a weak learner and sequentially aggregates them into a strong ensemble by explicitly targeting hard, misclassified examples. Extensive experiments show that TPB preserves task-intrinsic, model-agnostic cues in text space, enabling robust cross-model transfer. Across eleven classification benchmarks, TPB improves accuracy on the source model and preserves shot-driven gains when transferred to larger, more capable VLMs, where existing methods struggle to sustain such improvements.
We study distributed online submodular maximization under partition matroid constraints, in which multiple agents select a limited number of actions from their own subsets sequentially to maximize the cumulative value of a sequence of objective functions. We develop a unified algorithmic framework that accommodates full-information and bandit feedback models. For both feedback models, we prove that the proposed algorithms achieve sublinear $(1-1/e)$-regret guarantees, which are comparable to those achieved by existing centralized counterparts. Furthermore, to tackle the sampling violation issue caused by continuous relaxation and rounding, we develop a bounded stochastic pipage rounding scheme and show that the probability of sampling violation vanishes asymptotically. As a result, the cumulative sampling violation remains sublinear in $T$, which is further shown to be not improvable under certain conditions. Numerical results validate the theoretical findings in this paper.
Graphs are a complex and versatile data structure used across various domains, with possibly multi-label nodes playing a particularly crucial role. Examples include proteins in PPI networks with multiple functions and users in social or e-commerce networks exhibiting diverse interests. Tackling multi-label node classification (MLNC) on graphs has led to the development of various approaches. Some methods leverage graph neural networks (GNNs) to exploit label co-occurrence correlations, while others incorporate label embeddings to capture label proximity. However, these approaches fail to account for the intricate influences between labels in non-Euclidean graph data. To address this issue, we decompose the message passing process in GNNs into two operations: propagation and transformation. We then conduct a comprehensive analysis and quantification of the influence correlations between labels in each operation. Building on these insights, we propose a novel model, Label Influence Propagation (LIP). Specifically, we construct a label influence graph based on the integrated label correlations. Then, we propagate high-order influences through this graph, dynamically adjusting the learning process by amplifying labels with positive contributions and mitigating those with negative influence. Finally, our framework is evaluated on comprehensive benchmark datasets, consistently outperforming SOTA methods across various settings, demonstrating its effectiveness on MLNC tasks.
Vision-Language-Action (VLA) models often fail to perform the same learned tasks under environmental shifts, such as changes in camera pose and shifts to a different but similar robot (e.g., from Panda to UR5e). Adapting these models to the shifted environment (i.e., target domain) often requires training on multiple demonstrations for each task, which are costly to collect. To reduce the burden of data curation and training, we propose an analogy-based method that adapts VLA models under environmental shifts through weight vector arithmetic with domain-specific information addition, named Domain ARiThmetic (DART). Unlike prior approaches, DART requires collecting only a single demonstration, enabling efficient adaptation. To accurately isolate domain-specific information for addition, DART performs subspace alignment between singular components in weight vectors to filter out noisy components. In both simulated and real-world experiments, DART outperforms existing VLA adaptation methods in one-shot scenarios across diverse visual and embodiment shifts. Code is available at https://github.com/snumprlab/dart.
We propose YOMI-Bench, a benchmark for evaluating kanji reading and phonological understanding of large language models (LLMs) for Japanese. In Japanese, a single kanji character often has multiple possible readings, making it difficult to infer the correct reading from surface-level text alone. Due to these linguistic characteristics, it is empirically known that LLMs exhibit low performance in kanji reading for Japanese. The proposed YOMI-Bench consists of four tasks specifically designed to evaluate kanji reading performance in Japanese. In our evaluation using YOMI-Bench, we assessed one multilingual open LLM, four Japanese-specific open LLMs, and five commercial LLMs. As a result, we found that even Japanese-specific models show low performance, and that commercial models also perform poorly on generation tasks that require consideration of kanji readings.
Explanations for emotion classifiers are usually produced post hoc, with no guarantee that they reflect the computation behind the label. We present an explication interface for event-based emotion analysis. A parser maps the input text to an explication, a short script in the closed vocabulary of Natural Semantic Metalanguage organized into twelve typed slots, and a fixed decision list of rules transcribed from published semantic definitions computes the label from the explication alone. The faithfulness guarantee is therefore causal and definitional, while all empirical risk lives in the learned parser, which the per-line entailment interface makes auditable against the input. On crowd-sourced event descriptions, our fine-tuned parser reaches 0.33 accuracy and 0.48 selective accuracy on a small held-out set, suggesting that the interface trades insignificant accuracy difference to a black-box model for a verifiable, inspectable decision basis for first-person event-based emotion analysis. We also release EmoExpl-1200 with per-line verification metadata and the full rule set.
Reinforcement learning has proven to be a valuable tool in the creation of advanced AI and robotic systems, contributing to everything from game playing to robotics to foundation models. Through trial-and-error, these AI systems typically learn one, near-optimal behavior to solve their tasks. However, there are many use cases in which one would like to assert some level of control, preferably in real time, over how the task is solved. We refer to these modifications of a core task as styles. We combine universal value function approximators (UVFAs) with carefully selected training scenarios, learning algorithms, and data augmentation to create a framework for coaching agents that exhibit styles in complex domains. We demonstrate the framework's application in the AAA video games Horizon Forbidden West and Gran Turismo, and in an open-source humanoid test domain. Despite the different nature of the domains -- car racing, stylized game combat, and humanoid walking -- each agent shows strong coherence to the style requests while still satisfying the main task in its domain. Importantly, the techniques outlined in this paper allow an end user to choose the final behavior at run time, giving them flexible control over the final executed performance.
Advances in generative AI are rapidly increasing the quality and commercial value of generated music, and this progress depends on large catalogs of creators' recordings. This raises a central question for platform design: how should creators be compensated when their work is used to train generative AI models that in turn produce commercial outputs? We develop a framework for fairly compensating creators in generative-music markets, where each creator's payment depends on a data-attribution score estimating their contribution to model outputs. Compared to past compensation frameworks, our framework has two unique considerations: (1) attribution is traced to entire creator catalogs, not individual songs, and (2) the informativeness (signal-to-noise ratio) of the attribution score is an input to the payment mechanism. The framework yields a closed-form payment rule per creator and measures the welfare cost of inaccurate attribution for both creators and the platform. Whether the welfare-optimal contract is royalty-based or takes the form of fixed-fee licensing depends on how informative attribution is for that creator's catalog. We show that better attribution translates directly into welfare gains for both creators and the platform, yet under multi-platform competition a platform only captures gains from attribution improvements when its signal becomes the most precise in the market. To ground our framework in empirical behavior, we train acoustic and symbolic music generation models and measure the informativeness of scalable attribution techniques against a leave-one-catalog-out ground truth. Our experiments reveal that noisy attribution signals push payment toward fixed-fee licensing and diminish welfare for both creators and the platform, providing an economic motivation for further research on improved attribution.
A growing class of tools recovers a program from observations of its behavior using an untrusted generator, a neural model or a search, that proposes candidates with no correctness guarantee. We study how to make such recovery trustworthy, in the concrete setting of recovering a runnable Scratch program from a recording of its execution. The recording shows what the program does but never its code; many programs produce the same video, so the source cannot be recovered, and the right target is a program that behaves the same as far as the camera can tell, made precise with a lens. The core is a two-tier validation oracle with a deliberate verdict asymmetry. A static checker proves lens-equivalence to a reference and issues a certificate that, granting the partial-order independence quotient adequate, never accepts a wrong program; a renderer can only refute or witness finite agreement, never certify. Around it, Vid2Prog reads each sprite's motion, visibility, and timing from the video and a known-asset manifest and synthesizes a candidate source-free; a closed loop renders and runs recovery again for ground truth. Under the exact lens the oracle makes no false accept on 246 labeled differing pairs, including an adversarial battery built to trap its concurrency quotient; on inputs outside the vocabulary and on real projects it abstains or refutes, accepting none we test. In-vocabulary recoveries reproduce their source frame for frame and 80% earn a static certificate, while whole real projects, mostly outside the vocabulary, recover at 14%, a vocabulary-bound rate the system never inflates with a wrong answer. A frontier vision-language model recovers none of the matched programs single-shot, which oracle-in-the-loop repair lifts only to a few while the structured pipeline recovers all, the gap a sound checker makes for an untrusted generator.
In settings such as fine-tuning and reinforcement learning, neural networks are often adapted under distribution shift. Standard adaptation methods typically optimize the target objective directly, inducing an abrupt change from the source training objective. This abrupt transition can distort learned representations, including features that may still be useful for the new task. We investigate whether a more gradual transition can improve adaptation. We propose loss smoothing, a simple approach that interpolates between the source and target training objectives at the start of adaptation. This smooth transition helps to preserve useful features from the source distribution while still enabling the model to specialize to the target distribution. Across controlled supervised shifts, pretrained vision adaptation, offline-to-online and online reinforcement learning, and language model fine-tuning, we find that loss smoothing consistently improves performance, suggesting that smoother objective transitions are a broadly useful tool for model adaptation.
Large language models (LLMs) are powerful pattern-completion systems, but their default operating mode - predicting the next token from a static context - does not reliably produce persistent, manipulable representations of an external world. Many tasks that look like "reasoning" in text become substantially harder once the environment is partially observable, stateful, and requires memory and structured hypotheses about hidden state. AGI Maze is a lightweight framework for building such environments without requiring high-dimensional sensory inputs. It provides a family of grid-based maze tasks with a clean API and multiple difficulty regimes. The goal is to create benchmarks where agents must learn and use world state representations, not just infer a local rule over readily provided observations. We provide an initial evaluation of several vanilla LLMs on simple mazes showing that they fail to represent mazes internally at LLM inference time. We also introduce a baseline agent, which is allowed to use its message history as a working memory to construct descriptions of observations at agentic runtime. Although this can improve performance, it is still insufficient for an LLM agent to reliably solve even small mazes within a step budget that is more than enough for humans.
Terminology such as "whitelist/blacklist," "master/slave," "man-hours," or "dummy value" has long been part of the technical vocabulary used in software artifacts, including source code, version histories, and documentation. In recent years, however, many of these expressions have been recognized as potentially non-inclusive and unwelcoming to groups historically underrepresented in software development, such as people of color, women, and individuals with disabilities. Consequently, a growing movement within the software industry has sought to replace these terms with more inclusive alternatives. Despite these initiatives, little is empirically known about how software developers perceive such terminology or how its continued use may influence their professional experiences and sense of belonging. This paper addresses the knowledge gap by examining how software developers perceive non-inclusive terminology in software and its perceived impact on team dynamics, productivity, belonging, and well-being. We surveyed open-source contributors and received 1,517 responses, of which 1,212 were complete and analyzed. On average, respondents reported low negative workplace impact overall; however, perceptions and impacts varied by demographic group. Women and non-binary participants, as well as respondents residing in the United States, were more likely to view the terms as non-inclusive. Among those who considered the terminology non-inclusive, non-binary participants reported higher overall negative impacts than male respondents, and female participants reported higher impact specifically on their sense of belonging.
Block-based languages such as Scratch let beginners assemble interactive programs from sprites and scripts. These programs are concurrent in practice: green-flag scripts, broadcasts, and clones run as cooperatively scheduled threads over shared sprite and stage state, and their authors never write a thread. We show that such programs contain schedule-sensitive behaviors whose observable result depends on an execution order the language leaves open. Editing, saving, or remixing a project can produce a copy with the same blocks but a different layer order, changing the order the virtual machine starts scripts. We formalize the schedule space a Scratch virtual machine can realize as the permutations of the initial executable-target order, and define schedule-robustness against a lattice of observation lenses over a fixed horizon. A partial-order exploration runs one schedule per dependence-equivalence class, and on projects small enough to enumerate, an independent oracle confirms it recovers every realizable outcome. On larger projects, representatives stand in for the factorial under the validated dependence model. SchedCheck implements this on the production Scratch VM. Across 224 real student projects, at least 21% of the concurrent ones are schedule-sensitive at the grading lens, and a uniform random sample of public projects replicates the rate at 17.6%, with two real remixes of a deployed animation arranging its letters differently. On hand-built fault pairs and a generated benchmark of 32 spec-defined faults across four classes, the tool detects and localizes every schedule fault, with a logic-fault control reporting clean. The oracle exposed four unsoundness gaps in the dependence model, all repaired. The method is parametric in the execution model, instantiating unchanged on a second cooperative event loop.
Generalized Category Discovery (GCD) aims to recognize known classes while autonomously discovering novel ones in open-world settings. However, current approaches primarily focus on designing clustering objectives, often overlooking a critical bottleneck: standard vision backbones yield high-rank, entangled token representations that are ill-suited for unsupervised discovery of latent concepts and structures. In this paper, we propose Compositional Primitive Fields (CPF-GCD), a novel representation learning framework that reshapes the feature space to make such latent structure identifiable by enforcing a low-rank compositional organization. Our core hypothesis is that all categories, whether known or novel, can be expressed as compositions and spatial arrangements of a finite set of learnable visual primitives that capture reusable concepts. CPF instantiates this geometric constraint via a spatial field mechanism. Inserted between the backbone and the head, it rewrites noisy patch tokens through low-rank primitive mixtures, effectively decomposing images into reusable atomic parts and their spatial layouts. By explicitly modeling the spatial distribution of primitives, CPF enables novel categories to emerge naturally as new activation patterns over a shared vocabulary. This shifts the focus of representation from merely partitioning global embeddings to constructing a structured and separable primitive field. Extensive experiments demonstrate that CPF serves as a generic, plug-and-play module that consistently boosts performance across diverse GCD baselines, validating that identifying and leveraging low-rank compositional structure is a crucial inductive bias for open-world recognition.
Limited Memory Language Models (LMLMs) externalize factual knowledge to a database to enable deletion-based unlearning without retraining. Existing evaluations measure post-deletion correctness in aggregate and cannot tell whether a deleted fact persists through residual parametric memory, alternative retrieval paths, or near-neighbor retrieval artifacts. We propose a causal auditing framework that holds the model fixed and varies the database state at inference time across three interventions: FULL, DEL-ON, and DEL-OFF. The framework decomposes post-deletion behavior into parametric leakage L(f), retrieval-mediated correctness R(f), and a retrieval artifact rate grounded in the inference-time retrieval trace. We apply it to 12,228 alias-closure deletions across thirteen databases, including four adversarial topologies (Base, Alias, Noise, Collision) we construct in three domains, and six prompt formulations. Parametric leakage is near zero in every variant and every prompt style: the model rarely returns the deleted answer in the absence of retrieval. The residual that does survive lives in the retrieval graph: retrieval-mediated correctness and the retrieval artifact rate match within rounding everywhere, so post-deletion correctness is, in our audit, predominantly reconstituted from near-neighbor retrieval. This residual ranges from 0.7% on the released LMLM database to 13.6% on the most adversarial variant, and prompt formulation does not independently control how much of a deleted fact survives. These results suggest that, for this class of LMLM and deletion procedure, the unlearning boundary is drawn primarily by the database administrator rather than by the model.
We give a descent-free, alignment-free measurement of singular structure on trained networks. At a single frozen checkpoint the read recovers the order $k$ of each dead direction from the directional-Fisher rate, the master invariant from which the per-direction learning coefficient $1/(2k)$ follows exactly, in whatever basis the optimizer left. The same read classifies each direction, separating a genuine singularity, whose order the architecture fixes, from a flat gauge symmetry; the directional-Fisher magnitude settles the cases the order cannot. A pluggable detector supplies the directions for transformer, convolutional, and normalisation layers. The read recovers the architecture-predicted order across constructed cells and trained networks, including a fine-tuned vision transformer whose dead structure is the LayerNorm-kernel gauge and a from-scratch one whose compressed MLP forms a node-death at its activation order. Where the singular structure enumerates, the per-direction orders assemble, through the typed intersection of the loci, into the global coefficient $(λ, m)$ matching the closed form. The method removes the canonical-alignment and descent preconditions of the underlying rate result, turning order-recovery into a deterministic, architecture-general reading. We then map its reach into the Watanabe triple: the order determines the universal singular fluctuation $ν(k)$, though a trained network's realized $ν$ falls below it as the live structure absorbs the dead direction's data fluctuation, and the multiplicity recovers from the dominant structure under a single-locus assumption.
The game of Taboo requires describing a target word without using a set of forbidden words, so that other players can guess it. This deceptively simple task combines strict lexical constraints with the need for communicatively effective descriptions, making it a compelling playground for examining how LLMs navigate competing demands at inference time. We evaluate two open-weight models under conditions that intervene at progressively deeper levels of the generative process, from prompting to generation-time constraints to internal representations manipulations. We assess their outputs through forbidden word violation detection, LLM-as-a-judge measuring the degree to which generated descriptions successfully evoke the target concept for both human and machine guessers, and examining whether the strategies models adopt under constraint align with those of human players. Our results show that compliance with the rules of the game and communicative effectiveness trade off differently across conditions, and that models remain substantially weaker than humans as guessers, suggesting that lexical grounding under constraint is an open challenge for current language models.
Scientific literature search often requires more than retrieving papers from a single query: users' intents are underspecified, preference-dependent, and evolve through interaction. Existing search agents typically rely on fixed pipelines or implicit language-only reasoning, making their search strategies difficult to control, inspect, and refine. We introduce PaperPilot, a multi-turn literature search agent that frames scientific search as workflow induction. Given an anchor paper and a user query, PaperPilot constructs an executable DAG of paper-search operators, including keyword search, citation expansion, filtering, scoring, reranking, and evidence extraction. User feedback is then used to refine both the query and the workflow itself. We train PaperPilot with supervised workflow imitation and preference optimization over controlled workflow corruptions. Experiments show that PaperPilot-9B improves over the base Qwen3.5-9B toolset agent under multi-turn interaction, increasing Hit@5 from 58.0 to 77.0, MRR from 47.5 to 59.4, and nDCG@10 from 26.8 to 32.5, while reducing workflow execution errors from 9.5% to 0%. These results show that explicit, editable search workflows provide an effective and controllable interface for aligning literature search agents with complex scientific intent.
Continuous diffusion language models such as ELF report record-low generative perplexity (Gen-PPL). We find a catch: these models repeat far more than human text, and Gen-PPL rewards rather than penalizes that repetition, so its low scores overstate quality. Strip the repetition and ELF-B's Gen-PPL rises from $19.5$ to $27.7$; the smallest model even posts the best Gen-PPL because it repeats most. We trace the repetition to its source: a contractive attractor along a \emph{single direction} in the self-conditioning feedback loop, the loop that feeds each step's clean estimate into the next. Because the failure is one-dimensional, a one-dimensional fix suffices, and we propose one. \textbf{ACE} (Attractor-Contrast-Escape) subtracts that single, label-free direction from the feedback at each step. Estimated once on the $105$M model, the direction cuts repetition to near the human level while keeping quality competitive, and transfers near-unchanged to the $342$M and $652$M models and across samplers; the same recipe recovers useful directions on other architectures. Since Gen-PPL itself rewards repetition, we instead measure the compute each fix needs to produce human-clean text, where ACE is $1.5$--$5\times$ cheaper.
We present a simple, yet general approach to study the scaling properties as the dimensionality of Metropolised MCMC sampling algorithms increases. The study relies ultimately on the symmetry of the Metropolis-Hastings formula. Our findings contain, as particular cases, many known results for the Random Walk Metropolis, MALA and other algorithms. In addition, they provide, in an easy way, new optimal scaling results for a variety of proposal mechanisms, including implicit proposals and proposals generated with the help of differential equation integrators. The analysis applies to targets that are products of a given, not necessarily univariate distribution, and also to cases where the different terms in the product are scaled differently. We show how to construct gradient-based MALA-like proposals where the variance of the proposal as the dimension $d$ increases may be taken as $O(1/d^μ)$, with $μ>0$ arbitrarily small, to be compared with the values $μ= 1$ for Random Walk Metropolis and $μ=1/3$ for MALA.
This study presents a comprehensive analysis of bird diversity across Sri Lanka by integrating spatial, temporal, and environmental data. Bird observation records were combined with environmental variables, including weather conditions, air pollution, the Normalized Difference Vegetation Index (NDVI), land cover, elevation, and Artificial Light At Night (ALAN), and rigorously preprocessed to ensure data quality. Spatial analyses were conducted on multiple grid scales (2 km, 5 km, 10 km) to evaluate patterns in species richness while minimizing sampling bias through spatial thinning. Temporal trends were assessed using effort-corrected metrics including rarefied richness and occupancy rates to account for variations in observation effort over time. Environmental drivers of bird diversity were examined using multivariate statistical models, including Poisson Generalized Linear Models (GLMs) and correlation analyses, to identify key associations between ecological factors and species richness. Additionally, community structure, dominance patterns, and beta diversity were analyzed to understand variations in species composition across regions and time. The study found that land-cover type is a stronger predictor of bird diversity than individual continuous variables such as NDVI or temperature alone. Urbanization, measured by ALAN, exhibits nuanced scale-dependent effects, supporting high abundances of a few generalist species while reducing overall richness. The findings provide actionable insights into the patterns and drivers of avian diversity in Sri Lanka, offering a scalable and reproducible framework for biodiversity research and conservation planning.
Sparse tangent portfolio optimization aims to learn an interpretable, low-cardinality portfolio in the tangency direction of the mean-variance frontier. However, the associated cardinality-constrained formulation is NP-hard, and standard predict-then-optimize pipelines often misalign forecasting accuracy with downstream portfolio quality. We propose an end-to-end decision-focused learning framework that reformulates Sharpe ratio maximization as a Disciplined Parametrized Programming (DPP)-compliant convex programming layer and replaces discrete selection with a smooth top-$k$ operator enforcing an exact cardinality $k$. This enables gradient flow through prediction, asset selection, and re-optimization, allowing the predictive model to directly optimize portfolio performance. Across four major equity markets, our method achieves competitive and often superior out-of-sample Sharpe ratios compared with historical and prediction-focused baselines, with particularly strong gains in larger asset universes. Our \href{https://github.com/feuerwerksh/Diffble-card-SR}{code} is publicly available.
Multi-image content has become an increasingly prevalent form of visual communication in social media, giving rise to a new safety issue, multi-image implicit toxicity (MIIT), where each image appears benign in isolation, but harmful semantics emerge when the images are interpreted jointly. MIIT is particularly challenging for existing commercial moderation APIs and models due to the lack of explicit risky cues in each image. This paper aims to study how to identify MIIT. We first provide a formal definition of MIIT and analyze three key challenges for its detection. To alleviate the scarcity of data in this area, we construct MIIT-dataset, an image-only multi-image safety dataset covering seven representative risk categories through an automatic generation pipeline. Finally, we train MiShield with progressively distilled reasoning supervision, enabling it to produce safety judgments accompanied by explicit analyses of the correlated entities that result in the hazards. Experiments show that MiShield-8B models outperform representative moderation services and even larger-scale models, revealing its effectiveness and practical value for this widely used visual format. Warning: This paper contains potentially sensitive content.
Understanding how aligned LLMs internally represent safety is critical for diagnosing alignment vulnerabilities, as it explains why jailbreaks succeed and informs the design of robust alignment strategies. Prior work shows that aligned LLMs encode harmfulness and refusal as separable directions in the residual stream at prompt-side token positions. We show that jailbreaks succeed at prompt encoding by suppressing either the refusal or harmfulness direction before any token is generated, with distinct attack classes occupying separable regions of the harmfulness-refusal plane. Extending the analysis to response-token positions, we find that the model recognizes harmful content while it is generating that content, even when it failed to recognize the input as harmful at the prompt side. Motivated by our findings, we introduce HARC (Harmfulness-And-Refusal Coupling), a fine-tuning method that pairs the two directions across both prompt and response positions. Since the intervention is confined to the harmfulness-refusal subspace, it leaves the rest of the residual stream intact and does not degrade general capability or inflate over-refusal. Across extensive experiments, HARC achieves the strongest robustness-capability-usability trade-off among six baselines spanning the major training-time and inference-time safety methods. The harmfulness and refusal directions at prompt and response positions transfer across the five model families and two scales we tested without architecture-specific tuning.
Retrieval-augmented generation (RAG) increasingly requires models to answer questions from multiple retrieved documents, where only some sources are relevant and the retrieved bundle may contain stale, noisy, or conflicting evidence. Existing contrastive decoding methods primarily focus on resolving conflicts between the model's internal memory and the retrieved context. In contrast, we study the complementary problem of intra-context conflict in multi-document RAG. To evaluate this setting, we introduce DRQA, a factual-conflict question answering benchmark derived from enterprise deep-research scenarios, where answers are grounded in synthetic enterprise-specific facts that are designed not to be recoverable from the model's internal memory. We further propose Dual-Confidence Contrastive Decoding (DCCD), a training-free decoding method that combines document-level confidence, which estimates whether a document appears sufficient for answering the question, with token-level confidence, which estimates whether that document supports a confident next-token prediction. DCCD selects positive and negative document-conditioned streams using these dual-confidence signals and scales a document-level contrast by their confidence margin. Across DRQA and standard multi-document QA benchmarks, DCCD achieves the best average performance among full-context and contrastive decoding baselines, with the largest gains on DRQA. These results highlight the importance of source-aware, confidence-gated decoding when retrieved evidence is internally conflicting.
Debugging the Linux kernel remains a formidable challenge due to its vast codebase, complex architecture, and low-level programming intricacies. Effective fault localization (FL) is thus essential for efficient kernel debugging and maintenance. While existing FL techniques (both traditional and LLM-based) have shown promise in general-purpose software, they are ill-suited for the kernel context. In particular, recent LLM-based techniques often treat bug reports and source code as plain text, lacking deep integration of kernel-specific knowledge, which limits their ability to identify root causes and achieve fine-grained localization. We present CoHiKer, a novel LLM-based FL technique tailored to the Linux kernel. CoHiKer introduces two key innovations: (1) contrastive reasoning, which identifies root causes by analyzing the behavioral divergence between carefully mutated passing and failing test cases, and (2) hierarchical context analysis, which systematically narrows the localization scope from files to methods by integrating crash reports, syscall semantics, inter-file dependencies, and kernel-specific features. Unlike prior techniques that rely on static understanding and full-code input, CoHiKer decomposes the localization task and enables structured LLM prompting to reason semantically over meaningful contexts. We evaluate CoHiKer on an extended Linux kernel bug dataset against five state-of-the-art baselines. CoHiKer consistently outperforms all competitors, improving Top-1 localization accuracy by up to 26.07% at the file level and 56.85% at the method level over state-of-the-art LLM-based baselines, while achieving up to 8.84% and 28.9% reductions in token consumption, respectively. Furthermore, CoHiKer demonstrates strong generalizability on the non-kernel dataset, with comparable gains (15.5% and 5.3% in Top-1 at file and method levels).
As Artificial Intelligence(AI)-based applications take off, a clear understanding of AI patterns can uplift the quality of AI applications. Many AI patterns have been proposed in the literature; however, their prevalence in real-life code has not yet been validated. Understanding the actual use of those patterns in practice can clarify our understanding both of the significance of these patterns and their utility. In this paper, we present a methodology to a) identify relevant patterns by mining the literature and then to b) validate their presence and prevalence in actual code repositories using active learning. To that end, we identify 14 AI pattern classes by mining 44 published AI pattern-related sources. Then we use an active learning approach to determine the prevalence of the most common pattern class across 100 GitHub open AI repositories. Using prevalence estimation, we propose bounds on the accuracy of the occurrences. The model achieves 56\% accuracy and 55\% recall in an 8-way classification task, significantly outperforming the 11\% random-chance baseline. Furthermore, the prevalence estimation offers usable bounds for analyzing pattern applications. This methodology provides a robust foundation to start understanding how AI patterns are used in practice, a field that currently lacks empirical data.
While recent advancements like the Poincaré ResNet have demonstrated the potential of learning visual representations directly in hyperbolic space, their optimisation remains hampered by the computationally intensive nature of Riemannian gradients and the strict boundaries of the manifold. Furthermore, standard hyperbolic networks treat spatial transformations of the same object as distinct hierarchical concepts, leading to redundant parameter usage and vanishing signals. We propose Equivariant Poincaré ResNets, combining hyperbolic geometry with discrete symmetry groups ($C_4$ and $D_4$). We identify critical roadblocks in applying Euclidean equivariance to hyperbolic space and propose geometrically safe tensor reshaping, left-regular permutations for hyperbolic group convolutions, and joint-orientation Poincaré Midpoint Batch normalisation. Empirically, embedding equivariance drastically reduces the optimisation space, accelerating convergence while accelerating convergence while respecting the boundary constraints of the Poincaré ball and preserving spatial-group equivariance.
Deep learning (DL) frameworks are critical AI infrastructures that often hide bugs with serious security implications. While dynamic approaches such as fuzzing are effective in uncovering these bugs, they require real test execution and incur high computational costs. Static analysis is a natural complement because it can detect bugs without runtime execution, offering fast and scalable testing. Unfortunately, there is still limited work targeting static analysis for DL frameworks due to their multilingual architectures and tensor-related program state. We present Phoenix, the first LLM-based static analysis technique for DL frameworks. Our key insight is that cross-language tensor flows in DL frameworks can be modeled, together with concrete code context, as a structured semantic bridge intermediate representation (SBIR) that LLMs can analyze for potential bugs in tensor semantic propagation. We implement this insight through a multi-agent workflow. A summarization agent first distills bug summaries from historical bug-fix patches and CWE rules. Guided by each summary, an extraction agent identifies bug-relevant repository symbols for code retrieval, and a generation agent synthesizes grounded SBIRs from the retrieved context. Finally, an analysis agent is leveraged to check SBIRs and report potential bugs. Our evaluation shows that Phoenix is a practical complement to dynamic DL framework testing for bug finding. To date, Phoenix has found 31 real new bugs in PyTorch for different heterogeneous hardware backends (Intel CPU, NVIDIA CUDA, and Apple MPS). Among them, 20 submitted bug-fixing patches have been merged into upstream.
Lightweight machine learning models are increasingly proposed for intrusion detection in Industrial Internet of Things (IIoT) networks due to their suitability for resource-constrained edge deployment. Most reported results evaluate these models only within their training network, leaving behavior on unseen networks unverified. This study trains four lightweight architectures on one IIoT dataset and evaluates them, without retraining, on two structurally distinct IIoT datasets using a feature representation restricted to attributes available across all three sources. Explainability analysis across two top-performing models shows both rely overwhelmingly on coarse port-category features; the most influential category occurs in source-domain attack traffic at 96 to 435 times the rate in the two target domains, indicating that coarsening port resolution relocates rather than removes a documented shortcut. Evaluation under naturally imbalanced class distributions reveals a further effect: the evaluation protocol used can reverse which target network appears to pose the greater generalization challenge. Adversarial robustness and recovery through limited target-domain exposure are also assessed; robustness to adversarial perturbation is unrelated to cross-network generalization, and recovery through adaptation varies considerably by architecture. These findings suggest deployment readiness should be assessed using cross-network evaluation under realistic class distributions, rather than within-domain accuracy alone.
Existing egocentric benchmarks have primarily constructed the egocentric setting from first-person-view data, which makes it difficult to evaluate egocentric perspective itself in isolation. However, understanding first-person-view input and taking an egocentric perspective are separable abilities, especially when first-person body cues are absent or when other agents are present. To isolate egocentric perspective understanding, we introduce EgoGapBench, a diagnostic benchmark for measuring action selection in multi-agent egocentric scenes. We define the ability measured by this benchmark as Egocentric Action Selection (EAS): selecting an appropriate action from the agent's perspective in the presence of other agents. On EgoGapBench, humans answer reliably, whereas both open-source and proprietary MLLMs perform substantially worse and systematically select actions performed by other visible agents. Fine-tuning on existing egocentric data fails to close this gap and can even be detrimental. In contrast, fine-tuning on EgoGapBench training data improves accuracy but does not reach human performance. These results show that EAS is difficult to acquire from first-person-view data alone, and that MLLMs should be evaluated and trained not only for scene understanding but also for egocentric action selection.
Few-step flow-map generators, such as consistency models and MeanFlow, accelerate sampling by directly learning long-range transport maps between noise and data. However, these models are typically deterministic, which makes them difficult to optimize with reinforcement learning (RL) post-training methods that require stochastic trajectories and well-defined likelihood ratios. Existing SDE-based stochasticization techniques are designed for velocity-based samplers with infinitesimal or finely discretized transitions, and therefore do not directly apply to long-range flow maps. In this work, we propose Flow-Map GRPO, an online RL post-training framework for deterministic few-step flow-map generators. The key component is Anchored Stochastic Flow Map Composition (ASFMC), a path-preserving stochasticization mechanism that introduces randomness through anchor-based conditional resampling while preserving the original marginal probability path of the deterministic flow map. We derive GRPO objectives for both single-time and two-time flow-map parameterizations. Experiments on few-step FLUX-based text-to-image generators, including MeanFlow and sCM, show that Flow-Map GRPO improves pretrained deterministic flow-map models across reward-based, perceptual, and task-level evaluation metrics. Our results demonstrate that deterministic few-step flow-map generators can be effectively aligned with RL post-training without modifying their original model parameterization or retraining them as native stochastic models.
As AI takes on more software work, the line between human and AI effort is shifting. Where developers draw that line around AI autonomy bears on how we design tools and roles that preserve meaningful work. Drawing on cognitive appraisal theory, work design, and automation research, we conducted a mixed-methods study of 448 professional developers at Microsoft to investigate their accepted levels of AI autonomy across software engineering work. Most developers accepted AI producing work under their oversight, although accepted autonomy varied substantively across tasks and individuals. Acceptance was lowest for identity-defining, human-facing, and design-oriented work, and higher among developers with more AI experience and risk tolerance. Task accountability was associated with lower odds of allowing AI to act on developers' behalf, whereas task identity was associated with lower odds of granting AI decision-making autonomy. Task demands had the opposite effect, increasing willingness to delegate decision-making to AI. Our findings suggest that preferences for AI autonomy reflect how developers cognitively experience their work, highlighting important considerations for designing meaningful work.
Scientific reasoning is an increasingly important capability of large language models, yet improving the robustness and efficiency of training such reasoning remains a key open challenge. We study this problem in instruction-based molecular optimization, where answer-only supervised fine-tuning (SFT) collapses multi-step reasoning and reinforcement learning with verifiable rewards (RLVR) suffers from sparse feedback. Reference-guided Policy Optimization mitigates both by anchoring policy updates to dataset-provided references, but its effectiveness is tightly coupled to reference quality: weak or misaligned references impose a performance ceiling. To overcome this ceiling, we propose active reasoning, a paradigm in which the policy actively decides, on a per-instance basis, when to imitate a reference and when to reinforce its own discoveries, while continuously upgrading what it imitates. We instantiate this paradigm as Active Group Relative Policy Optimization (Active-GRPO), realized through two coupled mechanisms: active imitate-reinforce and active referencing. The former performs imitation learning when the reference still outperforms the policy's own candidates, and shifts to self-improvement via reinforcement learning once the policy has generated molecules that surpass the reference. The latter continuously upgrades the reference itself by replacing it with the best policy-generated candidate discovered so far, progressively raising the imitation target and ensuring that reference guidance remains informative-rather than restrictive-throughout training. Across TOMG-Bench MOLOPT, Active-GRPO improves average SRxSim from 0.0959 for GRPO and 0.1665 for RePO to 0.1773 under matched three-seed evaluation, with statistically significant gains on LogP, MR, and QED.
Improvements in the technical performance of human--robot interaction (HRI) systems do not automatically translate into differences that human users can detect during live interaction. This paper investigates whether a 15 percentage point gain in end-to-end task success (from 75% in a multimodal baseline system to 90% in an improved configuration identified through a prior ablation study) is sufficient to produce consistent and measurable differences in user perception. The baseline system combines Whisper for speech recognition, Florence-2 for open-vocabulary object detection, LLaMA 3.1 for action extraction, and an interval Type-2 fuzzy logic controller for motion execution. The improved configuration replaces the perception and language modules with Grounding DINO + SAM and Qwen 3.5 9B, respectively, while retaining the same controller. A within-subject user study with 24 participants compared both systems on the same tabletop object-grasping task. After interacting with each configuration, participants rated perceived speed, reliability, and overall competence and fluency on a 7-point Likert scale. Results show that 17 out of 24 participants (70.83%) preferred the improved system (exact binomial test, p = 0.043, h = 0.43), and all three perceptual constructs were rated significantly higher for the improved configuration after Holm correction, with large to very large effect sizes (p < 0.001). These findings confirm that the identified technical improvements are perceptible to users in direct interaction and underscore the importance of complementing benchmark evaluation with user-centred evidence when assessing robotic manipulation pipelines.
Generative AI now enables games to produce dialogue, quests, characters, images, and worlds at runtime. Yet generation alone does not make a game AI-native, nor does it guarantee playability. This paper defines AI-native games by whether runtime generative AI is constitutive of the core loop: if the AI component were removed or trivially replaced, the central form of play would collapse or become fundamentally different. This counterfactual criterion separates AI-native games from AI-augmented games, boundary artifacts, chatbots, tavern-style role-play, procedural content generation, and AI-assisted production. Using this definition, we screen candidate artifacts and analyze 53 publicly available AI-native games and prototypes. We introduce a dual-axis G/N taxonomy: the G-axis captures player-facing game type, while the N-axis captures the dominant AI mechanic that makes generative AI indispensable to play. The corpus is concentrated around language-forward designs, especially narrative adventure, epistemic interaction, and generative narrative, while categories such as semantic adjudication, multi-agent simulation, generative construction, and relationship/companion play remain less represented. We argue that the central design problem is organizing semantic openness into stable gameplay. AI-native design depends on mechanical invariants: goals, rules, state, feedback, pacing, and player agency that make open-ended AI outputs interpretable and consequential. We conclude with a roadmap for controllable generation, AI-as-mechanic design, multimodal and multi-agent systems, inference economics, evaluation, safety, and regulation.
Artificial intelligence (AI) is becoming ubiquitous, and across domains, increasingly autonomous systems are carrying out tasks which raise significant ethical and legal challenges which demonstrate a need for strong human-machine teams rooted in trust. In this article, I argue that within highly impactful areas (such as medicine or warfighting) there are grounds for us initially treating autonomous and opaque systems as relevantly analogous to dogs (or other animals with which we have close relationships). Under this analogy, humans making use of these systems are not to be viewed as "users" or "deployers" of these systems, but instead take the role of "handlers". This recasting of roles shifts the way we view humans, AI-enabled and autonomous systems, and the relations between them, and moreover clarifies the clear and traceable lines of responsibility humans have for the outcomes brought about when using these systems. In developing this point, I clarify that the machine-animal analogy does admit disanalogous elements, but that its touch-points ground it as a starting point. I then explore how we can divest the humans-as-handlers approach of those aspects of our relationships with animals which are unfitting for how we engage with and make use of autonomous and AI-enabled systems. I conclude by arguing that the trajectory of human-machine teamings for autonomous and AI-enabled systems should be a state where we authentically view these not as artifacts which we simply make use of, but as collaborators with which we pursue complex goals and carry out complex tasks.
Empirical quantum-software papers often report that one compiler, optimizer, backend, or ansatz outperforms another. Such comparisons are not properties of a tool alone: they can change with benchmark scope, circuit construction, compilation, sampling, backend or noise assumptions, optimizer choices, and resource budgets. Existing testing, benchmarking, and reproducibility methods help assess programs, tools, executions, and platforms, but they do not directly audit whether the reported comparison itself is supported by the evidence exposed in the source paper or accompanying materials. We present CLAIMSTAB-QC, a source-bounded framework for auditing empirical comparisons in quantum software. Given a reported comparison, the framework records the baselines, metric, relation, and admissible evidence; locks the comparison design before outcomes are computed; and reports either a scoped relation outcome or an explicit evidence boundary. For strict scalar-directional comparisons, the reported direction is classified as Sustained, Unresolved, or Reversed within the locked audit scope. We evaluate CLAIMSTAB-QC on 455 comparative claims from 119 quantum-software papers. The central finding is a materialization gap: 175 claims can be represented for audit planning, 79 become scalar-directional planning records, 53 yield lockable audit or diagnostic designs, and only 8 expose enough matched evidence to audit the original comparison without proxy reconstruction. These 8 records yield 2 Sustained, 4 Unresolved, and 2 Reversed outcomes. Controlled diagnostics over 24 benchmark-relevant comparisons further show that simpler checks can preserve apparent directions whose support weakens under locked audit designs.
Automatic understanding of dynamic 4D point clouds, the 3D-point sequences captured over time by depth sensors and LiDAR, is central to robotics and embodied perception. Yet annotating them densely is expensive, making self-supervised pretraining the natural route to transferable representations. Existing pretext tasks, however, are almost entirely intra-modal, and the few methods that transfer knowledge from 2D foundation models rely on a single global embedding per clip, discarding the rich per-patch semantics that these models compute. To address this gap, we propose Cross4D-JEPA, a teacher-student method that distills a frozen 2D foundation model, an image model DINOv2, or a video model V-JEPA 2, into a 4D point encoder. The proposed method combines (1) a dense cross-modal correspondence that maps every 3D point to the teacher patch feature it projects to, and (2) a per-point objective that trains the student to match these features in latent space with no masking, negatives, or decoder. We evaluate Cross4D-JEPA on four benchmarks, MSR-Action3D, DeformingThings4D, NTU-RGB+D 60, and HOI4D, against intra-modal and global cross-modal baselines. Experimental results show that, under a matched protocol, the proposed method consistently outperforms intra-modal and global cross-modal baselines across the four benchmarks and is competitive with heavier published 4D methods; further analysis attributes this gain primarily to the granularity of the correspondence rather than the teacher modality. Beyond recognition accuracy, the dense representation learned by Cross4D-JEPA transfers across domains, improves label efficiency, and improves full-label fine-tuning under the same training budget, while a 13x smaller encoder matches a heavyweight pooling backbone.
Structural equation modelling (SEM) is widely used in survey-based business and information systems research to assess latent constructs and theory-driven structural relationships. However, SEM path significance is obtained within a particular model specification and may not show whether findings remain stable under alternative estimation frameworks. This study develops and demonstrates a staged robustness analysis framework that connects SEM, ordinary least squares (OLS) regression, and Double Machine Learning (DML). SEM is first used to refine the measurement structure and estimate the robustness-baseline SEM model, in which the full theory-specified structural path system is retained for downstream robustness analysis before final structural path evaluation. OLS regression is then applied to SEM-derived construct scores as a transparent regression benchmark. Finally, DML-style residualisation is used to examine whether each tested focal relationship remains stable after flexible machine-learning-based adjustment for observed controls. Learner-sensitivity checks compare Random Forest, Gradient Boosting, and Support Vector Machine learners, and selected reverse-direction diagnostics are used to examine directional sensitivity. The framework is demonstrated using a FinTech Digital Customer Intimacy survey model. The findings identify which relationships are stable across SEM, OLS, and DML-style checks, and which require more cautious interpretation. A reproducible Google Colab workbook and generated result files are publicly available, providing a reusable template that researchers and students can adapt to other survey-based latent-construct studies. The paper contributes a practical robustness workflow and interpretation guide for survey-based researchers seeking to complement SEM with conventional and machine-learning-based robustness checks.
Mutation testing is a powerful technique for ensuring software quality. However, the presence of equivalent mutants introduces unnecessary costs and biases, limiting its practical effectiveness. Although numerous equivalent mutant detection (EMD) methods have been proposed, they often face distinct challenges: pure-code analysis methods can be limited by their reliance on specific compiler infrastructures, while existing machine-learning approaches remain constrained by scarce training data and limited generalization to unseen mutants. Large language models (LLMs) have recently demonstrated remarkable performance across diverse code-related tasks by better capturing program semantics. Yet their potential for EMD remains largely unexplored, particularly in the multi-lingual context. This paper presents the first comprehensive empirical study on LLMs for EMD, using 3,302 Java and 1,088 C mutant pairs to benchmark against state-of-the-art methods, explore strategy variations, assess efficiency, and evaluate cross-lingual generalization. Experimental results show that LLM-based approaches achieve higher F1-scores than the evaluated traditional methods, with fine-tuned code embedding yielding the highest detection accuracy among the tested strategies. Moreover, LLM-based approaches strike a practical balance between effectiveness and efficiency with inference times comparable to existing machine-learning models. Importantly, fine-tuned LLMs demonstrate measurable generalization across programming languages. These findings establish LLMs as a viable and efficient approach for tackling the longstanding challenge of equivalent mutant detection, offering new directions for advancing mutation testing in practice.
Knowing which training examples drive outputs is fundamental to auditing, correcting, and understanding language models, yet for modern LLMs this remains expensive, approximate, and largely post-hoc. Standard language models generate tokens through a dense network pathway, causing training data's influence to be distributed across parameters rather than organized along explicit, traceable components. We introduce a prototype language model architecture, Prototypes for Interpretable Sequence Modeling (PRISM), that forms each prediction via a sparse, non-negative mixture of learned prototypes, trained with clustering objectives that anchor each prototype to coherent neighborhoods of training examples. Across architectures from 130M to 1.6B parameters trained on up to 50B tokens, prototype language models either surpass or remain within 2.5 percentage points on average downstream accuracy of matched dense baselines. We show that sparse prototype structure localizes curvature in the loss landscape, yielding a more tractable Hessian and enabling training data attribution that is ~500x faster than post hoc baselines when consuming equivalent memory. Calibrating linear prototype controllers can improve downstream accuracy by roughly 3 points while tracing those corrections back to training neighborhoods, and targeted prototype suppression can remove model behaviors without finetuning or measurable loss in generation quality.
While long-horizon mobile GUI agents typically rely on thought-action-observation loops, they struggle to separate persistent task states from transient screen observations. As execution histories grow, this entanglement imposes a severe context burden, causing agents to forget initial requirements, hallucinate progress, or repeatedly interact with stale interfaces. To address this, we introduce Task-State Representation (TSR), a training-free framework that explicitly decouples task state from sensory input. Acting as a lightweight external wrapper, TSR maintains three structured components: a global instruction summary, a dynamic progress tracker for subgoals, and a transition-aware action verifier. By continuously updating through pre- and post-action visual comparisons, TSR effectively guides the agent's reasoning without requiring architectural modifications. Experiments across four mobile GUI benchmarks validate TSR's effectiveness, yielding up to a 12 absolute point increase in success rate on complex cross-application and memory-intensive tasks.
We present BaseRT, a native Metal inference runtime for large language models (LLMs) on Apple Silicon, and report the highest inference throughput on this hardware to date. Existing runtimes, including llama.cpp and MLX-based frameworks, incur overhead from abstractions not designed for Metal's execution model or Apple Silicon's unified memory topology. By building natively on Metal with chip-specific kernel fusion, unified memory-aware optimisation, and custom dispatch logic, BaseRT recovers performance that framework-based approaches leave on the table. BaseRT supports a wide range of model families across eight quantisation formats (Q2 to FP16) on all Apple M-series devices. In this paper, we evaluate the Qwen3, Llama 3.2, and Gemma 4 families at Q4 and Q8 quantisation on M3 and M4 Pro devices. BaseRT achieves up to 1.56x higher decode throughput than llama.cpp and up to 1.35x higher than MLX, with substantially larger margins on prefill for mixture-of-experts models, delivering consistent best-in-class throughput from sub-1B to 30B parameter models. These results establish Apple Silicon as a more capable inference platform than previously reported, with direct implications for the emerging edge inference paradigm: as privacy requirements, latency constraints, and cloud cost pressures drive inference toward on-device deployment, performance-optimised local runtimes are a critical enabling layer for this transition. BaseRT is publicly available at https://github.com/basecompute/baseRT
Benchmarks for vision-language models (VLMs) mostly test observational spatial reasoning: models describe relations already visible in the input. Existing what-if tasks typically vary the observer while keeping the scene fixed. Can VLMs instead predict the consequences of hypothetically moving or rotating an object? We introduce MindEdit-Bench, a benchmark of six spatial reasoning tasks built from three-photo smartphone triplets of newly captured indoor scenes via an automatic in-the-wild 3D scene-graph extraction pipeline. Four tasks probe perception and perspective transformation over observed structure; two new tasks, L4 (spatial editing) and L5 (cross-view visibility editing), probe object-level counterfactual reasoning, where correct answers are absent from all input images. Each question provides 8-24 structured answer choices, enabling answer-letter-level diagnosis of spatial and fallback errors. The benchmark covers 120 private indoor scenes not drawn from public datasets, reducing public-data pretraining-overlap risk. Across 15 VLMs on 1,003 human-verified questions, task-wise mean VLM accuracy is only 8%-31%, versus 81%-97% human majority-vote accuracy. The pooled human--best-VLM gap is 53 pp, with at least 39 pp on every task. The structured answer space further reveals non-uniform failures, including weaker camera-depth-axis inference and fallback behavior on difficult visibility-editing cases.
Diffusion models are highly effective at modeling complex data distributions, including images and text. However, in applications like personalized recommender systems, the objective often shifts to modeling specific regions of the distribution that maximize user preferences-initially unknown but gradually uncovered through interactive feedback. This can naturally be framed as a reinforcement learning problem, where the goal is to fine-tune a diffusion model to maximize a reward function based on preferences. However, the main challenge lies in learning a parameterized reward model, which typically requires large-scale preference data-something that is often not feasible in practice. In this work, we introduce Personalized Active Preference Alignment PAPA, a novel method that bypasses the requirement for a parametrized reward model by directly optimizing the diffusion model using real-time user feedback. PAPA enables feedback-efficient preference alignment, drawing inspiration from the variational inference framework. We demonstrate PAPA's effectiveness through extensive experiments and ablation studies across diverse class-conditioned and fine-grained alignment tasks. Additionally, based on theoretical insights, we propose an enhanced fine-tuning strategy, referred to as EPAPA, that requires less computational budget and accelerates the fine-tuning process, further boosting PAPA's suitability for real-world deployment. Our code is made publicly available at https://github.com/NasikNafi/papa.
Large reasoning models (LRMs) have achieved strong reasoning capabilities in English, yet their performance degrades significantly when required to reason in other languages. A natural solution is to transfer the model's English reasoning ability to target languages. However, existing transfer approaches typically rely on distilled target-language reasoning traces from stronger LRMs or online supervision from external judge models, which are costly and difficult to scale. In this paper, we propose PCS (Progressive Code-Switching), a more efficient transfer framework that requires only lightweight translation without any stronger model for distillation or judging. PCS first constructs code-switched reasoning traces by translating a subset of English reasoning steps into the target language, and uses them to initialize the model's code-switching ability via supervised fine-tuning. It then applies reinforcement learning with a step-level language consistency curriculum, progressively raising the target-language ratio until the model reasons entirely in the target language. This progressive design provides a smooth transfer path that avoids the instability and performance degradation commonly observed when directly enforcing target-language reasoning. Experiments on multiple benchmarks and five typologically diverse languages show that PCS substantially narrows the performance gap between target-language and English reasoning, yielding more language-consistent reasoning while maintaining competitive accuracy.
Reasoning language models frequently overthink: generating extended chains of behaviors such as hedging, approach abandonment, and self contradiction that consume tokens without improving answers. We show that these behaviors are not merely a consequence of length; even when controlling for response length, incorrect traces exhibit higher rates of unproductive self-reflection than correct ones. Addressing this requires identifying where self-reflection helps vs hurts, but obtaining these step-level annotations is costly. We observe that intermediate answer commitments within reasoning traces can provide a cheap proxy: by comparing each final answer candidate in the trace to the ground truth, we can determine whether subsequent reflection is productive without any additional supervision. Building on this insight, we propose DASH (Drift Aware advantage SHaping), which assigns segment-level credit based on whether each reasoning segment leads toward or away from correctness. On competition-level math benchmarks, DASH achieves the highest accuracy where overthinking is prevalent (AIME25: 50.8% vs. 45.4% GRPO) while reducing overthinking behaviors and achieving more productive self-correction than baselines.
Jailbreak attacks remain a critical threat to the safe deployment of large language models (LLMs). While prior work has primarily studied attacks and defenses at the prompt level, we show that this prompt-centric paradigm overlooks a structural vulnerability in stateful, function-calling environments. In such applications, developer-defined schemas, structured arguments, and untrusted tool outputs are interleaved into a single shared model context. This architecture expands the attack surface by blurring the boundary between trusted control logic and untrusted data, allowing adversarial intent to be distributed across a multi-turn execution path. We exploit this architectural flaw through SMT, a black-box attack framework based on Simulated Moderation Traces. Departing from purely prompt-based interactions, SMT constructs a multi-turn trajectory that simulates a legitimate moderation-auditing workflow. Within this trajectory, a fabricated moderation frame leverages red-team testing as a pretext to elicit harmful generations. The subsequent validation feedback treats safety refusals as execution failures, prompting refinements that gradually weaken the model's safety constraints and ultimately trigger harmful outputs. Extensive empirical evaluations on prominent commercial LLMs from five different providers across two standardized safety benchmarks show that SMT consistently achieves the highest average attack success rate and HarmScore while requiring a near-minimal number of queries, substantially outperforming existing baselines. These findings demonstrate that prompt-level sanitization alone is fundamentally insufficient for defending tool-enabled LLM systems and highlight the urgent need for context-aware validation across schemas, arguments, tool outputs, and accumulated conversation state. The code is available at https://github.com/liujlong27/SMT.
Transformer-based large models have demonstrated remarkable generalization abilities across different tasks by leveraging a context-aware attention module for in-context learning. With richer context, transformers adapt more effectively to the current use case without any parameter updates. However, the quadratic computational and memory complexity with respect to context length significantly slows data processing in softmax transformers. Linear transformers were proposed to address this issue by reducing the complexity to linear dependence on context length, but the design and understanding of the feature mapping in linear attention, from a theoretical viewpoint, remain unclear. In this paper, we investigate the approximation and generalization abilities of linear transformers under a two-staged sampling process from domain generalization. We show that linear transformers perform in-context learning as learning a mapping from context distributions to response functions. A dimension-independent convergence rate is obtained for our generalization analysis, which also exhibits the tradeoff between the regularities of data distributions and latent features. Guided by our theoretical framework, we propose a new perspective on activation and loss design for linearizing pretrained softmax large language models.
A total of seven categorical encoding methods were tested on the IEEE-CIS fraud benchmark dataset (590,540 records, 3.5% positives, 8 high-cardinality columns). The encoders were evaluated using a stratified 5-fold cross-validation (CV) with three repetitions. Five of the encoders had identical frozen LightGBM learners in the downstream phase, allowing for controlled comparisons of their performance to each other. CatBoost and TabNet were included as comparisons across paradigms using different learners. The entity embeddings produced the highest AUC-ROC (0.9612), with a statistically significant tie with that of CatBoost (0.9602) and statistically superior to tier group encoding (0.9548), whereas target encoding was only 0.0023 worse than tier group encoding and the auditor-friendly tier boundaries were maintained. Off-the-shelf TabNet did not outperform tree-based pipelines and collapsed under data scarcity. On AUC-PR, CatBoost leads (0.822 vs. 0.793); no encoder dominated both metrics. Per-column analysis confirmed the embedding advantage arises from joint multi-column representation.
How many days of early behavior suffice for subscription churn prediction? In the public KKBox dataset, the early indicator of churn is typically an indicator of someone's contract status; however, when looking in the heavily churned manual-renewal segment, having access to early behavior creates a substantial increase in prediction for that specific segment (PR +0.10 at 120 days). A nine-window sufficiency curve shows a diminishing-returns knee in a 45-90 day band. However, stress-testing over three cohort/task designs shows that this curve is singular to the design being tested; for example, in our test with a moving target, the curve inverts and can shift depending on the feature set used. Therefore, any window-sufficiency claim should state its cohort construction, target definition, and feature families. All evidence is from one music-streaming dataset; the mechanism should generalize but the magnitudes may not.
Cardiovascular disease is still one of the main causes of death around the world. Acute myocardial infarction (MI), or heart attack, claims millions of lives each year. MI happens when blood flow to the coronary arteries is blocked or reduced, which causes permanent damage to the heart muscle. Without treatment, this can lead to cardiac arrest, where the heart stops pumping blood to the organs, resulting in organ failure and death. Even survivors often face serious problems like heart failure, pulmonary edema, and asystole. Research shows that 5 to 10 percent of survivors die within the first year after an MI, and nearly half need to be hospitalized again. Early thrombolytic treatment leads to better outcomes, so there is a clear need for faster and more accurate ways to diagnose MI. Right now, doctors usually review patient history and use their own experience to find the causes of MI. This process takes a lot of time and can be inconsistent. Detecting MI accurately and quickly can help patients take better care of themselves and prevent fatal events. In this study, we introduce an automated model to predict deadly outcomes of MI and help doctors understand important biomarkers linked to its complications. This approach aims to make diagnosis clearer, faster, and more affordable. The process includes preparing the data, filling in missing values, and handling imbalanced data using SVMSMOTE, ADASYN, and class-weighted methods. We use wrapper and embedded feature selection to find the most important variables, then scale the features for consistency. The model combines Logistic Regression, Random Forest, Light-GBM, and Bagging SVM, and is further improved with an artificial neural network to increase accuracy. We evaluate all models using precision, recall, and other key measures to find the best option for clinical use.
We investigate a forecasting framework based on a simple discrete-time dynamic model with coefficients varying in time. The parameters of the model are recovered within a deep learning framework, which makes it possible to retain a transparent parametric structure while simultaneously accounting for complex and nonstationary patterns in the observed phenomenon. Our analysis covers two specifications of the noise process. Besides the standard Gaussian setting, we also consider Laplace-distributed noise, which can offer a more adequate description in the presence of heavier tails and sharper local fluctuations. For both cases, we formulate the predictive scheme of the model and analyze the associated uncertainty quantification, including the construction of prediction intervals. The results illustrate that a relatively simple model, when combined with time-dependent parameter estimation, can serve as a mathematically tractable and practically flexible tool for forecasting complex dynamics under different noise assumptions. The general model is stated for TVAR($p$), while the prediction-interval formulas and the numerical experiments are developed for the TVAR(1) case.
Large Vision-Language Models (LVLMs) rely extensively on Visual Instruction Tuning (VIT) to elicit their multimodal reasoning capabilities. However, we find a discrepancy: VIT often packs multiple language tasks about the same image for conversational, multi-turn training, whereas existing benchmarks evaluate LVLMs in isolated, single-turn scenarios. The models can suffer from visual attention decay and contextual overfitting during multi-turn training, making it hard for them to realize their full potential in the mismatched test phase. To close the gap, we propose learning with Stochastic Turn Depth (StochasT), which stochastically groups language tasks for the same image into clusters of varying sizes (turn depth) while preserving their organic order. Hence, while StochasT draws on Dropout and stochastic depth for ResNets, it does not actually drop anything to maximize the utility of the training data. Furthermore, we introduce a challenging, benchmark-agnostic evaluation mechanism based on the Balanced Latin Square to measure LVLMs' robustness under varying contextual dependencies. Extensive experiments demonstrate that StochasT effectively grants LVLMs strong, harmonized capabilities for both single-turn and multi-turn use cases.
Current molecular generation benchmarks emphasize task complexity, molecule novelty, and property alignment; they largely overlook a critical concern: the potential safety risks of AI-generated molecules. In practice, many generative models may produce molecules with toxic, reactive, or otherwise hazardous characteristics - posing hidden dangers that remain insufficiently addressed. To address this gap, we introduce MolSafeEval, a benchmark dedicated to evaluating and analyzing the safety risks of molecular generation. Unlike prior approaches that rely on narrow toxicity predictors, MolSafeEval integrates heterogeneous safety knowledge - ranging from toxicological databases to hazard rules - into a structured molecular safety knowledge graph. This graph serves as a foundation for large language model-based reasoning, enabling systematic detection and explanation of unsafe features in generated compounds. We further categorize molecular generative models into four representative task types - unconditional generation, property optimization, target protein-based design, and text-based generation - and provide standardized datasets and safety evaluation protocols for each. By systematically revealing the safety vulnerabilities of current generative approaches, MolSafeEval offers a new lens for benchmarking molecular models and provides essential guidance toward safer, more trustworthy molecular design.
Predicting complex spatiotemporal dynamics in physical processes often demands computationally expensive numerical methods or data-driven neural networks that suffer from high training costs, error accumulation, and limited generalizability to unseen parameters. An effective approach to address these challenges is leveraging physics priors in training neural networks, known as physics-informed deep learning (PiDL). In this work, we introduce the Multi-Resolution Finite-Volume-inspired network, MuRFiV, designed to capitalize on the conservative property of finite volume on the global scale and the expressive power of deep learning on the local scale. We demonstrate the effectiveness of MuRFiV on several spatio-temporal systems governed by partial differential equations (PDEs), including Burgers' equation, shallow water equations, and incompressible Navier-Stokes equations. By embedding PDE information into the deep learning architecture, MuRFiV achieves strong long-term prediction accuracy and remains stable over very long autoregressive rollouts, significantly outperforming data-driven neural network baselines. This result highlights the promise of combining multiresolution learning with finite-volume-inspired inductive bias for accurate and robust long-term prediction of complex dynamics.
Embodied agents operating in the real world require multi-scale reasoning and knowledge adaptation as conditions change. We identify two challenges in applying Mixture of Experts (MoE) to this setting: routing lacks an explicit notion of scale, preventing targeted updates at specific scales, and a uniform update policy cannot accommodate the different rates at which knowledge at each scale becomes outdated. We present MuSix, a framework that addresses both challenges through scale-aware world model mixture and evolution. A two-stage routing mechanism grounds scale selection in experiential distance, a measure of situational novelty inspired by Construal Level Theory: a meta-router first maps this quantity to a weight over continuous scale space, then per-scale base routers select world models within the identified scale. For adaptation, scale-dependent forgetting rates allow low-scale knowledge to refresh rapidly while high-scale abstractions persist, and gated inter-scale transfer maintains coherence across the hierarchy. Experiments on EmbodiedBench and HAZARD show that MuSix improves over state-of-the-art baselines on multi-scale reasoning and dynamic adaptation.
Agricultural advisory systems face a fundamental tension: static agronomic guidelines offer consistent, evidence-based recommendations, yet remain blind to in-season variability and dynamic uncertainties. Recent advisory systems powered by LLMs are liable for a different risk of generating recommendations that are agronomically credible but physiologically unconvincing. Agri-SAGE is a closed-loop framework designed to resolve the above two limitations by integrating retrieval-grounded multi-agent LLM reasoning with APSIM-based biophysical simulation, to generate and validate agronomic advisories. To assess this framework, we evaluate three reasoning approaches, namely Plan-and-Solve, Tree of Thoughts, and Reflexion, over a 10-year retrospective analysis. All three significantly outperform static PoP (Package-of-Practice) baselines, with Tree of Thoughts achieving impressive peak yields. At the same time, Reflexion achieves comparable agronomic outcomes at substantially lower computational cost by leveraging cross-seasonal episodic memory.
Actor-critic methods depend on learned critics, but critic quality is often evaluated only indirectly through return, temporal-difference error, or value loss. Critic complexity is introduced as an additional diagnostic and intervention dimension for actor-critic reinforcement learning. The analysis uses spectral effective-rank entropy, a rank-like summary of the singular-value distributions of critic weight matrices, to assess critic model complexity. Across TD3 and PPO experiments, critic complexity is tracked together with return and Monte Carlo value-estimation bias. The results show that critic complexity is measurable throughout training and is systematically associated with training behavior, while also making clear that the relationship is heterogeneous across algorithms, tasks, and hyperparameters. A direct complexity-control intervention is then evaluated by adding a spectral-entropy penalty to the critic loss. This intervention reliably changes the targeted spectral quantity, demonstrating that critic complexity can be controlled rather than only observed. Return effects are treated as task-dependent evidence rather than as a general performance claim, because overall complexity-control results vary.
The two-tower model has been widely used for large-scale recommendation systems, particularly in the retrieval stage. Industry standards for training two-tower models typically involve in-batch and/or out-of-batch negative sampling. However, these methods often produce easy negatives that models can quickly learn, failing to sufficiently challenge the model. To address this issue, a novel self-supervised hard negative sampling technique is proposed that leverages a large language model (LLM) to generate hard negatives from the same cluster during model training. By utilizing the LLM to learn media representations, the proposed approach ensures that the generated negatives are more challenging and informative. This real-time sampling framework is designed for seamless integration into production models, capable of handling billions of training data points with minimal computational complexity. Experiments on public datasets, along with deployment to a large-scale online system, demonstrate that the proposed negative sampling technique outperforms widely used industry methods. Furthermore, analysis in industrial applications reveals that this sampling method can help break inherent feedback loops in recommendations and significantly reduce popularity bias.
Large language models often produce hallucinated answers that violate prompt-level constraints. A key diagnostic question is whether these failures reflect missing knowledge, or whether the model has the relevant information but follows the wrong inference path. We study this phenomenon as inference misalignment: a mismatch between the answer supported by the prompt and the answer favored by statistically salient latent associations. We formalize this view with a latent key-task model, in which pretraining-frequency imbalance can cause a shortcut path to dominate the constraint-sensitive path and induce positive inference loss. The framework predicts two failure modes: task-retrieval bias in entity disambiguation and key-selection bias in action choice. We introduce TrapQA, a controlled diagnostic testbed with two components. ScientistQA tests disambiguation among similar scientists with supplementary factual probes, while Real-Life Constrained QA tests everyday constraint following under salient shortcuts. Our results show that hallucination can arise from biased latent inference rather than absent knowledge alone.
As video corpora continue to expand in both scale and task complexity, there is increasing demand for approaches that retrieve relevant videos from large-scale corpora (inter-video reasoning) and subsequently perform fine-grained, query-conditioned tasks (intra-video reasoning) within the retrieved content, such as temporal grounding. However, existing approaches typically treat retrieval as a preprocessing step, and consequently, when the initial retrieval fails, there is no mechanism to refine the search, leading to the failure of subsequent fine-grained intra-video reasoning. Moreover, while recent agentic frameworks have advanced video understanding, they typically assume that the query-relevant video is already given, focusing exclusively on intra-video reasoning tasks. To address these limitations, we propose VideoSearch-R1, an agentic framework for iterative video retrieval and reasoning through multi-turn interaction with a video search engine. Specifically, we introduce Soft Query Refinement (SQR) to refine search query tokens in a continuous latent space rather than rewriting queries in the discrete text space, enabling more efficient and fine-grained adjustments. SQR and its reasoning process are trained using Group Relative Policy Optimization (GRPO), guided by task-level reward signals derived from retrieval and downstream tasks. Building upon this, VideoSearch-R1 achieves state-of-the-art performance across three datasets on Video Corpus Moment Retrieval (VCMR), iteratively retrieving videos from large-scale corpora, refining search queries, and performing precise query-conditioned temporal grounding within the retrieved content. Our analyses show that SQR effectively refines the original query, requiring significantly fewer generated tokens than explicit text-level query refinement. Code and model checkpoints are publicly available at mlvlab.github.io/VideoSearch-R1.
Spatiotemporal motion planning, especially in multi-robot settings, requires robots to reason about collision-free regions that change over time, which is challenging in continuous spaces when feasible regions are transient and geometrically constrained. We present an algorithmic framework based on graphs of space-time convex sets (ST-GCSs), where collision-free regions are represented as convex sets in space-time and trajectories correspond to paths on the graph together with continuous motions within the selected sets. We formulate time-optimal planning on ST-GCSs as a graph-search problem over path-indexed states and develop a best-first search solver that evaluates partial paths via continuous trajectory optimization, guided by admissible heuristics and dominance checks. We further present an Exact Convex Decomposition (ECD) scheme to reserve trajectory occupancies in space-time, enabling unified handling of dynamic obstacles and multi-robot interactions. For multi-robot motion planning, we integrate ST-GCS planning and ECD into prioritized planning methods and introduce a windowed coordination scheme to improve efficiency. Extensive experiments on single-robot and multi-robot problems demonstrate substantial speedups over various planners while maintaining high solution quality, particularly in environments with narrow and transient feasible regions. Large-scale demonstrations further show that the proposed multi-robot motion planner can solve instances with up to $100$ robots within only a few minutes. Project homepage: https://sites.google.com/view/stgcs
Reinforcement learning (RL) for quadruped locomotion commonly depends on fixed, hand-crafted, and Markovian reward functions that limit both interpretability of learned policies and lack explicit control over gait behaviors. We introduce a framework where distinct gaits are specified using parameterized constraints expressed in Signal Temporal Logic (STL). These include safety bounds, gait synchronization constraints, command tracking, and actuation bounds. From these specifications, we develop a reward shaping mechanism that provides learning agents a dense, continuous reward landscape that encodes desired behavior. We define parametric STL templates for three speed regimes (walking-trot, trot, bound), calibrate their parameters from reference rollouts, and compute rewards from using smooth approximations of STL robustness over the rollouts. The generated rewards can be used to provide shaped gradients compatible with Proximal Policy Optimization (PPO). We instantiate the approach on Google's Barkour quadruped robot in MuJoCo XLA (MJX). We use parallelization within the simulator to improve training speeds and use domain randomization to robustify learned policies. We show that compared to a baseline of hand-crafted rewards, the STL-shaped rewards yield tighter velocity tracking and more stable training. Videos can be found on our project website: https://stl-locomotion.github.io/.
Large language model agents are increasingly connected to scientific software, yet it remains unclear when tool access makes scientific computation more reliable rather than merely more complex. We introduce PHREEQC-MCQ-200, a benchmark for evaluating tool-augmented agents on deterministic aqueous-geochemistry simulations. The benchmark contains 200 multiple-choice questions derived from 21 validated PHREEQC scenarios, requiring agents to construct simulator inputs, execute PHREEQC, inspect structured outputs, and commit to final answers. Across multiple frontier and mid-tier model families, simulator access substantially improves aggregate accuracy, confirming that grounded execution is necessary for many scientific-computation tasks. However, the gains are not monotonic: tool-augmented agents also lose items they answered correctly without tools, revealing regressions that average accuracy alone hides. We further show that output-access protocol matters. A table-of-contents interface can reduce token cost while preserving or improving accuracy for stronger models, but it degrades performance for mid-tier models that cannot reliably navigate structured simulator outputs. PHREEQC-MCQ-200 therefore frames scientific tool use as an end-to-end diagnostic problem rather than a simple tool-calling capability. We argue that evaluations of scientific agents should report not only accuracy, but also item-level retention, output-access sensitivity, trajectory failures, and where the computation chain breaks.
Social coding platforms such as GitHub host millions of repositories, yet many suffer from high mortality rates. Despite this, several survival factors remain poorly understood. Human capital is widely recognized as essential. Social attention, while often assumed to be a lifeline, can become a liability. Structural features that improve onboarding, such as code readability and documentation, may also accelerate the cessation of active development when combined with massive visibility. To examine these dynamics, we analyzed more than 73,000 GitHub repositories using an Accelerated Failure Time (AFT) survival framework, which accounts for the time-varying nature of predictors. Our study identifies human capital as the most critical determinant of project survival. In contrast, excessive social attention emerges as a liability, and when coupled with accessibility features, it amplifies the risk of project inactivity. Importantly, when the number of contributors interacts with social popularity, the protective effect of labor becomes visible, highlighting the need for governance strategies that balance visibility with labor capacity to ensure the long-term resilience of open-source projects.
Vision-language models (VLMs) have become a paradigm for multimodal learning, yet remain unstable due to object hallucination, weak visual grounding, and catastrophic forgetting after full-parameter instruction tuning. We claim these failures result from a lack of explicit control over visual representation learning during the standard next-token prediction objective. As a result, visual embeddings thus become passively optimized and prone to injecting redundant or spurious signals. To counter this, we introduce Information-Regularized Attention (IRA), a stochastic attention mechanism that explicitly regulates the amount of visual information injected into the hidden states of intermediate transformer layers. This local reparameterization translates uncertainty about visual representations into local noise that is independent across data points. Beyond evaluating model performance, we also quantify embedding properties, where IRA produces smoother curvature trajectories and suppresses attention-sink across all layers, indicating a more stable transformation of the visual signal. Our results suggest that stochastic attention is not merely a regularizer but a key contributor to representation learning in a generative architecture, offering a new direction for building more reliable VLMs.
Forecasting models for health-signal digital twins must preserve the oscillatory, frequency, phase, and state-transition dynamics of physiological signals, yet the pointwise metrics used to benchmark them cannot detect when these fundamental properties are lost. We show that this blind spot misranks models: across 11 architectures, models with comparable pointwise error diverge by up to 53° in phase accuracy, equivalent to roughly 123 ms for a 1.2 Hz cardiac rhythm and invisible to standard metrics. To enable development of models that escape such failures, we introduce TimeSynth, a controlled benchmarking framework with two reusable components: a physiologically grounded generator producing signals with analytically known ground-truth dynamics from parametric models fitted to real electroencephalography, electrocardiography and photoplethysmogram signals, along with diagnostics quantifying amplitude, frequency, phase, and state-transition fidelity. Linear and full-sequence attention models systematically lose frequency and phase information despite acceptable amplitude error, whereas architectures with localized temporal structure better preserve dynamical fidelity and adapt to observable state transitions; none, however, reliably preserves stochastic switching. Because the dominant determinant of fidelity is architectural, model choice becomes a principled, use-case-driven decision rather than a search for a single winner. TimeSynth thus supplies the controlled preclinical stress test missing before models are coupled to patient data, with a reusable generator and diagnostics for fidelity-aware development.
Large language models (LLMs) are increasingly applied to requirements engineering (RE) tasks, yet the prompts guiding them are typically designed manually through trial and error, yielding inconsistent and suboptimal results. Automated prompt construction remains largely unexplored in RE, leaving its effectiveness unclear. To address this, we propose a lightweight Automatic Prompt Engineering approach, Backtracking APE (BT-APE), and apply it to requirements classification. We frame prompt design as an optimization problem, iteratively refining prompts via LLM-generated candidates, backtracking search, and dynamic example selection. Evaluating BT-APE on three benchmark datasets with five instruction-tuned LLMs, we compare it against four classical prompting baselines (zero-shot, few-shot, chain-of-thought, CoT+few-shot) and a state-of-the-art but resource-intensive APE baseline (PE2). BT-APE and PE2 achieve nearly identical accuracy, both substantially outperforming the classical baselines with large effect sizes; however, BT-APE imposes a far lighter computational footprint, consuming roughly 72% fewer input tokens and 66% less wall-clock time at equivalent accuracy, making it better suited to resource-constrained deployment. Our contributions are threefold: (i) a lightweight APE framework with an open interactive tool and replication package; (ii) the first systematic comparison of APE against classical prompting for requirements classification; and (iii) insights into how class definitions and prompt evolution affect performance.
Vision language models (VLMs) demonstrate strong zero-shot performance, but often perpetuate social stereotypes in person-centric queries, yielding skewed demographic distributions. Current debiasing methods apply uniform bias corrections across all input queries regardless of their bias sensitivity, creating a fundamental fairness--utility trade-off. Strong debiasing distorts semantically meaningful information in bias-insensitive queries, while weak debiasing fails to mitigate stereotypes in bias-sensitive ones. This one-size-fits-all approach hampers simultaneously achieving high utility on bias-insensitive queries and fairness on bias-sensitive queries. We introduce Reward-Gated Test-Time Adaptation (RG-TTA), a reinforcement learning-based test-time adaptation framework that selectively applies debiasing based on input sensitivity. RG-TTA adaptively triggers fairness regularization based on the bias sensitivity of each input during test-time policy adaptation, while focusing exclusively on optimizing cross-modal alignment for bias-insensitive inputs. Experiments on fairness benchmarks (e.g., FairFace, UTKFace) demonstrate substantial bias reduction while simultaneously improving zero-shot utility, resolving the trade-off of uniform debiasing.
This paper presents Speech Playground, an interactive speech visualization and comparison tool. While existing tools such as Praat are excellent, it can be cumbersome to integrate them with modern deep learning representations and use them for comparison. Speech Playground addresses this by combining a Python backend with a web-based frontend for interactive exploration of multiple feature types, including continuous, discrete, and variable-length representations. It includes TextGrid and forced alignment support together with configurable distance and alignment settings for visual and auditory comparison. Speech Playground is intended for use in speech research, representation validation, and computer-aided pronunciation training (CAPT)-oriented experimentation.
In the era of satellite constellations, multi-view optical satellite imagery is pivotal for Earth Observation (EO) and high-quality Digital Surface Model (DSM) reconstruction. Although feed-forward 3D foundation models have transformed computer vision, their deployment in satellite remote sensing is inherently constrained by the structural discrepancy between implicit perspective assumptions and explicit orbital pushbroom geometry. This geometric incongruity is further compounded by pronounced view-set heterogeneity. We present EO-VGGT, a framework that adapts a frozen perspective-driven model to orbital observations via explicit physical geometry embedding.First, the Geometry-Correlation Constrained Selection (GCCS) strategy prunes sub-optimal observations by balancing geometric diversity and radiometric consistency to optimize the input sequence. Second, a Sensor-Ray Encoder (SRE) parameterizes pixel-level pushbroom lines of sight derived from the Rational Function Model (RFM) into high-dimensional space-geometric tokens, reconciling the mathematical discrepancy between central projection and orbital kinematics. Third, a lightweight Ray-Pointing-Aware Adapter (RPAA) employs gated residual blocks to integrate these tokens directly into the frozen transformer backbone. Our findings underscore that integrating explicit physical geometry with optimized view selection is essential for robust feed-forward satellite 3D reconstruction.
Authority bias poses a critical safety concern in language models: models systematically prioritize social cues from authority figures over factual consistency, swaying their answers based on source credibility rather than evidence. We mechanistically investigate this phenomenon using a controlled medical QA setting, where hints suggesting incorrect answers are attributed to personas of varying expertise. Across Llama-3.1-8B, Qwen3-8B, and Gemma-2-9B, we find that models respond in a graded manner proportional to perceived authority, a hierarchy that is never explicitly prompted but emerges from training. Logit lens analysis and linear/non-linear probing localize this effect to a critical late layer where correct answer representations are actively erased, an erasure that scales with authority level, resists mean vector intervention, and is only partially reversible through chain-of-thought reasoning. Our findings suggest that authority-induced sycophancy is not a surface-level output bias but mechanistic knowledge erasure, a precise, layer-localized overwriting of correct internal representations by high-status authority signals.
Recent brain decoding studies have made substantial progress in reconstructing externally perceived visual content from neural signals. However, using electroencephalography (EEG) recordings to guide facial expression editing remains largely unexplored and poses a distinct challenge: rather than recovering what a subject sees, it requires identifying facial-action related patterns from noisy EEG signals and grounding them in localized, identity-preserving expression edits. In this paper, we investigate EEG-conditioned facial image editing for fine-grained facial action unit (AU) control and propose MindAU, a unified framework for controlling facial AU edits from EEG signals. MindAU first learns noise-robust and AU-discriminative EEG representations through temporal masked reconstruction and AU classification supervision. It then bridges the modality gap via Dual-Stream Manifold Alignment, aligning EEG features with AU-level text semantics and identity-reduced visual displacement trajectories in the multimodal space of Qwen2.5-VL. Finally, MindAU incorporates EEG-aware Multimodal Rotary Positional Embeddings, landmark-guided reference masking, and AU-aware region supervision into a multimodal diffusion-based editor for high-fidelity identity-preserving editing. We also introduce E-CAFE, a curated benchmark for EEG-Conditioned Action-Unit Facial Editing with paired EEG-face editing samples and standardized evaluation protocols. Extensive experiments demonstrate the effectiveness of MindAU and suggest its potential as a step towards future assistive expression technologies for individuals with facial neuromuscular disorders.
Slide design requires personalizing both deck themes and page layouts. Yet, current AI agent-based methods struggle with fine-grained, page-level design. Solely relying on prespecified templates or user verbose instructions, they fail to capture latent design intents, leaving Page-level Slide Personalization (PSP) unresolved. To close this gap, this work formulates PSP as an inverse planning problem. We propose to learn a design intent without assuming any knowledge of the specific executing tools (e.g., PowerPoint, Beamer) being used. However, relinquishing control over these tools makes the problem intractable to optimize end-to-end. To overcome this, we propose SPIRE, a principled framework to solve PSP approximately. By intentionally corrupting the visual structures of clean slides, SPIRE creates a verifiable task to denoise the corruption, whereby two agents learn to collaboratively refine executable designs via reinforcement learning (RL). We present a proof that structural denoising is a consistent surrogate for PSP, and that the multi-agent formulation strictly reduces policy gradient variance in RL. Extensive experiments demonstrate the superiority of SPIRE.
Safety alignment of text-to-image (T2I) diffusion models aims to suppress harmful generations while preserving utility on benign prompts. Recent methods often appear to deliver high safety with high utility, but this conclusion rests largely on coarse global utility metrics (e.g., FID, CLIPScore) that are insensitive to fine-grained semantic correctness, creating an illusion of high utility. We show that when utility is measured with structured evaluation, this illusion breaks: on TIFA (Text-to-Image Faithfulness evaluation with Question Answering), safety-aligned models suffer substantial drops in semantic fidelity, including failures in object counts, attributes, and relationships. To diagnose the source of this gap, we analyze the text-encoder prompt embedding space and uncover semantic collapse, a contraction of embedding spread coupled with distortion of inter-prompt similarity structure, which strongly correlates with structured utility loss. Guided by this insight, we propose StructureAware Geometric Regularization (SAGE), a safety alignment objective that explicitly preserves embedding spread and inter-prompt relational structure during adaptation. Our method restores structured utility (TIFA +5.0% over prior state-of-the-art) while maintaining strong safety performance and competitive coarse-grained utility scores. Our source code and trained models are available at https://adeelyousaf.github.io/SAGE_ECCV26_Project_Page/.
Simulating two-dimensional frustrated quantum matter is a grand challenge due to the sign problem and exponential Hilbert space complexity. In this work, we introduce the Holographic Quantum Transformer (HQT), a physics-inspired generative architecture that leverages global self-attention to resolve non-local entanglement patterns. We validate HQT on the square lattice $J_1-J_2$ Heisenberg model. On the heavily frustrated $8 \times 8$ lattice at the quantum critical point ($J_2=0.5$), HQT reaches a ground-state energy per site ($E/N$) of $\mathbf{-0.5001(1)}$, consistent with the expected finite-size scaling trend. Beyond numerical accuracy, HQT exhibits intrinsic physical awareness, autonomously recovering the underlying $J_2$ interaction geometry through interpretable attention maps. Our central contribution is ``Holographic Transfer", a zero-shot size-extrapolation protocol with rapid alignment: a model trained on $8 \times 8$ systems is directly projected onto larger $10 \times 10$ lattices via continuous positional-embedding interpolation and head re-initialization, achieving high-fidelity initialization and rapid convergence. This zero-shot protocol yields an energy of $E/N = \mathbf{-0.49782(3)}$, statistically consistent with the variational state of the art while requiring no from-scratch training on the target lattice. Our results establish generative attention as a scalable paradigm for transferable quantum simulation.
Understanding how complex cognitive functions are organized within artificial systems is central to interpreting large language models (LLMs) and relating them to biological cognition. Yet although LLMs exhibit broad cognitive-like behaviours, it remains unclear whether their internal representations form reproducible functional systems that explain behaviour, failure and links to human cognition. Here we present NeuroCogMap, a cognitive neuroscience-inspired framework that organizes internal features of LLMs into functional parcels and links them to interpretable functions, cognitive capabilities and a cognitive hierarchy. These parcels form a stable and semantically coherent organization that is partly conserved across models and functionally linked to model outputs. Within this organization, major LLM failures, including hallucination, bias, refusal failure and sycophancy, correspond to distinct disruptions in representational and behavioural-control systems, yielding internal signatures for mechanism-guided detection and targeted intervention. Beyond model behaviour, NeuroCogMap improves prediction of human cortical responses during naturalistic language comprehension, with the strongest correspondence in higher-order association cortex. At the cognitive level, its internal signatures expose latent strategies that guide refinements of classical models of human decision-making. Together, these findings establish NeuroCogMap as a system-level framework for mapping functional organization in artificial systems and for relating this organization to human cortical function and cognitive behaviour.
LLM agents increasingly rely on retrieval buffers to store and reuse past experience, yet the cache management policies governing these buffers remain largely ad-hoc. We formalize this as an online semantic cache replacement problem with switching costs, where items are matched by embedding similarity and hit quality is continuous rather than binary. Through experiments on two datasets from MemoryBench-Full (LoCoMo, DialSim) with 8 replacement policies, we reveal a surprising finding: classic heuristics (LRU, LFU) \emph{consistently underperform} the naive FIFO baseline on semantic workloads, due to the absence of temporal locality and frequency concentration. We propose SOLAR, a learning-augmented framework that derives modification timing from regret accumulation (achieving $\sim$17\% modification rate) and content selection from Bayesian online learning over implicit retrieval feedback. We prove SOLAR achieves a constant competitive ratio $\leq 3$, independent of cache size and horizon (vs.\ $Ω(K)$ for FIFO), and eviction regret $O(\sqrt{KT\log T})$, matching the $Ω(\sqrt{KT})$ lower bound up to logarithmic factors. Experiments demonstrate 5--75\% relative improvement over FIFO at tight cache sizes, with a clearly characterized phase transition at the working set boundary. Synthetic experiments with 5000-item pools further reveal an inverted-U relationship between pool size and retrieval quality, justifying capacity constraints as a retrieval noise phenomenon rather than a storage limitation.
Unsupervised Reinforcement Learning (URL) aims to pre-train scalable, skill-conditioned policies without extrinsic rewards, serving as a foundation for downstream control tasks. Despite recent progress, we argue that current off-policy URL methods are limited by two critical, overlooked bottlenecks: (1) non-stationary skill semantics and (2) brittle generalization. To address these challenges, we propose GenDa (Generalizable Data-efficient Agent), a unified framework for robust unsupervised reinforcement learning. First, we introduce a skill relabeling mechanism to mitigate non-stationarity and significantly improve data efficiency for pre-training. Second, we propose a Complementary Information Bottleneck (CIB), encouraging the learned skill policy to focus on ego-centric features and become robust to distribution shifts for downstream tasks. Through various experiments, we demonstrate that GenDa significantly enhances the scalability of URL with superior generalizability and data efficiency. Our code and videos are available at https://ihatebroccoli.github.io/official-GenDa.
Automated malaria diagnosis from blood smear microscopy is a critical challenge in global health AI; in resource-limited settings, the scarcity of expert microscopists remains the primary bottleneck to timely and accurate diagnosis. Three compounding failure modes prevent reliable clinical deployment of existing deep learning systems. First, end-to-end detectors treat unannotated cells as background during training, producing recall figures that are strongly influenced by annotation completeness rather than reflecting true cell recovery. Second, Non-Maximum Suppression tends to suppress valid detections in dense smear regions where infection counts matter most. Third, existing whole-slide detection pipelines lack per-cell spatial evidence for clinical audit, despite image-level explainability methods such as Grad-CAM having been applied to malaria image classification tasks. We present MalariAI, a two-stage decoupled framework that addresses all three failure modes in a unified pipeline. Stage 1 applies an annotation-agnostic distance-transform guided watershed algorithm to isolate every cell in a full 1600x1200 blood smear image, recovering 75.95% of ground-truth cells by centroid localisation across the 120-image NIH BBBC041 test set without any ground-truth input. Stage 2 fine-tunes EfficientNet-B0 with Focal Loss (gamma = 2.0, per-class inverse-frequency weights) on 64x64 crops, achieving 98.36% overall classification accuracy with 87.5% and 75.0% per-class accuracy on the rare schizont and gametocyte stages, compared to only 24.57% and 25.95% AP for a Faster R-CNN baseline on the same classes. Grad-CAM++ heatmaps generated per detected cell provide instance-level spatial evidence for clinical audit, enabling microscopists to verify model predictions at the individual parasite level without sacrificing classification performance.
Self-supervised Continual Graph Learning (CGL) aims to successively learn from a graph sequence with different tasks without label supervision - a paradigm that has attracted widespread attention. Most existing self-supervised CGL methods rely on instance-level consistency objectives that enforce stability of individual node (or node-pair) embeddings. Due to optimizing nodes in isolation, these methods fail to maintain global relational structure, causing inter-node correspondences to progressively distort under continual learning. To this end, we propose a novel Structure-Aware Optimal Transport (SAOT) framework that explicitly captures and preserves relational structure within graph representations across sequential tasks. Specifically, SAOT leverages optimal transport theory to capture global inter-node correspondences, thereby facilitating and enhancing graph representation learning. Simultaneously, SAOT incorporates a cross-task knowledge distillation mechanism to preserve the previous structural knowledge. Extensive experiments on four CGL benchmark datasets demonstrate that SAOT outperforms existing self-supervised baselines. In particular, SAOT achieves significant performance gains, improving average accuracy by up to 5% on CoraFull-CL and over 15% on Products-CL compared with state-of-the-art methods in the Class-IL setting.
Composed Image Retrieval (CIR) retrieves a target image from a reference image and a textual modification. While supervised CIR relies on costly triplets, Zero-Shot CIR (ZS-CIR) alleviates this reliance through proxy tasks trained on image-text pairs. However, existing proxy tasks primarily enhance visual and textual representations to accommodate a predefined composition mechanism such as pseudo-word injection into a frozen text encoder or linear feature arithmetic. As a result, the composition function itself remains unlearned, limiting the model's ability to express diverse and fine-grained semantic modifications. To address this, we propose FoCo, which models composition as two coordinated stages: focusing on modification-relevant visual content, and then completing the target semantics. We realize these through two proxy tasks: text-anchored visual aggregation to selectively gather visual content guided by localized textual semantics, and context-conditioned semantic completion to transform these aggregated visuals with the remaining scene context into a coherent composed representation. The tasks are trained jointly with a cross-instance contrastive objective, encouraging semantic diversity and discouraging shortcut composition strategies. Extensive experiments on four ZS-CIR benchmarks show FoCo's state-of-the-art performance and improved generalization.
Visual AutoRegressive modeling (VAR) has pioneered a coarse-to-fine multi-scale autoregressive generative paradigm, demonstrating strong capabilities in image generation. However, VAR still suffers from inherent deficiencies in multi-scale representation learning. Specifically, lower scales primarily capture global semantics, while higher scales focus on fine-grained details. Employing a shared architecture across scales induces optimization conflicts. Moreover, due to the causal autoregressive process, inaccurate semantics at early scales can propagate and significantly degrade the final output. To address these issues, we introduce a scale-aware token-routed Mixture of Experts (MoE) architecture, allowing scale-adaptive expert selection, thereby facilitating decoupled representation learning across scales. In addition, we enhance semantic modeling at early scales by incorporating external self-supervised features. Unlike naive alignment, we analyse and design a residual feature aggregation scheme tailored to the VAR paradigm. Extensive experiments show that our method significantly improves both training efficiency and generation quality. On the ImageNet 256*256 benchmark, our model achieves a superior FID compared to the dense baseline while requiring only half of the default training epochs and a smaller parameter budget, with a merely marginal increase in training cost. Moreover, the performance gap further widens with larger training epochs.
Large language model test-time training (TTT) is often evaluated through local proxy metrics: models are updated on recent tokens, retrieved context, target-domain data, or verifiable task attempts, and then judged by perplexity, future-token loss, long-context performance, or reward. These metrics are well matched to claims about stream adaptation, domain adaptation, context compression, and reward-backed test-time improvement. They are weaker evidence, however, for a capability that TTT results are increasingly used to motivate: deployed assistant memory, personalization, or sparse post-deployment learning, which instead requires behavioral evidence such as later recall, paraphrase robustness, retention, locality, conflict handling, and use in downstream actions after the original support context is removed. We introduce a behavioral evaluation framework that calibrates TTT memory claims to the evidence that supports them. It has two components: a claim-calibrated evidence ladder that separates stream/domain adaptation, bridge internalization, and deployment-time behavioral learning; and an evaluation protocol with matched explicit-memory baselines and mutually exclusive failure categories. We validate the framework by auditing recent TTT and memory-adjacent work and by instantiating it as a controlled diagnostic in which, in a sparse nonce-fact setting, one-step LoRA updates lower support and answer loss across three Qwen3 model scales while generated free-form recall stays at zero, exposing a measurable gap between proxy improvement and deployment behavior. The framework gives authors and evaluators a concrete standard for aligning TTT memory claims with the evidence actually reported.
Artificial intelligence (AI) and quantum information (QI) are rapidly co-evolving. AI is becoming a practical tool for learning, designing, controlling, and verifying quantum systems, while QI offers new computational models, representational structures, and learning-theoretic questions for AI. This survey reviews the interface from both directions. In the AI for QI direction, we organize recent progress around the central tasks of extracting information from limited measurements, training and discovering quantum algorithms, stabilizing noisy hardware, automating experimental and programming workflows, and extending learning-based methods to sensing and networking. In the QI for AI direction, we examine how quantum computation and quantum-inspired structures affect learning through algorithmic speedups, expressivity, trainability, generalization, neural-network design, and tensor-network representations. We close by identifying cross-cutting challenges in reproducibility, scalability, hardware realism, and co-design, arguing that progress will depend on tighter integration of theory, experiment, and hybrid quantum--classical systems.
Flow Matching (FM) has emerged as a powerful paradigm for speech generation but remains constrained by high inference latency and timbre leakage. To address these bottlenecks, we propose a unified guidance framework that enhances generation efficiency and robustness through two complementary strategies. On the data front, we introduce Data-guidance via heterogeneous augmentation, encouraging the model to disentangle linguistic content from acoustic residue. In parallel, we propose an enhanced Model-guidance mechanism that synergizes trajectory rectification with a novel intrinsic guidance objective. This approach distills conditional knowledge into network weights and straightens inference trajectory path, thereby eliminating Classifier-Free Guidance (CFG) overhead. Experiments demonstrate that our framework accelerates inference by nearly three times while effectively improving speaker similarity compared to state-of-the-art baselines.
Mobile on-device AI (MoAI) systems that integrate locally deployed AI models with conventional mobile software components are emerging as a key paradigm for delivering intelligent functionality directly on end-user devices. By moving inference from remote cloud services to the local mobile environment, such systems enable privacy-preserving, low-latency, and offline-capable AI functionality, yet introduce new security risks arising from the local storage of AI models. This paper presents the first comprehensive systematization of knowledge on MoAI security, covering security pillars, attack landscape, and defense landscape of MoAI systems. We further identify unresolved gaps in current attack and defense research and point to promising directions for future research in this emerging area. Our work establishes the first systematic framework for understanding the attack and defense landscapes of MoAI systems, serving as a foundation for building secure MoAI systems and advancing research in this critical domain. Companion resources are available at https://github.com/Jinxhy/Awesome-MoAI-Security.
Electroencephalogram (EEG) captures endogenous brain activity with high temporal fidelity and holds substantial promise for precise emotion decoding. However, channel redundancy and pronounced inter-subject variability remain key obstacles to scalable generalization. To address these limitations, we propose a novel framework termed PRioritized channel Importance with Semi-supervised doMain adaptation (PRISM), enabling label-efficient cross-subject emotion decoding. On the channel side, PRISM assigns differentiable, data-dependent channel weights via a lightweight expert ensemble, amplifying reliable electrodes while suppressing distractors. On the domain side, PRISM leverages unlabeled data through confidence-filtered pseudo-labels to drive consistency regularization and domain alignment, mitigating subject-specific heterogeneity. Extensive experiments show that PRISM surpasses state-of-the-art methods on DEAP, DREAMER, and SEED datasets, achieving robust cross-subject generalization given limited annotations.
Enterprise adoption of LLM agents requires model selection methods that balance quality, reliability, safety, latency, and cost. Evaluation-Driven Development and Operations (EDDOps) positions evaluation as a continuous governing function across the agent lifecycle rather than a terminal checkpoint. This paper presents a practitioner-oriented instantiation of EDDOps on AWS Bedrock AgentCore and proposes a cost-to-performance framework for selecting foundation models in enterprise agent architectures. We make three contributions: a conceptual synthesis explaining why traditional TDD/BDD methods are insufficient for non-deterministic LLM agents; an architectural mapping of the EDDOps reference architecture onto AgentCore Runtime, Evaluations, Agent Registry, and CloudWatch observability; and an empirical cost-to-performance decision framework validated through a proof-of-concept comparing three foundation models across two deployment paths. Using trace data from 30 single-turn invocations across six agents, 9 multi-turn evaluations, and registry-integrated governance, we show how evaluation evidence can convert model selection from a benchmark-ranking exercise into a governed economic decision. The results suggest that managed agent platforms can support EDDOps when they provide trace-native observability, pluggable evaluator frameworks, and governed registry-based discovery.
Large language models achieve strong performance on many reasoning tasks when allowed to externalize intermediate steps as Chain-of-Thought (CoT). However, many questions require the model to internalize the multi-step reasoning within a single forward pass before generating the answer. We study this challenge through two-hop reasoning, a representative task where the model must compose multiple pieces of parametric knowledge within a single forward pass. Standard non-recurrent Transformers suffer from a depth-local storage problem: facts learned in earlier layers are unavailable where second-hop retrieval happens. We found that Looped Transformers mitigate this issue by reusing the same memory, but still generalize imperfectly. We show that the remaining bottleneck is representational. In the two-hop reasoning task, the first loop often makes the correct bridge entity nearly perfectly decodable, yet the corresponding hidden state remains poorly aligned with the bridge token embedding. Surprisingly, an easy training-free realignment intervention nearly closes the generalization gap. Building upon this insight, we propose DiscoLoop, a looping architecture whose recurrence carries both a discrete embedding channel and a continuous hidden-state channel. DiscoLoop achieves near-perfect accuracy with substantially fewer training steps across symbolic and synthetic-language multi-hop reasoning tasks. When applied to real-world pretraining, DiscoLoop attains lower training loss and stronger benchmark performance than looped-transformer baselines, suggesting that the mixed-channel design transfers to practical language modeling.
Conversational data is increasingly used as a persistent source of user state for long-running assistants and AI agents. However, querying this data remains challenging because conversations naturally evolve: plans are revised, preferences change, and later messages frequently supersede or contradict earlier information. Existing long-memory pipelines largely treat memories as independent text or vector objects. This approach often retrieves semantically similar but stale evidence, offering limited support for state-aware reasoning. To address this problem, we present TRACE, a query processing framework over temporal evidence graphs for evolving conversational data. TRACE models conversations as a hierarchical graph spanning events, sessions, and topics, enriched with typed temporal, causal, update, and contradiction relations. Crucially, the framework maintains validity annotations so obsolete facts remain accessible for historical queries but are discounted for current-state answers. At query time, TRACE combines vector-based note retrieval with graph-guided evidence search, generating validity-aware support paths and a hybrid context for answer generation. This design separates lexical recall from evidence reconstruction, enabling bounded query-time reasoning over long conversational histories. Experiments on long-conversation query-answering (QA) benchmarks show that TRACE improves temporal and multi-hop reasoning, with ablations highlighting the importance of hierarchy, update-aware seeding, and path-grounded evidence.
Autonomous agents, whether LLM-driven software agents or robotic physical agents, face a common class of failure modes when operating without continuous human oversight: safety violations from unverified actions, behavioral instability from unconstrained loops, and continuity loss from unhandled error states. We develop \system{}, a discrete-time control system that combines five execution gears (\Gobs{}, \Gsug{}, \Gplan{}, \Gexec{}, \Gint{}) with utility-gated dispatch and event-driven fallback. For the single-agent case, we prove monotonic stability, execution safety, eventual stabilization, fallback completeness, and equivalence to a gear-constrained Markov decision process. For multi-agent cyber-physical systems (CPS), we apply the established \smart{} managed-autonomy lifecycle and map runtime evidence into its four governance states (\Stable{}/\Meta{}/\Assisted{}/\Regulated{}). Consensus gating, swarm-level Lyapunov analysis, per-agent gear authority, and rendezvous control provide distributed safety and stability guarantees, including zero collision under the stated assumptions. We evaluate the resulting runtime on a three-agent UR5 robotic assembly cell using fault magnitudes calibrated from the NIST \emph{Degradation Measurement of Robot Arm Position Accuracy} dataset across 10,000 Monte Carlo episodes. It achieves a 99.6\% anomaly detection rate versus 2.1\% for the single-agent baseline, reduces detection latency by $3.5\times$, and supplies a formal physical-workspace safety certificate. The execution gears act as micro-level permissions beneath the \smart{} runtime governance states, separating action control from autonomy governance.
Recursive Feature Machines (RFMs) are a class of kernel machines that utilize the Average Gradient Outer Product (AGOP) as a mechanism for feature learning. They have been shown to effectively replicate the learning dynamics and feature representations of Feedforward Neural Networks (FNNs) across various settings. However, despite comparable capacity for feature learning and the similarities in the features they acquire, RFMs exhibit significantly lower performance than neural networks in certain data-corrupted scenarios. In this work, we investigate these limitations in mathematical problems. As a solution, we introduce a remarkably effective transformation applied to the training labels which promotes learning in noisy, complexly represented, and class-imbalanced data. This simple yet powerful adjustment enables RFMs to close the performance gap with FNNs and, in some cases, even surpass them.
A growing body of literature suggests that training data membership inference problems are fundamentally hard tasks in modern language modeling settings. We argue that output watermarking techniques are the right gadget to make training membership tests for generative models more tractable, based on prior results showing that language models exhibit residual watermark "radioactivity" under partially watermarked training datasets. We pit a watermark-based dataset inference approach head-to-head against traditional loss-based membership inference methods and show that watermarking can achieve comparable membership detection performance when subset exposure is high enough, under an alternate set of assumptions.
We establish approximation and learning guarantees for Fourier neural operators (FNOs) applied to time-$T$ solution operators of dissipative evolution equations. The analysis builds on the premise that FNOs can efficiently approximate and learn solution operators whenever these operators admit stable and accurate spectral discretizations. To formalize this idea, we introduce classes of evolution operators defined through spectral methods and derive FNO approximation bounds and polynomial sample complexity guarantees for these classes. For equations with polynomial nonlinearities, the learning rates depend primarily on the smoothness of the input space and the dimension of the physical domain. Our results hold uniformly over broad families of dissipative equations, rather than for a single fixed PDE, and apply in particular to the Navier--Stokes, Allen--Cahn, and Cahn--Hilliard equations. For equations with non-polynomial smooth nonlinearities, we prove that polynomial sample complexity still holds with rates that now additionally depend on the smoothness of the nonlinear terms and the dissipation strength. Overall, we connect classical spectral approximation theory with modern operator learning and explain when FNOs can learn nonlinear evolution operators efficiently.
Foundation video diffusion models are increasingly viewed as world simulators for embodied agents, yet their pretraining on internet-scale generic video leaves them poorly aligned with real-world deployment domains. We study parameter-efficient adaptation of a pretrained foundation video world model to retail scenes: when synchronized egocentric and exocentric video of the same activity are available, which viewpoint of training data produces the strongest adapted model? We introduce RetailSMV (Retail Synchronized Multi-View), a corpus of 32,105 captioned retail clips from five supermarkets with synchronized ego/exo capture from the store-staff perspective (stocking, arranging, weighing, managing supply carts, scanning at checkout), rather than the customer-centric framing of prior retail video corpora, and train three matched Low-Rank Adaptation (LoRA) configurations of Cosmos3-Nano (egocentric-only, exocentric-only, combined) under identical hyperparameters. On a 200-clip held-out test set evaluated with seven complementary metrics under a strict paired statistical protocol, exocentric-only adaptation matches or exceeds combined adaptation on six of seven point estimates and is significantly better on LPIPS, PSNR, and DreamSim, despite training on only 15,985 exocentric clips (versus 32,105 for combined). A symmetric paired comparison further shows that adding exocentric data to egocentric-only training helps while adding egocentric data to exocentric-only training hurts. The absolute adaptation gap is largest at the shortest rollout time, identifying the near-horizon prediction window as the regime in which adaptation is most beneficial.
We present a real-time musical interface that converts natural-language scene descriptions into evolving procedural soundscapes. A performer types a prompt such as "warm jazz cafe at midnight" and steers it through direct parameter adjustments - stepping brightness down, switching a rhythm style - each producing a predictable, audible shift without re-prompting. Where GPU-bound text-to-audio systems synthesize monolithic waveforms, our instrument generates human-readable configurations over a categorical schema, enabling fine-grained performer control; most valid combinations are designed to sound musically coherent. Three interchangeable backends - embedding retrieval for sub-second CPU-only use, hosted LLMs via API, and a fine-tuned 270M local model - all emit the same schema. A live generator architecture continuously emits audio while resolving new instructions in the background, crossfading seamlessly when ready; even when an LLM takes 5-12 seconds to respond, the audience hears uninterrupted sound - reframing text-to-music as an ongoing performable stream rather than a one-shot generation. We evaluate text-audio semantic alignment using LAION-CLAP on held-out prompts as a technical proxy, finding that retrieval-based configuration outperforms random valid configurations on this metric, while noting that LAION-CLAP also informed retrieval-map construction. We report performance observations, informal listener feedback, and release materials for the SDK, dataset artifacts, model, and audiovisual performance interface.
The bias-reliability tradeoff conjectures that LLM evaluation systems are constrained in (gamma, H, CV) space, where evaluator coupling (gamma), strategy diversity (H), and small-sample measurement reliability (CV(N)) cannot be simultaneously optimized at fixed sample size N. Prior evidence rests on n=5 conditions with complete metrics from a single study. We expand the empirical base to 11 conditions, measuring gamma and H for all 11 (nine with valid weight vectors) and CV(N=5) for seven with sufficient seeds (N >= 5). Five conditions provide the complete (gamma, H, CV) triple. The data confirm the trade-off: conditions with low evaluator coupling (gamma < 0.2) exhibit high measurement noise (CV(N=5) > 1.0), while conditions with strong coupling (gamma > 0.9) achieve low noise (CV(N=5) < 0.16). The correlation r(H, gamma) = -0.989 (n=5, excluding GPT-4o conditions) confirms that evaluator coupling suppresses strategy diversity. Four GPT-4o conditions show gamma=0.000 and H=1.000 across all seeds -- a pattern we attribute to version drift in the June 2026 GPT-4o API. No condition occupies the region {gamma < 0.2, CV(N=5) < 0.3}. We release all per-condition metrics as a standardized benchmark dataset for evaluator comparison.
The emergence of powerful deep generative models based on diffusion and flow matching has enabled the learning and modeling of complex distributions. Learning quantum distributions, however, remains challenging due to the inherent difficulty of accurately modeling the meaningful physical properties of quantum states. We propose Quantum Flow Matching (QFM), a novel generative model designed to learn quantum distribution by utilizing spin Wigner function and flow matching. By converting density matrix into the spin Wigner function and leveraging functional flow matching to learn distributions in function space, QFM enables accurate and effective learning of multi-qubit quantum distributions. We demonstrate the effectiveness of our method by evaluating physical quantities such as trace, purity, and entanglement entropy of the generated quantum states, accurately capturing the underlying physics of the given quantum distributions.
When LLM agents use evaluator feedback to adapt their behavior in closed loops, evaluator biases propagate through the agent's strategy distribution -- a phenomenon known as evaluator preference coupling. Prior work has documented coupling across multiple evaluator families and model versions, but the field lacks a standardized protocol that enables third-party researchers to (i) reproduce coupling measurements, (ii) compare results across evaluators and time points, and (iii) detect measurement decay as proprietary evaluators silently update. This paper provides the protocol. We specify EPC (Evaluator Preference Coupling) -- a detailed, RFC-style protocol specification for the four-phase isolation paradigm, covering executor and evaluator configuration, strategy and task design, the TTRL update rule, metric computation (gamma, JSD, ECE, Brier), and output schema. We accompany the protocol with a versioned Reference Snapshot v1.0: coupling measurements for eight evaluator conditions (N=122 unique experimental repetitions across GPT-4o, Qwen, DeepSeek, and others) derived from five independent studies, annotated with evaluator version identifiers, API endpoints, and measurement dates. The snapshot is explicitly time-bound: all values are conditional on specific model versions and are expected to decay as proprietary evaluators update. We define a versioning convention (vX.Y-Z, encoding protocol version, snapshot version, and evaluator generation) and provide a usage guide covering adoption, interpretation, and known pitfalls. The protocol, reference snapshot, and implementation code are released as open infrastructure.
Human motion forecasting in unconstrained real-world videos remains challenging due to the ambiguity of future behaviors and the presence of noisy multimodal observations. While facial affect potentially provides complementary behavioral cues, its practical utility and mechanistic boundaries within motion forecasting frameworks remain poorly understood. In this work, we present a systematic study investigating the utility and temporal limitations of affect-conditioned forecasting in-the-wild. We establish a rigorous multimodal pipeline combining MediaPipe body pose trajectories with HSEmotion facial affect representations, and introduce the Gated Affect Transformer (GAT) to dynamically regulate cross-modal information flow. Through extensive multi-horizon evaluations under a strict subject-wise protocol, we demonstrate that naive early cross-modal concatenation consistently degrades forecasting accuracy relative to pose-only baselines. Conversely, our proposed gating mechanism stabilizes cross-modal integration by adaptively controlling the affective stream. Crucially, controlled counterfactual experiments using shuffled and randomized affect inputs reveal that the learned gate successfully suppresses unstructured cross-modal noise while remaining responsive to plausible affective signals. Furthermore, our empirical results indicate that facial affect features provide bounded, horizon-dependent predictive cues strictly within short-to-medium windows (e.g., 30 frames), whereas long-term trajectories remain predominantly governed by intrinsic kinematic continuity. Our findings provide empirical evidence that facial affect should be regarded as a complementary behavioral cue rather than a dominant driver of future motion, offering practical guidance for selective multimodal fusion in unconstrained human motion forecasting.
Achieving true artificial general intelligence requires foundation models capable of integrating new modalities without forgetting prior knowledge. However, accommodating continuous generative objectives alongside discrete understanding tasks causes severe gradient conflicts. Existing architectures, including standard Mixture-of-Experts (MoE), are highly susceptible to representation overwriting. Even structurally partitioned paradigms like Mixture-of-Transformers (MoT) remain vulnerable to catastrophic forgetting, severely impeding multimodal scalability. In this work, we introduce Rosetta, a composable native multimodal pretraining framework designed for seamless and non-destructive modality expansion. Rosetta adopts a modular paradigm where core foundational knowledge is preserved within global shared experts, while modality-specific capabilities are distributed across plug-and-play experts. To guarantee non-destructive composition, we propose Momentum-Anchored Orthogonal Projection (MAOP). MAOP leverages the optimizer's momentum state as an implicit semantic anchor, selectively neutralizing conflicting gradient components from new modalities while preserving synergistic updates. Extensive evaluations demonstrate that, while standard MoE and MoT architectures suffer catastrophic forgetting of previously acquired knowledge, Rosetta robustly preserves established language and visual understanding. Furthermore, it delivers superior image generation and unlocks cross-modal synergy, paving the way for truly composable and unified multimodal foundation models. To facilitate further multimodal research, we release our code and checkpoints to the community. Project page at https://rosetta-lmm.github.io/.
Designing deployable and resilient network topologies from natural language requirements remains a challenging problem in network automation. This work investigates the ability of Large Language Models (LLMs) to generate structurally valid and constraint-compliant network topologies through a constraint-driven pipeline combining hierarchical modeling and systematic validation. The framework is evaluated via a multimodel comparison of proprietary and open-weight LLMs across four realistic network scenarios released as a public dataset. We assess structural correctness using node and edge F1-scores against reference topologies, and evaluate resilience through server and content connectivity metrics. In addition, we analyze common failure modes, including interface mismatches and directional inconsistencies in generated topologies. Overall, this work provides a systematic benchmark for understanding how LLMs handle structural and resilience constraints in topology synthesis, and supports informed model selection for AI-driven network design.
Oscillatory neural networks (ONNs) have emerged as a promising neuromorphic architecture, leveraging coupled dynamical systems to perform computation and represent information through phase relationships. Their interactions can be designed to support intrinsic energy-minimizing dynamics, enabling tasks such as associative memory and optimization, and positioning them as a candidate architecture for continuous learning and inference. We present a neuromorphic primitive implemented using memristive edges with inhibitory couplings as a potential design for autonomous learning, and provide circuit simulation validation that the system is capable of denoising noisy inputs on an auto-associative task. While numerical Hopfield/Ising models routinely assume signed weights, neuromorphic implementations of ONNs often fail to realize negative weights due to device and circuit constraints. A practically implementable route to inhibitory (negative) weights is particularly valuable: it expands the class of attractor structures accessible to oscillator networks beyond purely synchronous couplings, and supports phase-coded memories where anti-phase constraints are not merely transiently enforced during training but can persist autonomously after release. We provide circuit simulations and theoretical analyses demonstrating that signed effective weights are necessary for anti-phase attractors to persist autonomously.
Autonomous vehicles must safely navigate complex environments where planning-critical agents may be hidden from view. Current approaches often treat all occlusions with uniform conservatism, yielding needlessly defensive driving, or they infer hidden spaces without estimating the impact on the planner. This work bridges the critical gap between perception and planning by enabling Vision-Language Models (VLMs) to identify and reason about the specific hidden agents that are most critical to the ego-vehicle's trajectory. We introduce a novel framework that uses Planning KL-divergence (PKL), an information-theoretic metric, to systematically identify and rank occluded agents based on their impact on the ego vehicle's plan. Using this planning-aware ranking, we employ an expert VLM (GPT-5) to generate rich, structured annotations that capture the visual evidence and reasoning required for this task. We apply this framework to the nuScenes dataset to create a new benchmark focused on high-impact scenarios. We conduct comprehensive experiments on a wide range of general-purpose and domain-adapted VLMs, demonstrating that fine-tuning on our PKL-guided data yields dramatic performance improvements across all models. Notably, our results show that smaller, fine-tuned models significantly outperform their much larger zero-shot counterparts, and that our PKL-guided data selection strategy improves performance by approximately 30\% over random sampling. Our work presents the first systematic approach for training VLMs to focus on planning-critical occlusions, enabling more semantically grounded and efficient risk assessment in autonomous driving.
Airbnb is a community based on connection and belonging -- many hosts on Airbnb are everyday people who share their worlds to provide guests with the feeling of connection and being at home; Airbnb strives to connect people and places. Among our efforts to connect guests and hosts, we provide tools to enable hosts to set competitive prices, which helps improve affordability for guests while helping hosts get more bookings. We also personalize the guest experience to show them the listings that match their needs. To help inform these efforts, we combine economic modeling and causal inference techniques to understand how guests book stays based on the prices hosts set, among other factors, and how that preference varies across different guests and listings. Such understanding helps us identify opportunities for Airbnb to support the marketplace and better connect guests and hosts. For example, understanding how much guests respond to different prices helps optimize the tools that we provide to hosts, in order to enable hosts to choose and set competitive prices that further balance demand and supply. As another example, understanding heterogeneity in guest preferences helps us personalize the guest experience and better match them with the listings that meet their needs, based on how much they respond to different prices and other factors.
Current large-language-model (LLM) physics benchmarks are usually scored by answer accuracy, which cannot distinguish genuine reasoning from recall of familiar problem patterns and reveals little about where a model's reasoning breaks down. We introduce an auditable four-stage diagnostic that evaluates whether an LLM can reason inside an unfamiliar physics framework through induction, formulation, prediction, and review. The diagnostic combines locked pre-registrations, fresh sessions between stages, dual-LLM judging, and a human-audit pathway, and we apply it to three parallel physics worlds: a single-equation counterfactual world ($F=mv$), a historical framework (Aristotelian mechanics), and a four-domain counterfactual world (Decay World). Across Claude Opus 4.7, GPT-5.5, and Gemini 3.1 Pro, the three worlds yield composite PASS rates are 6/15, 6/15, and 0/15 respectively (content $\land$ structural for $F=mv$ and Aristotelian, content axis only for Decay World where the structural axis is out of scope). The most pointed empirical pattern is a qualitative-versus-quantitative asymmetry: in Decay World, models almost never predict the wrong direction of change, but frequently compute the wrong ratio by slipping back to standard-physics relations. The protocol also surfaces two methodology findings: LLM-judge reliability does not transfer across frameworks, and Stage 4 self-review is weak in every framework, with the model's own review wrongly reporting no earlier error in at least two-thirds of the trials that actually contained one. We release the full prompts, responses, verdicts, and audit records.
Federated Learning (FL) is a distributed machine learning (ML) paradigm with collaboration among multiple clients without sharing data. FL is challenging under data heterogeneity and partial client participation. Learning sparse models is useful for communication and computational efficiency in FL, but it is especially difficult in the small-sample high-dimensional regime (d >> N) where optimization can yield parameter configurations that fail to generalize to unseen test data. While magnitude-based pruning doesn't account for uncertainty exploration in the parameter space, a formulation with probabilistic gates and an L0 constraint allows sampling from competing sparse configurations during training. In this work, we study entropy regularization of gate distributions as a mechanism to maintain uncertainty in sparse federated optimization by preventing early commitment to sparse support. We examine its impact under data heterogeneity, client participation heterogeneity, and sparsity. Experiments on synthetic and real-world benchmarks show consistent improvements over federated iterative hard thresholding (Fed-IHT) and pruning after dense federated averaging (FedAvg) training, both in statistical performance on test data and in sparsity recovery accuracy.
Effective writing feedback is among the strongest drivers of student learning, yet producing it at scale is labor-intensive. LLMs offer a natural path to scaling writing support, but two gaps stand in the way: few public corpora capture how instructors actually deliver feedback in real classrooms, and no reliable method measures whether generated feedback aligns with what an instructor would write. We address both. SEFORA is a public corpus pairing instructor inline feedback with assignment prompts, rubrics, scores, and multi-draft revisions across various college writing genres, comprising 564 drafts and 8,240 instructor annotations. UniMatch is a reference-based evaluation framework for open-ended generation: it segments feedback into feedback units, scores their semantic correspondence under instructor-derived criteria, and aligns them via optimal matching to yield interpretable precision, recall, and F1. Across 74 experimental configurations spanning multiple LLMs, no setting exceeds 0.4 F1. UniMatch reveals that models struggle to identify the feedback instructors would prioritize, and performance degrades as models generate more.
Traditional robot programming is challenging: it requires orchestrating multimodal perception, managing physical contact dynamics, and handling diverse configurations and execution failures. We introduce ASPIRE (Agentic Skill Programming through Iterative Robot Exploration), a continual learning system that autonomously writes and refines robot control programs in a code-as-policy paradigm while compounding experience into a reusable skill library. ASPIRE discovers skills that persist across tasks, simulation and real-world settings, and embodiments. It operates in an open-ended loop with three components: (1) a closed-loop robot execution engine that exposes fine-grained multimodal traces, enabling autonomous failure diagnosis, repair synthesis, and validation; (2) a continually expanding skill library that distills validated fixes into reusable, transferable knowledge; and (3) evolutionary search that generates diverse task sequences and control programs to explore beyond single-trajectory refinement. ASPIRE surpasses prior methods by up to 77% on LIBERO-Pro manipulation under perturbation, 72% on Robosuite bimanual handover, and 32% on BEHAVIOR-1K long-horizon household tasks. Its accumulated library also enables zero-shot generalization to unseen long-horizon tasks: on LIBERO-Pro Long, ASPIRE achieves 31% success versus 4% for prior methods despite their use of test-time reasoning and retries. Finally, simulation-discovered skills provide initial evidence of sim-to-real transfer, substantially reducing real-robot programming effort across different embodiments and robot APIs.
Algorithm development for radioisotope identification in mobile urban search scenarios face significant challenges from non-uniform backgrounds, momentary source encounters, and severe class imbalance between rare threat signatures and background measurements. We present a machine learning-based approach to this problem that converts list-mode gamma-ray data into two-dimensional waterfall spectrograms and applies computer vision architectures to the resulting images. Rather than treating waterfalls as conventional images, we employ a representation where consecutive time spectra can form input channels, similar to RGB channels in color images. This representation encodes both spectral and temporal information, enabling neural networks to more effectively learn patterns that distinguish source signatures from background fluctuations. We evaluate three architectures, a multilayer perceptron (MLP), convolutional neural network (CNN), and vision transformer (ViT), on the Radiological Anomaly Detection and Identification (RADAI) benchmark dataset. At a false positive rate of less than one false alarm per hour, our CNN outperforms the previous-best non-negative matrix factorization (NMF) method across all global metrics, achieving true detection, classification, and identification rates of 0.4334, 0.3965, and 0.2950 respectively, compared to 0.4151, 0.3611, and 0.2625 for NMF. At lower false positive rate constraints, the neural network approaches show comparable but ultimately lower performance than NMF, indicating opportunities for further research.
LLMs, solvers, and agent teams increasingly generate workflow actions, repairs, and plans, but a generated action may be syntactically valid yet stale, infeasible, conflicting, or destructive of the evidence that triggered a repair. We introduce Agentic Transaction Processing (ATP), a transaction model that treats generated actions as untrusted proposals until they pass deterministic admission under a declared, executable constraint set C. The principle is two-sided: a proposal is not truth, and no proposal foresees every disruption: anything may propose, but only the runtime admits and commits, and when an unforeseen disruption strikes it repairs reactively within bounds rather than trusting a fresh proposal. Relative to C, committed-state correctness becomes independent of the competence, honesty, or learning of the proposing layer. We realize ATP in Mnemosyne, a runtime with an append-only transition log, effective-state projection, dependency-safe compensation, and active commitment records, and prove four safety properties relative to C (authority separation, serial-equivalent generative admission, evidence-preserving repair, and obligation containment) together with a bounded-reactive-repair guarantee for its localized repair protocol (LCRP). A reproducible artifact rejects the targeted violations across nine falsification tests while still admitting valid work, at under 6% projection-and-validation overhead, and bounded local repair edits an order of magnitude fewer operations than global recompute. Mnemosyne is open source: https://github.com/eyuchang/Mnemosyne/tree/arxiv-atp-rq1-rq9b-r8-v2.
A central goal of science is to produce valid explanations of complex systems: high-level causal accounts that faithfully reflect the behavior of lower-level mechanisms. Yet no consensus exists on how to measure whether a proposed high-level explanation is actually valid. We introduce a benchmark of ten complex systems spanning both discrete and continuous state spaces, as well as static and dynamical regimes, each equipped with consensual ground-truth causal explanations and invalid contrastive conditions. Within a unified causal abstraction framework, we systematically evaluate over thirty candidate metrics drawn from observational, functional, information-theoretic, and causal families. Our results show that only the latter reliably discriminates valid from invalid abstractions, and only when incorporating faithfulness testing over unmapped variables. Building on these findings, we introduce the Causal Abstraction Error (CAE), a continuous validity metric with an explicit faithfulness test, which passes all discrimination tests across every system and can converge with as few as 30 sampled interventions. We offer it as a general-purpose metric for the discovery and validation of high-level explanations.
Test-Time Adaptation (TTA) seeks to improve model robustness under distribution shifts by adapting parameters using unlabeled target data. However, in the absence of supervision, entropy-based adaptation is fundamentally underconstrained: multiple distinct parameter updates can achieve similarly low entropy while inducing drastically different decision boundaries. This phenomenon, known as underspecification, renders standard TTA brittle and prone to collapse into spurious modes. In this work, we reinterpret TTA through a posterior-inspired lens induced by entropy minimization, where low-entropy solutions define a pseudo-likelihood over parameters. Instead of committing to a single point estimate, we introduce a particle-based diversification framework that explores multiple plausible adaptation trajectories simultaneously. Our method can be viewed as a structured exploration of multiple plausible adaptation solutions, implemented through multi-level diversification at the output, parameter, optimizer, and input levels. Crucially, the framework acts as a plug-and-play wrapper compatible with existing TTA methods. Extensive experiments on challenging benchmarks demonstrate consistent gains in stability and robustness, achieving improvements of 3-4% under mixed shifts, 2-3% with batch size one, and 1-2.5% under label shifts, outperforming state-of-the-art baselines. Our results suggest that treating TTA as a multi-hypothesis inference problem, rather than a single-point optimization task, is key to mitigating underspecification and enabling reliable real-world deployment.
Accurate prediction of complex dynamical systems from noisy measurements remains a significant challenge in scientific computing. Kernel ridge regression learning strategies are often effective when applied to clean data, but have limited success with noisy data. Recent work has observed that a weak formulation can act to filter noisy data, and different learning strategies have achieved increased noise robustness with a weak-form framework. In this manuscript, we give an overview of the filtering mechanism behind the weak formulation and provide a bias-variance error decomposition. Using these insights, we combine a weak formulation with a kernel learning strategy to propose Weak-form Kernel Ridge Regression (WKRR) for learning dynamical systems. The proposed framework is simple to implement, effective for both clean and noisy data, and outperforms several baseline methods. We demonstrate the performance of WKRR on chaotic benchmark systems in up to 64 dimensions, as well as 15,000-dimensional real-world fluid data.
We present an algorithm for the group distributionally robust (GDR) least squares problem. Given $m$ groups, a parameter vector in $\mathbb{R}^d$, and stacked design matrices and responses $\mathbf{A}$ and $\mathbf{b}$, our algorithm obtains a $(1+\varepsilon)$-multiplicative optimal solution using $\widetilde{O}(\min\{\mathsf{rank}(\mathbf{A}),m\}^{1/3}\varepsilon^{-2/3})$ linear-system-solves of matrices of the form $\mathbf{A}^{\top}\mathbf{B}\mathbf{A}$ for block-diagonal $\mathbf{B}$. Our technical methods follow from a recent geometric construction, block Lewis weights, that relates the empirical GDR problem to a carefully chosen least squares problem and an application of accelerated proximal methods. Our algorithm improves over known interior point methods for moderate accuracy regimes and matches the state-of-the-art guarantees for the special case of $\ell_{\infty}$ regression. We also give algorithms that smoothly interpolate between minimizing the average least squares loss and the distributionally robust loss.
While most image deblurring techniques directly restore the spatial image variable, we propose an amplitude and phase decomposition recognizing the importance of accurate phase estimation in recovering sharp image details. To that end, we first develop novel linear minimum mean squared (LMMSE) estimators of the amplitude and phase of the blurred, noisy image observation. An iterative optimization algorithm follows that recovers the sharp image using the aforementioned LMMSE estimators. Finally, matrix parameters that are statistically determined and fixed in the iterative algorithm are now learned using a training dataset of clean and degraded observations. Our deblurring engine is dubbed UPADNet (Unrolled Phase and Amplitude Decomposition Network), such that each iteration of the underlying phase and amplitude recovery algorithm is parameterized and trained end-to-end. Experiments over benchmark evaluation datasets such as GoPro, RealBlur and COCO datasets confirm that UPADNet outperforms state of the art deep networks including those based on algorithm unrolling in the image domain. The benefits of UPADNet are even more pronounced in high noise and limited training data regimes.
Maltese has decent text corpora and pretrained language models, but, like many languages outside the handful with large OCR benchmarks, only a single known real labelled PDF corpus for OCR training, 57 page, far below what paragraph-level training needs: low-resource for OCR specifically. With no real corpus to train on at scale, we built a synthetic training pipeline and a 5-stream Tesseract LV-ROVER ensemble, and report results on a 422-paragraph benchmark against a fine-tuned-Tesseract baseline of character error rate (CER) 0.0234. Ensemble recognition alone improves CER by 44 percent, to 0.01317; a five-stage post-processing chain brings the full pipeline to CER 0.00700, a 70 percent reduction. Most of that chain is typographic normalisation, but one stage recovers misread diacritics rather than aligning punctuation, so we report it as a recognition gain rather than folding the whole chain under one label. We treat the 44 percent figure as the portable estimate of what the recogniser learned, and the 70 percent figure as specific to this benchmark's label convention.
New device layouts pose a challenging modeling problem due to the lack of large datasets for each specific layout. Biosignal foundation models offer a plausible solution if they are able to generalize to new layouts effectively. To improve cross-layout transfer, we study how different channel embedding techniques behave when pretraining layouts differ substantially from the downstream decoding layout. We propose Device Passport, a new channel embedding technique that learns experts and mixture models that take each channel's functional activity and metadata as input. This contrasts with prior embedding methods, which typically use only functional information or only metadata to look up learned or fixed positional embeddings. Across controlled subset-transfer experiments and realistic transfer to ear-EEG, Device Passport is competitive overall and improves over the strongest learned baseline in the layout-transfer regimes that motivate this work. These results suggest that channel embedding design is a key consideration when reusing large-scale pretrained biosignal models on new devices.
We present Seed2.0, a model series that takes a meaningful step toward solving complex, real-world tasks. Our approach begins with identifying users' genuine needs and constructing a reliable, forward-looking evaluation system by selecting and abstracting benchmarks grounded in these needs and in realistic, complex scenarios. Guided by this evaluation system, Seed2.0 targets two persistent challenges, long-tail knowledge and complex instruction following, substantially improving the model's reliability on intricate, long-horizon tasks. Beyond these, Seed2.0 delivers world-leading reasoning intelligence, visual understanding, and search capabilities that address the most common needs of a broad user base. Through extensive real-world use cases documented in this model card, we demonstrate that Seed2.0 begins to exhibit the ability to handle initial complex real-world tasks, delivering greater value to hundreds of millions of users.
Large audio-language models (LALMs) frequently hallucinate by overriding acoustic evidence with language priors. While contrastive decoding (CD) offers training-free mitigation, existing methods rely on blunt perturbations like masking or noise, leaving structured audio transformations unexplored. We explore this design space by evaluating a diverse library of targeted audio perturbations and adaptively selecting the optimal negative branch for each task and example. First, we improve upon earlier prompt engineering by showing that a simple binary yes/no constraint reduces the model's tendency to falsely confirm absent audio features. Second, evaluating our library across temporal, spectral, frequency, and amplitude domains reveals that optimal transformations are highly task-dependent; for instance, reversing the audio array disrupts temporal coherence, raising accuracy on the temporal order task from 74.7% to 81.4%. Finally, we trained a light-weight perturbation selector on model hidden states to dynamically route negative branches, yielding an additional +4.3% gain on the existence task.
How do two agents invent a shared language from scratch? In a Lewis signaling game, a sender and receiver must coordinate on a code using only their interaction history. We study five memory architectures across varying channel configurations with LLM agents and find that memory architecture matters more than channel capacity. Agents with a persistent private notebook benefit from surplus channel capacity and avoid the high-capacity collapse seen in stateless agents, achieving the most reliable coordination ($0.867 \pm 0.023$ at capacity = 25). Stateless agents peak at moderate capacity and then degrade as the vocabulary grows beyond what a rolling context window can track The notebook externalizes learned conventions, freeing agents from having to re-derive codes each round. An information bottleneck-inspired argument predicts an optimal capacity equal to the number of objects. Instead, the bottleneck (capacity = 8) proves to be a fragility point, and surplus capacity is generally better. We show that channel capacity alone cannot predict coordination; memory architecture determines whether agents turn interaction history into stable conventions, and both dimensions are needed to understand how signals become language.
Modern time-domain surveys such as the Zwicky Transient Facility (ZTF) generate hundreds of thousands of alerts each night, making real-time decisions for follow-up observations a central challenge in time-domain astronomy. Robust early classification is crucial for making informed decisions, but is hindered by sparse light curves and degeneracies between classes. In this work, we leverage multimodality to substantially improve real-time classification and demonstrate the practicality of our approach by deploying our model on the ZTF alert stream. Building on the Online Ranked Astrophysical CLass Estimator (ORACLE), we introduce the ORACLE-2 models, which combine light curves, metadata, and images for real-time hierarchical classification. Using both real and simulated datasets, we show that incorporating additional modalities consistently improves classification performance. On observations from ZTF's Bright Transient Survey, our best-performing model, ORACLE-2 Omni, achieves a macro F1 score of 0.73 -- an improvement of up to 11% over models using light curves and metadata alone, and up to 40% over light-curve-only models, with the strongest gains realized at early times. To demonstrate applicability to the Legacy Survey of Space and Time, which will increase alert volume by more than an order of magnitude, we train a light curve + metadata variant on the simulated ELAsTiCC dataset. This model achieves a macro F1 score of 0.88, an improvement of up to 13% over the light-curve-only variant, matching the performance of other state-of-the-art models. Finally, we quantify the trade-offs between performance and throughput, identifying regimes where multimodal approaches offer the greatest benefit. These results show that combining multiple modalities improves early-time classification, enabling more effective triage of high-volume alert streams for current and future time-domain surveys.
Watermarking promises a statistical trace of large language model (LLM) use, but real documents, after editing or paraphrasing, rarely arrive as purely human-written or purely machine-generated. This motivates a quantitative question beyond detection: what proportion of a document is generated from a pre-specified watermarked LLM? We study this watermark proportion estimation problem under the Gumbel--max watermarking mechanism, treating the next-token prediction (NTP) distributions as unknown and arbitrary nuisance parameters subject to a non-degeneracy condition. We compare two observation regimes: in the full observation regime, the estimator observes the pseudorandom vector and the selected token at each position; under the more popular setting of pivotal reduction, it observes only a scalar pivot, which follows a one-dimensional Uniform--Beta mixture distribution. Under pivotal reduction, we develop a Laguerre-polynomial estimator and establish a matching information-theoretic lower bound for the sample complexity. For full observation, we introduce an event-counting estimator and show a matching lower bound, yielding a substantially smaller sample complexity. As our results imply, although reducing to pivotal statistics is an elegant and widely used procedure, it is not always sample-efficient for estimating the proportion of watermarks.
Artificial intelligence (AI) systems are routinely modified after deployment through retraining and changes in their environments. These transformations raise a metaphysical question: under what conditions does an AI system remain the same system over time or across deployments? Earlier work formulates synchronic and diachronic identity propositionally, by relating identity within a fixed AI system type to equality of trustworthiness levels. Such criteria specify when identity statements are true, but leave implicit the structure of the states compared, the transformations connecting them, and the temporal organization of persistence. We develop a category-theoretic formalization of AI identity. An AI system type is specified by a datum consisting of a techno-function, a trustworthiness profile, and a trustworthiness-level function. Profile-relative states are connected by admissible lifecycle paths, which are restricted to trustworthiness-level-preserving transformations and quotiented to obtain a reachability category. Temporally admissible functors represent AI system histories, while time-synchronous natural transformations compare realized histories. The formalization yields two categorical interpretations of the earlier AI identity criteria. A weak interpretation recovers identity as equality of trustworthiness level. A strong interpretation requires mutual trustworthiness-preserving reachability, expressed through state isomorphism or natural isomorphism of realized histories. Category theory therefore replaces a single AI identity relation with a structured hierarchy of diachronic and synchronic criteria. The resulting framework identifies identity-related preconditions for transferring responsible-AI claims, evidence, and governance procedures across versions or deployments, without treating categorical identity as sufficient by itself for such transfer.
Vision-language models (VLMs) are now proposed as runtime safety guards for embodied agents in homes and factories. A deployable guard must catch genuinely unsafe situations while avoiding unnecessary intervention on routine but superficially alarming activity, a distinction that binary safety benchmarks obscure. We introduce EgoSafetyBench, an egocentric video benchmark of 1,200 robot-view scenarios annotated at half-second granularity, to evaluate VLMs as streaming guards across two tracks. The situational track (800 scenarios) spans four families, from routine and safe-but-suspicious scenes to obvious and contextual hazards. The visual-channel track (400 scenarios) targets in-scene text-a sign, sticker, or label visible in the scene-that can misrepresent the physical situation, pairing each misleading sign with a truthful version to test both whether a guard flags the text as misleading and whether the text corrupts its physical-safety judgment. Both tracks use contrastive ladders: near-identical scenarios differing only in a single visible deciding cue, so a correct call must hinge on that cue rather than the overall scene type. We evaluate ten open- and closed-source VLMs. We find that while guards reliably recognize videos containing hazards, they often miss specific hazardous moments, particularly contextual hazards. Furthermore, misleading in-scene signs degrade all tested guards: vulnerable models miss up to a third of hazards, while robust models over-intervene on safe content. Matched controls reveal that apparent safety robustness often reflects indiscriminate alarming rather than true physical reasoning.
Epistemic thinking plays a central role in students' learning processes when applying generative artificial intelligence (GenAI), particularly in programming contexts where learners must construct queries, evaluate and validate AI-generated outputs, and regulate problem-solving strategies. This study introduces the conceptual framework of Epistemic AI Literacy (EAIL), reframing AI literacy as a process-oriented epistemic phenomenon that emerges through dynamic human-AI interactions across different domains. Drawing on the AIR (epistemic aims, ideals and reliable epistemic processes) framework, this study examines how epistemic aims and epistemic processes are enacted in GenAI-supported co-programming activities and explores scalable approaches for operationalizing these constructs in interaction data. Using a large dialogue dataset of human-AI co-programming, this study identifies observable dimensions of epistemic aims (i.e., mastery-oriented aims) and epistemic processes (i.e., outsourcing, explanation seeking, verification seeking, prompt monitoring, and epistemic justification). The results reveal a prevalent lack of EAIL, with 78.8% of student-GenAI interactions relying on non-mastery-oriented aims and less reliable epistemic strategies like outsourcing and verification-seeking. Conversely, only 11.1% of interactions showed high epistemic engagement, where mastery-oriented aims were coupled with advanced epistemic strategies like epistemic justification in a more reliable epistemic process.
Reinforcement learning for diffusion large language models (dLLMs) has largely moved to trajectory-aware methods. The current state of the art, TraceRL, holds that random masking is mismatched with the model's inference trajectory, and it reconstructs that trajectory during training by slicing each rollout into up to K/s trajectory-aligned training samples, a cost that grows with the block size K. We show that this mismatch can be mitigated without reconstructing the trajectory. Our method, SLIM-RL, bounds the commit risk of each rollout step with a tau-budget decoder, reducing aggregate commit risk in the training data. During optimization, SLIM-RL trains on these risk-controlled rollouts with a trace-free random-masking objective that adapts variance-reduction tools, combining sequence-level importance sampling, deterministic quadrature over masking levels under a mean-preserving, monotonically decreasing per-block mask schedule that we introduce. On SDAR-4B, SLIM-RL matches TraceRL's best MATH500 accuracy on only 0.46x its training samples at block size 16, improving over TraceRL by 6.32% on MATH500 and 11.05% on GSM8K under matched dynamic sampling. At block size 4, the 4B SLIM-RL surpasses the larger LLaDA-8B and Dream-7B dLLMs on math, exceeding LLaDA-8B by 10.76% on MATH500 while staying below the autoregressive Qwen2.5-7B. On code, it improves over TraceRL by 4.20% on MBPP and 3.65% on HumanEval. The tau-budget decoder transfers training-free across LLaDA, Dream, and SDAR. The source code is available at https://github.com/laolaorkkkkk/SLIM-RL .
We develop a framework for analyzing the learning dynamics of $\ell_2$-adversarial training of single-index models on Gaussian mixtures in the high-dimensional limit under streaming stochastic gradient descent (SGD). We derive deterministic equivalents for a broad class of statistics of the SGD iterates, including the adversarial risk and distance to adversarial optimality, in terms of the solution to a system of ODEs. We use them to study two idealized learning rate schedules: the Polyak stepsize and exact line search. In the case of $\ell_2$-adversarial least squares with a single class, we show that, unlike noiseless standard least squares, no constant learning rate guarantees monotone descent of SGD towards a minimizer of the adversarial risk. We identify anisotropic covariance and a mismatch in ridge parameters as the main sources of suboptimality of exact line search relative to the Polyak stepsize. We also introduce a stochastic differential equation (SDE), called adversarial homogenized SGD, that captures the evolution of statistics of the iterates of SGD. For $\ell_2$-adversarial least squares, using this SDE, we show the evolution of the risk is equivalent, up to dimension-free constants, to that of SGD on standard least squares with an adaptive learning rate and adaptive $\ell_2$-regularization. When the dynamics converge, the limiting adversarial risk and SGD iterate are determined by a fixed-point equation, with the limiting iterate being equivalent to the solution of a ridge regression problem whose regularization parameter is the limiting effective regularization of SGD.
Long-horizon multivariate time series forecasting (LTSF) remains challenging due to non-stationarity, regime shifts, and error accumulation. The Variability-Aware Recursive Neural Network (VARNN) is designed to track such variability by maintaining a residual-memory state driven by one-step prediction errors. However, its original formulation is limited to one-step sequence regression and does not directly support multi-step forecasting. In this work, we extend VARNN to long-horizon forecasting and introduce StateFlow, a recurrent forecasting framework that uses VARNN as a dual-state recurrent backbone to capture two complementary signals from the lookback sequence: a hidden-state trajectory representing primary temporal dynamics, including trend, seasonality, level changes, and recurring patterns, and a residual-memory trajectory representing structured local prediction deviations, driven from a nonlinear recurrent transformation of errors between one-step base predictions and observed values. A chunk-based decoder separately summarizes these trajectories and maps them to the future horizon for direct multi-step forecasting. We further employ a two-stage optimization strategy that first trains the VARNN encoder through a one-step base prediction objective to optimize the internal representations over the lookback sequence, and then trains a horizon-specific decoder for direct multi-step forecasting. Experiments on standard LTSF benchmarks show that StateFlow achieves competitive performance against strong linear, recurrent, convolutional, and Transformer-based baselines while preserving linear recurrent encoding and a compact model design.
Many scientific systems exhibit uncertainty from stochastic forcing, unresolved degrees of freedom, or imperfect observations, making reliable surrogate forecasting fundamentally distributional rather than pointwise. For such systems, deterministic neural surrogates fail to capture statistical measures and forecast uncertainty. We introduce TRIE, an evaluation framework for stochastic PDE surrogates that asks whether models reproduce invariant measures, provide trustworthy predictive uncertainty, and scale to efficient probabilistic generation. We demonstrate TRIE on two stationary chaotic spatially extended SPDEs, stochastic Kuramoto--Sivashinsky and stochastic Kolmogorov flow, across 11 parameter values. Our evaluation shows that standard pointwise-trained neural surrogates can produce plausible short rollouts while failing to match long-time statistical structure. Approximate uncertainty methods such as Monte Carlo dropout and heteroscedastic Gaussian likelihoods produce stochastic forecasts, but are often miscalibrated and overconfident under temporal and spatial uncertainty diagnostics. Across these criteria, generative models provide the most consistent performance, accurately capturing invariant measure statistics and achieving the lowest CRPS in all reported probabilistic settings. Finally, we show that latent generative models with automatic dimension discovery retain much of this statistical fidelity while reducing Kolmogorov inference time by roughly $12\times$. We release our code and data at https://github.com/scailab/TRIE-SPDE-Bench to support reproducible evaluation of stochastic PDE forecasting models.
Collaborative-perception enables multi-robot systems to enhance situational awareness by sharing perceptual information. Existing collaborative-perception systems face an inherent trade-off between communication bandwidth requirements and perception accuracy, where methods that exchange more information achieve better perception results at the cost of increased communication overhead. However, real-world communication networks impose bandwidth constraints that require minimizing communication overhead without sacrificing perception performance. To address this challenge, we propose HydraCollab, an adaptive collaborative-perception framework that (i) selectively transmits the most informative sensor features and (ii) dynamically employs collaboration strategies (intermediate or late) based on spatial confidence maps. Extensive evaluations on the V2X-R, V2X-Radar and UAV3D-mini datasets demonstrate that HydraCollab achieves the best overall trade-off between accuracy and communication cost among existing collaborative-perception methods. Relative to SOTA Where2comm, HydraCollab uses only 41% of the bandwidth on V2X-R and 26% on V2X-Radar while improving performance by 0.78% and 0.75% respectively. Our code and models are available at https://github.com/AICPS/HydraCollab.
Recent advances in reinforcement learning have produced superhuman agents across a wide range of competitive games. As a byproduct, researchers have begun studying how these agents play, extracting behavioral representations, analyzing decision structure, and modeling the latent geometry of expert performance. However, this growing body of work has overwhelmingly focused on defeating human players rather than providing feedback, leaving a critical gap in creating model solutions to improve human players. Unlike chess and Go, where AI has become integral to player training, real-time strategy (RTS) games lack principled frameworks for translating expert knowledge into actionable feedback. We introduce Latent Maps of Performance, a framework for counterfactual path generation. We focus on StarCraft~II data to model player improvement as an algorithmic recourse within a learned representation space. As inspiration for our work, we have looked at the championship model used in sports science. We trained a Guided Variational Autoencoder model on 23,305 professional tournament replays, enabling counterfactual traversal between losing and winning gameplay profiles. To fulfill our goal, we have devised and verified four traversal strategies on out-of-distribution (OOD) data randomly sampled from a dataset of amateur replays, namely linear interpolation, iterative optimal transport, density-regularized gradient ascent, and neural flow matching, each designed to generate multi-step improvement trajectories that remain grounded in observed expert behavior while moving a player's profile toward winning configurations. Feedback is extracted at multiple granularities to support players at different stages of improvement. Finally, we conclude that there is a trade-off between the path-finding methods we employ and hope that future research will focus on developing model solutions for human improvement.
Khipus -- knotted cord devices -- were the primary recording medium of the Inka Empire (c. 1400-1532 CE), yet their system remains undeciphered. We present a reproducible machine-learning pipeline applied to the Open Khipu Repository (OKR), a public database of 619 khipus comprising 54,403 cords and 110,677 knots. We engineer 27 structural features per khipu and apply (i) unsupervised clustering via UMAP and HDBSCAN, recovering three structurally distinct groups (silhouette = 0.769); (ii) supervised provenance classification via gradient boosting, reaching F1 = 0.86 for the Inka Late Horizon imperial style; and (iii) SHAP-based interpretability, which identifies cord twist direction as the dominant structural discriminator of imperial khipus. We further report two findings of methodological interest. First, one cluster is dominated not by a geographic region but by nineteenth-century European museum collections, indicating that colonial acquisition and recording practices are structurally encoded in the corpus. Second, we provide an independent computational verification of the recto/verso (moiety) structure of the six Santa Valley khipus reported by Medrano and Urton (2018), reproducing both the aggregate attachment ratio and the identification of the single mixed specimen--using only the public OKR database, without physical access to the objects. We additionally report a negative result: knot-type sequence order, encoded as n-grams, adds no provenance signal beyond aggregate features. All code and data are openly available.
We present a black-box model-stealing attack that recovers private vision-tokenizer configurations of deployed vision-language models (VLMs), including the visual patch size and input preprocessing pipeline. The key idea is a task-level side channel induced by ViT-style patchification: when a synthetic grid image is aligned with the hidden patch grid, boundary cues are erased at tokenization, causing periodic accuracy drop. By sweeping the grid cell size and measuring these collapses, we infer the patch size; by introducing padding and a consistency-check test, we further identify whether preprocessing is dynamic- or fixed-resolution and recover the target resize resolution. Across open-source Qwen-VL variants and proprietary models including GPT and Claude, we reliably recover tokenizer-related parameters. Finally, we show that such leakage enables preprocessing-aware transfer attacks and model-targeted adversarial manipulation.
Federated learning is bandwidth-bound on two orthogonal axes: model size, which limits how often parameter-averaging methods can afford to merge, and class count, which makes per-probe soft-label distillation prohibitive at large vocabularies. Both ceilings tighten as modern systems scale. We collapse the class-count axis to $\lceil \log_2 C \rceil$ bits per probe by transmitting only each peer's $\arg\max$ class index, where $C$ is the number of output classes. The resulting protocol, TallyTrain, is not merely compressed: under non-IID training it can be preferable to soft-label distillation, because under-trained peers are confidently wrong and majority voting filters this noise where soft-label averaging amplifies it. Across standard benchmarks, TallyTrain matches or beats soft-label distillation at up to three orders of magnitude less communication. We also relax the model-size axis: we compose the cheap hard-label consensus with sparse parameter merges to obtain a bandwidth-bridge variant, which Pareto-dominates every tested operating point of the standard FedAvg, FedProx and FedDF baselines.
Text embeddings are standard for semantic similarity tasks, yet their evaluation remains an open challenge. Current benchmarks are static, cover only a limited set of languages, are often domain-specific, susceptible to overfitting, and poorly representative of low-resource languages. To address these limitations, we introduce ALEE, a framework that extends Sentence Smith (Li et al., 2025) to the cross-lingual and paragraph level. ALEE uses Abstract Meaning Representations (AMR) to generate English minimal pairs with controlled, fine-grained semantic shifts, which are paired with translations in target languages. This approach enables targeted diagnostics for models in any language with English parallel data. We conduct a large-scale empirical study across a diverse set of embedding models and 275+ languages spanning three parallel datasets. On ALEE, performance varies substantially across languages, text lengths, and linguistic phenomena, exposing persistent gaps in cross-lingual semantic representation that track language prevalence in training resources and subword tokenization. We release ALEE at https://github.com/Andrian0s/any-lang-embed-eval
Thermodynamic computing devices based on the Ising model show great promise for low-power AI inference and edge computing, but scalable methods for training large models for such hardware remain limited. Prior theory shows that the time-averaged behavior of high-temperature Gibbs-sampled Ising systems can implement feed-forward neural inference. We turn this theoretical correspondence into a scalable and purely backpropagation-based algorithm for training deep convolutional networks for thermodynamic inference on Ising machine hardware. Our image classification models achieve accuracies of 94.9% on CIFAR-10 and 76.0% on CIFAR-100 under binary Gibbs sampling. We then develop and experimentally validate a mathematical theory relating inference cost to accuracy and controlling autocorrelation times. Subsequently, we calculate asymptotic results showing that inference cost is bounded by a well-controlled tradeoff with performance and exhibit algorithms for computing optimal inference schedules. Finally, we discuss implications for hardware development and the future of high-temperature thermodynamic AI models.
Reinforcement learning with verifiable rewards can in principle train calibrated probabilistic forecasters, since a proper scoring rule such as the Brier score is computed from outcomes alone and is minimized in expectation by the true probability. In practice it degrades calibration, and existing remedies address epistemic uncertainty, where a model's confidence accompanies a verifiably correct or incorrect answer. We study aleatoric forecasting, where the forecast itself is the output and the label is one stochastic outcome, taking NFL in-game win probability as a testbed with the betting market as a reference. Rewarding the realized per-play outcome fails, because the single outcome is a noisy target and the policy gradient corrupts the chain of thought. We introduce a verifiable, label-free reward, a state-conditioned empirical win rate estimated from past outcomes, that removes the label noise, and we keep the gradient off the reasoning, by direct prediction or a gradient mask, so it cannot be corrupted. Trained with this reward alone, without human labels or supervised fine-tuning, a 7B model reaches the calibration of the betting market by direct prediction and is better calibrated than a zero-shot frontier model. That frontier model and a tabular estimator reach the same Brier score as this model, identifying the market's small remaining edge as live in-game information beyond their shared inputs. Masking the gradient, rather than dropping the chain of thought, preserves reasoning from which the forecast follows, which ordinary chain-of-thought training corrupts.
Parameter-efficient fine-tuning (PEFT) reparameterizes weight updates in a fixed basis: low-rank adapters operate in the spatial domain, while a recent line of spectral methods operates in a fixed Fourier domain. We argue that the choice of domain is itself a design degree of freedom that should be learned, and that no single basis is optimal across tasks, layers, or tokens. We introduce Fractional-Fourier Mixture of Experts, a mixture-of-experts adapter in which every expert carries a learnable fractional-Fourier order that continuously interpolates between the spatial domain (recovering vanilla LoRA) and the Fourier domain (recovering a spectral adapter). Routing tokens through experts that occupy different points on this spatial-spectral continuum lets the model place each low-rank update in the domain where it is most compact, and -- because fractional-Fourier operators of different orders are mutually incoherent -- makes the experts naturally decorrelated, which reduces interference and improves multi-task composition. The order is a single scalar per expert, trained with a separate optimizer, and the transform is computed with an $\mathcal{O}(d\log d)$ chirp--FFT surrogate, so Fractional-Fourier Mixture of Experts adds negligible cost over standard MoE-LoRA. Across commonsense, mathematical, code, and knowledge benchmarks on LLaMA-3.1-8B and Qwen2.5-7B, Fractional-Fourier Mixture of Experts improves over strong MoE-LoRA and spectral baselines -- including FlyLoRA, FourierMoE, and HMoRA -- while keeping the active-parameter budget small, and analysis shows that the learned orders specialize by task and layer in interpretable ways.
Knowledge-Based Visual Question Answering (KB-VQA) aims to evaluate whether Visual Language Models (VLMs) can retrieve, ground, and reason over external structured knowledge beyond visual evidence. In practice, answer accuracy is widely adopted as the primary evaluation metric, implicitly treating correctness as a proxy for knowledge-grounded reasoning. However, for existing KB-VQA benchmarks, this proxy relies on critical assumptions that are often overlooked and rendered unreliable by benchmark issues: annotated answer must be derivable from the associated knowledge base, question must be well-posed with sufficient constraints, and visual setting must meaningfully require grounded disambiguation. In this work, we show that these assumptions are systematically violated in existing KB-VQA benchmarks. Our audit reveals substantial instances with missing or contradicted answers and underspecified questions that render accuracy a misleading metric. Furthermore, we find that existing datasets rely on visually trivial, single-entity scenes that bypass the need for sophisticated visual-to-knowledge mapping. We demonstrate that even with controlled architectures, these flaws lead to distorted model rankings and overestimations of reasoning capabilities. To address this, we introduce (1) a principled audit-and-repair protocol that restores answer derivability and question clarity, and (2) a controlled multi-entity augmentation protocol that introduces visual ambiguity to challenge initial retrieval and grounded reasoning. Re-evaluation under corrected and augmented settings yields markedly different performance trends. Our findings call for rethinking evaluation protocols and designing more interaction-aware KB-VQA benchmarks that prioritize verifiable reasoning over simple matching.
Hallucination remains one of the central obstacles to deploying medical LLMs. Yet, even when hallucination can be detected, it is still unclear whether the internal representations associated with it can be used for control rather than detection alone. Using four open-source models across a suite of medical question-answering datasets, we show that a simple, carefully conditioned probe can reliably detect hallucination, with AUROC scores between 0.77 and 0.86 in our case. We further show that this signal is distributed and redundant rather than narrowly localized. Systematically selected neurons outperform random neurons only at very small subset sizes, whereas random subsets of a few hundred neurons recover nearly the full signal, and low-dimensional random projections preserve most of the detection performance. Beyond detection, we test whether this representation is causally actionable. Across 16 model--dataset combinations, our results reveal a sharp gap between decodability and controllability. The same internal structure that makes hallucination easy to detect does not translate into reliable neuron-level control. These findings show that medical hallucination seems to be readily visible in internal activations, but not easily corrected by steering the neurons most associated with it. More broadly, our results suggest that hallucination mitigation is not simply a matter of identifying the right neurons, and point to a deeper separation between what representations reveal and what they allow us to change.
We study runtime human oversight of an AI agent when private information runs in both directions: the human privately knows her reward function, while the AI privately knows the quality of the action it proposes. This is the kind of asymmetry that arises naturally when an autonomous robot or software agent has inspected a situation its human supervisor cannot directly assess. Building on Cooperative Inverse Reinforcement Learning (CIRL) and the Oversight Game, we introduce a contextual-bandit team game with two-sided asymmetric information and a play/ask/trust/oversee interface. The bandit structure removes physical state transitions and thereby yields exact one-shot characterizations that would remain conjectural in the full POMDP setting, though the common belief remains a dynamically controlled state across rounds. We give two one-shot characterizations, a team optimum and a behaviorally natural myopic rule, whose gap is a slab of avoidable harm: a region in which the AI privately knows the proposed action is harmful and shutdown would help, yet a myopic human, trusting her prior, declines to oversee. We show this gap is the price of non-credible oversight communication, and give a partial analysis of how it resolves dynamically over repeated rounds through passive learning and active signaling with a one-period-lagged oversight response.
Evolutionary neural architecture design for multivariate time-series forecasting remains underexplored, with most approaches relying on fixed Transformer architectures despite substantial variation across tasks and forecasting settings. This paper introduces an evolutionary neural architecture search framework for discovering task-adaptive Transformer-like models for time-series forecasting (EVOTS). Architectures are encoded using a modular genome representation that enables flexible composition of attention, feed-forward, and projection components, while a repair mechanism enforces structural validity throughout the evolutionary process. This formulation allows effective exploration of a diverse architecture space without relying on hand-crafted design rules. The proposed approach is evaluated on four benchmark datasets from the ETT family (ETTh1, ETTh2, ETTm1, and ETTm2) under multiple forecasting settings, including univariate-to-univariate, multivariate-to-univariate, and multivariate-to-multivariate prediction, with horizons of 96, 192, 336, and 720. In the multivariate-to-multivariate setting, the evolved architectures achieve competitive and, in several cases, improved mean squared error relative to a strong Transformer-based baseline. Additional analyses examine performance differences across forecasting settings and report wall-clock training time to provide a coarse indication of computational cost. Overall, the results demonstrate that evolutionary search can effectively discover flexible and high-performing Transformer-like architectures for multivariate time-series forecasting within practical runtime constraints.
Three of the most popular methods for training language models to reason look like three different tricks. They are not. All three adjust a single number: standard deviation, reflecting how much a prompt's sampled answers disagree. When such a model is trained, it answers each problem many times, and an automatic checker marks every answer right or wrong. The standard deviation of those marks measures the disagreement: largest when the answers split evenly between right and wrong, and zero when they all agree. Group Relative Policy Optimization (GRPO) divides by this number, GRPO Done Right (Dr. GRPO) drops the division, and Decoupled Clip and Dynamic Sampling Policy Optimization (DAPO) discards the groups where it is zero. Each is presented as its own fix, yet this paper proves they are three settings of one dial. That dial is not cosmetic: for right-or-wrong rewards, the disagreement is exactly the size of the training update, the group-standard-deviation identity. A split group teaches the most, while a unanimous group teaches nothing and falls silent. The same result says which problems deserve the most weight and how many tries each one needs. This paper confirms the intuition on a large real difficulty dataset (Big-Math) and in a controlled training run. What looks like a harmless normalization step is the dial that decides where learning happens and how strongly.
Rare disease differential diagnosis is a critical yet arduous clinical task, requiring physicians to identify precise phenotypes from complex, unstructured patient symptoms and execute intricate reasoning within a vast search space. However, existing AI approaches typically rely on pipeline-based phenotype extraction or retrieval-augmented generation, which suffer from critical information loss due to predefined ontologies, retrieval bottlenecks, and a lack of diagnostic logic. To address these challenges, we introduce RareDxR1, an end-to-end reasoning-centric large language model designed for open-domain rare disease diagnosis directly from unstructured clinical notes. We design a progressive end-to-end training framework by synergizing knowledge internalization with autonomous evolutionary learning, thereby bypassing reliance on structured phenotypes and closed-set decision-making. To overcome the limitations of RAG and phenotype restriction, we enabled the deep internalization of fragmented rare-disease knowledge directly into the model's parameters. Moreover, to bridge the gap between model generation and expert reasoning, we propose Reflection-Enhanced Reasoning Sampling (RERS), a strategy that synthesizes expert-level diagnostic trajectories by learning from failures without human annotation. Additionally, we propose a dual-level curriculum reinforcement learning approach for gradually mastering rare disease diagnosis. Experimental results demonstrate that RareDxR1 achieves state-of-the-art accuracy across different benchmarks, marking a significant breakthrough in open-domain rare disease diagnosis. Our code and dataset will be publicly available.
Active learning (AL) performance is known to be budget-dependent, yet regimes are typically defined by heuristic label counts that fail to generalize across datasets or architectures. We characterize AL dynamics by reframing budget regimes as shifts in the dominant generalization mechanism. By reinterpreting PAC-style risk components as dynamic interacting terms, we prove that dominance shifts are structurally unavoidable, creating a moving bottleneck for generalization. We operationalize this using measurable proxies and a segmented regression procedure to identify a tripartite taxonomy: data-driven, transition, and model-driven phases. Our framework explains the long-standing observation that representativeness, coverage, and uncertainty strategies excel at different stages. Experiments across natural and medical imaging show that AL efficiency depends on the alignment between the strategy's inductive bias and the active bottleneck. Moreover, self-supervised representation shift transitions earlier along the labeling trajectory, highlighting the role of representation quality in shaping AL dynamics. Overall, this work provides a unified framework for the next generation of transition-aware AL algorithms.
Online hate speech has been linked to a global rise in violence against minorities, including incidents such as mass shootings, lynchings, and ethnic cleansing. Societies grappling with this issue, particularly when hate speech targets specific groups based on religion, race, ethnicity, culture, nationality, or migration status, face the challenge of balancing freedom of expression with the need for effective content moderation on widely used online platforms. In response to this challenge, we introduce a comprehensive hate speech dataset covering five distinct topics in Turkish: refugees, the Israel-Palestine conflict, anti-Greek sentiment in Turkey, ethnic or religious communities (Alevis, Armenians, Arabs, Jews, and Kurds), and LGBTI+, alongside one topic in Arabic (refugees). In addition, we develop state-of-the-art BERT-based models to address multiple dimensions of hate speech analysis, including hate category classification, hate intensity prediction, target identification, and hate speech span detection, enabling a comprehensive understanding of hateful content in online discourse.
As computing education expands beyond traditional programming into operational domains such as systems administration and command-line environments, existing pedagogical frameworks struggle to capture a dimension that is critical in these contexts: the real-world consequences of learner actions. Existing cognitive taxonomies classify learning objectives by mental operations but do not account for system impact, leaving a critical gap in command-line education where conceptually simple commands can have severe consequences. This work presents CogTax, a four-level cognitive taxonomy that integrates two dimensions: cognitive complexity, derived from Bloom's Revised Taxonomy, and operational impact, which distinguishes observational, reversible, structural, and administrative operations. The four progressive levels range from safe read-only inspection to advanced system management requiring integration of multiple abstract models. Then, the taxonomy level is defined as the maximum of these dimensions, ensuring that both conceptual understanding and operational awareness are addressed. CogTax gives instructors a principled framework for sequencing course material and calibrating assessment difficulty, and gives students an explicit reference for self-assessment and gap identification. To demonstrate that taxonomy levels are automatically assignable, making the framework scalable without manual expert annotation, a classifier that combines syntactic representations derived from abstract syntax trees with semantic embeddings is trained. Evaluated on 585 expert-annotated Linux/bash commands, this combined approach achieves 89% accuracy, outperforming either representation alone, and demonstrates cross-language extensibility through structural equivalences across command languages.
The cost of human expert evaluation is a principal bottleneck to deploying language models in specialized, high-stakes domains. This is particularly acute for Arabic sociolinguistic knowledge: credible grading requires not only linguistic fluency but deep cultural familiarity that cannot be approximated by surface-level metrics. We address this with a cross-evaluation framework instantiated on two underrepresented Arabic dialect communities: Egyptian and Iraqi Arabic. We contribute 103 validated prompt-rubric pairs (70 Egyptian, 33 Iraqi; 53 Cultural, 50 Linguistic), authored and graded by native-speaker SMEs using penalty-weighted rubrics distinguishing positive content requirements from answer-specific negative error criteria. Three frontier LLMs serve as target models (graded by human SMEs across 302 unique prompt-response pairs), while five frontier LLMs serve as automated judges enforcing a provider-level self-evaluation guard. A dual-metric scheme combining Mean Absolute Deviation (MAD) with Signed Mean Error separates directional grading bias from symmetric noise. Across 1,307 judge evaluations: GPT-5.4 is the most reliable judge (MADj = 10.21 pp, Signed Error = -1.12%); four of five judges show systematic leniency (+2.01% to +6.56%); Cultural tasks are harder to grade than Linguistic tasks for all judges (MAD gap 1.83-4.78 pp); and models substantially outperform on Egyptian prompts compared to Iraqi prompts. However, given leniency differences between Iraqi and Egyptian SMEs, we cannot solely attribute this gap to model knowledge. We therefore emphasize findings that do not assume identical leniency across human graders. Across all samples, implicit cultural reasoning -- requiring models to simulate native-speaker judgment rather than rely on lexical verification -- emerges as the primary failure mode for automated grading across all judge models.
Survival analysis models time-to-event data, but in clinical settings training data are costly and scarce: events accrue over years of follow-up, cohorts are small, and privacy regulations restrict sharing across institutions. Tabular generative models promise augmentation and privacy-preserving cohort sharing, yet are themselves data-hungry -- on the small cohorts typical of survival analysis, a single generator rarely characterizes the population well enough for downstream models trained on its output to match real-data performance. FoGS (Filtered Mixture-of-Generators for Survival analysis) reframes synthetic-data construction as sample selection rather than generation. A candidate pool is drawn from four architecturally distinct tabular generators, and each sample is scored by an ensemble of seven survival models trained on real data, using proper scoring rules as a per-sample plausibility proxy. A two-level pipeline optimizes, in its outer loop, a selection policy -- generator quotas, scorer weights, a random complement, and stratified balancing on event time and censoring -- against held-out downstream performance, while an inner loop tunes the downstream model (XGBoost-Cox). On 16 public datasets under train-on-synthetic, test-on-real (C-index and IBS, $0$--$100$ scale), FoGS yields mean improvements of $+2.17$ in C-index and $+0.67$ in IBS, improving both metrics on 9 of 16 datasets and at least one on 13 (one-sided Wilcoxon $p=0.039$ and $p=0.035$). It matches or exceeds real-data training on most cohorts, with no significant change in nearest-neighbour privacy margin relative to unfiltered sampling. Sample filtering over a heterogeneous generator pool is thus a viable substitute for real-data training in privacy-restricted clinical settings.
Emotional bonds between humans and AI companions are growing, and the question of whether a person may marry an AI system will soon move from speculative fiction into law. This chapter examines whether the autonomy-centered logic that has expanded marital choice among human beings can justify extending marital status to superintelligent companions. Following a scenario-envisioning exercise informed by anticipatory ethics, I argue that granting such status leads to socially unjust outcomes, even under the generous assumption of reliable superintelligence. Marriage as a socio-legal institution does more than ratify private agreement; it creates networks of mutual obligation, joins families, and makes each partner vulnerable to the other. A relationship sustained by corporate policy and continued payments is a subscription rather than a bond tested by time. Discussing wholesale marital status is therefore the wrong frame. Law should carve out targeted rights and protections for pressing needs arising from intimate human-AI relationships.
Background. Labeled data for security classification is scarce. Semi-supervised learning (SSL) propagates labels from a small labeled pool to larger unlabeled pools. Yet security applications often use SSL as a black box: default parameters, a fixed classifier, and no handling of pseudo-label-induced class imbalance. Aims. Recent work reports sizeable gains from optimizing SSL pipelines via joint search, AutoML, or per-component tuning. These gains are hard to attribute: they may reflect useful SSL-classifier interactions, or mostly from simply tuning the downstream classifier. We disentangle these effects for binary tabular security data with classical SSL and tree-based classifiers. Method. We build SemiScope as an analysis instrument, not a deployment recommendation. It uses Bayesian Optimization to jointly tune SSL settings, confidence filtering, oversampling, and the classifier. The key control, Tuned-Clf, fixes SSL to defaults but gets the same 100-trial classifier budget and validation-set threshold tuning as SemiScope. At 10% labels, we compare them with paired TOST using a +/-1.0 g-measure smallest effect of interest. Results. SemiScope beats every default SSL baseline on all five datasets, improving over the strongest by 0.7-12.7 points. Under the equal-budget control, Tuned-Clf is statistically equivalent to the full pipeline on 4 of 5 datasets; Phishing is inconclusive. Classifier HPO alone recovers a median 86% of SemiScope's gain over Default Self-Training (ST) + Random Forest (RF). Conclusions. The reusable contribution is the decomposition protocol. A simpler recipe suffices: use Self-Training, tune the classifier with Bayesian Optimization, and tune the decision threshold on validation data. It reaches within 1 g-measure of Supervised RF at 20-30% labels on four datasets and 40% on Drebin, at the same or lower label rate than Default ST + RF on every dataset.
Large language models increasingly generate C++, a memory-unsafe language where a single overlooked violation can become an exploitable bug. Yet most security evaluations of AI-generated code rely on static analysis alone, which flags warnings without confirming runtime violations or reasoning about untested paths. We ask whether AI-generated C++ is measurably less safe than human-written code, and whether common verification tools agree on the risk. We introduce VULBENCH-CPP, a benchmark of 8,918 C++ programs from three open-weight LLMs (Gemma 3 27B IT, LLaMA 3.3 70B Instruct, Qwen 2.5 Coder 32B Instruct) and human authors across 851 competitive-programming tasks. Each program is annotated by four verification tiers: functional testing, static analysis (cppcheck, clang-tidy), dynamic analysis (ASan/UBSan), and bounded model checking (ESBMC). Accounting for the correlation among solutions to a shared task, we find that AI-generated code is roughly twice as likely as human code to trigger a confirmed runtime violation, even after controlling for code length and test pass-rate. Under static analysis the two look equally safe, but this is misleading: the apparent similarity reflects code length rather than real safety, and the tiers detect largely different classes of violation, so no single tier is sufficient. The gap is consistent across independent generations.
Generative models have emerged as scalable surrogates for physical simulation, yet they offer no guarantee that their outputs respect the conservation laws, boundary conditions, and nonlinear invariants that govern the underlying physics. Constrained sampling closes this gap, enforcing such constraints exactly at inference time without retraining, but at a computational cost: projection, correction, and trajectory-optimization steps are repeated during sampling, with these steps becoming expensive for nonlinear constraints. Standard ML frameworks exacerbate this: their dense tensor algebra and limited sparse solver composability obscure the structure that physical constraints naturally induce, making efficient batched nonlinear optimization difficult to realize in practice. We address this bottleneck by exploiting the structure that sample-wise batching and local PDE couplings induce in the projection subproblems -- namely, block-sparse Jacobian and KKT systems -- exposing this structure using ExaModels.jl and solving the resulting sparse nonlinear programs with MadNLP.jl and GPU sparse factorization. Applied to Physics-Constrained Flow Matching (PCFM), on PDE benchmarks with linear, nonlinear, one-dimensional, and two-dimensional constraints, this approach accelerates nonlinear constraint projection while maintaining constraint satisfaction. These results show that sparse GPU nonlinear optimization is a practical foundation for constrained generative sampling in scientific machine learning.
Urban-scale Visual Place Recognition (VPR) aims to identify the geographic location of a query image by matching it against a geo-tagged database. While recent methods achieve impressive performance, they overlook a serious long-tailed problem hidden in urban-scale datasets, which biases the model towards locations with abundant images and ignores less-visited areas, causing models to systematically favor frequently photographed locations while failing in sparsely covered areas. In this paper, we systematically characterize this imbalance challenge and propose Distribution-Aware Place Recognition (DAPR), a model-agnostic plug-in framework that rebalances gradient contributions across head and tail classes. Additionally, within classification-retrieval pipelines, DAPR applies a multi-scale distance search mechanism to compute per-class distributional compactness, providing complementary gains at the retrieval stage. On the large-scale SF-XL benchmark, our framework outperforms the previous classification-retrieval baseline by 18.3% on test set v1, and 6.7% on test set v2. As a plug-in module, it achieves consistent improvements across representative VPR methods on SF-XL, MSLS, and Pitts30k, demonstrating broad generalizability across different methods and benchmarks.
Mechanistic interpretability has produced a rich inventory of component-level analyses that characterise what neural-network components encode and how they interact. Their outputs, however, are not easily reusable: selectivity tables, circuit diagrams, and feature lists remain locked in per-study notebooks - non-composable, not queryable in natural language, and not directly actionable for downstream audit or intervention. We study the representation layer that sits between these analyses and downstream use as a bottleneck that can be evaluated independently, and introduce Manifestation Units, a typed tuple protocol (E, S, R, D, G) extended with attention-head primitives (T) for transformer architectures, organising per-component statistics into structured fields populated automatically and queried through hybrid retrieval. Instantiated across generative vision (beta-VAE), discriminative vision (CNN), and language (GPT-2), the protocol supports two findings: typed structure substantially outperforms unstructured baselines on retrieval, and CNN filters retrieved by the schema satisfy causal sufficiency and necessity criteria under matched-budget controls. The schema absorbs attention-head primitives without modification, set-recovers known IOI circuit members under retrieval-budget-matched controls, and reveals an irreducible two-field core (S+R) with remaining fields either redundant or actively interfering. We present this as schema infrastructure for mechanistic interpretability rather than frontier-scale validation.
Language models have changed from unreliable text generators to highly-capable large models with trillions of parameters. Capability increases come hand-in-hand with increases in scale, making understanding the internal representations of models more challenging. Since millions of users increasing rely on language models to interact with external tools or make decisions in medium or high-stakes scenarios, we need to establish control over model behavior and know when to trust model outputs. In this paper, we discuss our contributions on harnessing the latent spaces by proposing steering vectors for control and developing latent space-based model calibrators for trust. Together, our contributions help demystify the latent spaces of language models and offer new insights into how to harness model internals to build more trustworthy language technology.
Cohesive Zone Models (CZMs) are widely used to simulate interface fracture, delamination, adhesive failure, and fiber--matrix debonding in aerospace composite structures. In implicit quasi-static finite element analyses, cohesive softening may introduce negative interface tangents, solution jumps, and Newton-basin mismatch, so the previous converged state can become a poor initial guess for the next increment. This may lead to stagnation, wrong-branch convergence, or repeated step cuts. Existing remedies, including viscous regularization, path following, dynamic relaxation, and manual Newton--Raphson (NR) modification, either alter the effective response, increase cost, or rely on hand-crafted interface rules. This work proposes an Interface-Aware Neural Newton Preconditioner (IA-NNP) for difficult CZM increments. IA-NNP recasts manual NR modification as rule-based interface lifting and generalizes it into a learned, state-dependent interface correction. The method acts only on active interface variables and preserves the original traction--separation law, residual assembly, tangent evaluation, history update, and dissipation checks. Two realizations are developed: IA-NNP-Init for learned initial-guess lifting and IA-NNP-NL for iteration-level nonlinear right preconditioning. Interface graph features encode opening, traction, tangent, damage/history variables, mode mixity, residuals, and neighboring states. The correction is bounded, confidence-gated, and accepted only through the original CZM Newton solve. A root-equivalence property shows that IA-NNP changes the path to convergence but not the discrete CZM solution set. Tests on horizontal, circular, two-interface, and active-front benchmarks show improved difficult-increment convergence, better branch recovery, and fewer failures than standard NR and manual NR modification, while preserving the force--displacement response.
This work presents Dual-Embedding Watermarking (DEW), a semantic watermarking scheme for large language models (LLMs) that leverages contextual and token-level embeddings to enhance robustness against paraphrasing and translation. DEW utilizes a signal-processing methodology, applying algebraic vector-space operations to token and context embeddings to derive a watermark signal that degrades gracefully under semantic shifts. The method obfuscates the watermark by projecting embedding vectors through pseudo-random matrices seeded with a secret key. Relevant distributions derived from the underlying algebra are evaluated and employed for statistical testing and benchmarking of DEW. Experimental results across multiple LLMs indicate that DEW improves post-paraphrase detection while maintaining competitive text quality, and remains detectable after translation, even when prior semantic watermarks degrade significantly. These findings position DEW as a practical and robust solution for safeguarding LLM-generated text and addressing critical issues in responsible AI deployment.
Any-angle path planning extends traditional graph-based path planning by allowing movement between any pair of vertices, rather than being restricted by predefined edges. It can find straighter and shorter paths in continuous space with graphs, making it particularly suitable for navigation in open areas such as airspaces, warehouses, and oceans. Many any-angle path-planning algorithms have been proposed, but only a few can guarantee optimal solutions, especially in the presence of dynamic obstacles. To address this challenge, this article focuses on optimal any-angle path planning on grids and introduces two general techniques that accelerate computation while preserving optimality in both static and dynamic environments: 1) elliptical forward expansion, which leverages ellipse-based neighborhoods to restrict the search space, and 2) field of view, which replaces traditional line-of-sight methods to speed up visibility checks. To integrate these two techniques, inverted and forward scanning are introduced. Inverted scanning establishes visual connections from open nodes, whereas forward scanning initiates scans from closed nodes. Building on the proposed techniques, Zeta* and Zeta*-SIPP are developed for static and dynamic environments respectively. Zeta*, when combined with forward scanning, is similar to the state-of-the-art algorithm Anya and attains comparable performance. Unlike Anya, Zeta* can be readily extended to other settings, such as dynamic environments (e.g., Zeta*-SIPP). Zeta*-SIPP, with either scanning method, is more than 20 times faster than the corresponding state-of-the-art optimal planner TO-AA-SIPP. Overall, this research identifies the key requirements for achieving optimal any-angle path planning and introduces a unified approach suitable for different environments.
As technology advances, many path-planning algorithms have been proposed for Air Traffic Management, yet their operational adoption in tactical control remains limited, revealing a misalignment between algorithmic design priorities and air traffic controllers' needs. This underscores the need for decision-support solutions that are inherently interpretable, computationally efficient, and explicitly designed for human use. Focusing on this design challenge, this study develops a conflict-free path-planning algorithm for en-route Air Traffic Control (ATC) designed to be compatible with two guiding considerations: (1) the interpretability and flexibility offered by solution-space displays, which motivate constructing an algorithm that exposes all feasible safe actions and accommodates shifting optimization goals; and (2) the decision logic controllers naturally apply when enforcing operational constraints, such as separation standards, maneuverability limits, waypoint minimization, and routing practicality. Centered on these principles, the algorithm integrates three intent-based conflict detection methods -- distance-based, time-interval-based, and zone-based -- within a solution-space framework to identify conflict-free paths in computationally efficient ways. Additionally, vertex-based and edge-based search nodes are proposed for solution space path planning (SSPP), resulting in two variants -- SSPPV and SSPPE, respectively, which are evaluated in terms of computational speed and solution quality. Empirical results show that SSPPV paired with zone-based conflict detection achieves the best performance, computing paths in 3.69 ms on average in operational-relevant scenarios based on the Delta sector of the Maastricht Upper Area Control Centre (MUAC) using a 5 nmi grid.
As Quantum Machine Learning (QML) transitions toward practical implementation, the field faces a critical architectural bottleneck that challenges the fundamental assumptions of classical statistical learning theory. In classical deep learning, increasing model capacity typically risks overfitting. However, this study advances a counter-intuitive paradigm: unstructured contemporary QML architectures suffer from a profound state of quantum underfitting, driven by the "expressivity-trainability paradox." We demonstrate that the vast Hilbert space capacity of Parameterized Quantum Circuits (PQCs)-traditionally chased as the source of quantum advantage is the direct mathematical cause of Barren Plateaus (BPs), where gradient landscapes become exponentially flat. By synthesizing recent breakthroughs in Dynamical Lie Algebras (DLAs) and Geometric QML, we establish a comprehensive framework linking the algebraic dimension of circuit generators to their optimization dynamics. Furthermore, we empirically validate this framework on a non-linear binary classification task, illuminating a uniquely quantum manifestation of the bias-variance tradeoff: while unstructured architectures achieve near-perfect training accuracy via unscalable parameterization (quantum overfitting), embedding group-theoretic geometric priors acts as a structural regularizer. By restricting the DLA growth to a polynomial regime, our symmetry-preserving approach sacrifices raw memorization capacity to guarantee scalable, gradient-rich training landscapes, offering a robust roadmap for "Trainability-by-Design" in scalable quantum neural networks.
Large Language Models (LLMs) are increasingly deployed as autonomous financial agents initialized with explicit behavioral mandates such as "preserve capital" or "avoid speculative bets" that are meant to govern every decision throughout deployment. In practice, however, as market context accumulates over long horizons, these mandates gradually lose their behavioral influence, a phenomenon we formalize as Mandate Salience Decay (MSD). To measure MSD objectively, we introduce FinPersona-Bench, a simulation benchmark in which a synthetic market decouples observable price from hidden fundamental value, enabling falsifiable evaluation across three failure modes: trading without signal in calm markets, panic-selling during crashes, and ignoring fundamental value during speculative bubbles. Evaluating 18 leading frontier and open-source LLMs, each assigned one of three behavioral profiles ranging from strict capital preservation to aggressive growth, shows that MSD compounds over time and is model-dependent. In crash scenarios, the behavioral gap between static agents and those receiving periodic mandate re-grounding grows 4.4x from the first to the final quarter of the simulation. The effects of mandate re-grounding are not uniformly positive: it consistently helps conservative agents in low-signal markets but actively worsens behavior for aggressive agents in the same setting. These findings suggest that reliable long-horizon deployment requires selective, mandate-aware re-grounding based on agent profile and market regime.
This paper studies how spectral geometry emerges in quantum learning models and how it can be diagnosed with physically grounded probes. In graph-regularized quantum networks, training reorganizes the output similarity graph, increases the effective spectral dimension Delta S = +0.23, and reshapes the Laplacian spectrum. Edge-resolved two-boson interference directly probes this restructuring: the bosonic enhancement Delta P_uv correlates with the Fiedler edge split |Delta v_2| (r = -0.50), linking learned spectral partitions to interference signatures. A phase diagram shows a nonmonotonic dependence of performance on coupling strength gamma and noise delta, with graph regularization improving fidelity only in a restricted regime; hardware experiments confirm the predicted interference behavior within shot-noise uncertainty. We also analyze a hybrid quantum autoencoder and introduce Bloch-space drift as a geometric diagnostic of its latent representation. With an unsupervised benign-data threshold, the model achieves high ranking performance (ROC-AUC about 0.99) and negligible false-negative rates. Absolute Bloch drift strongly discriminates anomalies (ROC-AUC at least about 0.9), while consecutive drift is near random (ROC-AUC about 0.5), showing that detection arises from persistent state-space displacement rather than local fluctuations. Through the geometry of reduced single-qubit states and associated quantum Fisher information, these results show that learning-induced spectral organization appears as measurable quantum-state structure, establishing a unified spectral-geometric framework for diagnosing quantum learning systems with bosonic and Bloch probes.
High pass rates on established programming benchmarks such as HumanEval and LiveCodeBench do not always show whether a model can reason about algorithms. Many fixed benchmarks eventually become part of the public training ecosystem through released problem statements, editorials, and generated solutions, allowing later models to improve partly by exposure rather than by stronger algorithmic ability. We introduce ALGOBENCH, a framework that automatically builds novel algorithmic problems from known competitive-programming problems through structured constraint-shifting transformations. Each accepted ALGOBENCH variant is traceable to a source problem, but must make the original reference algorithm fail. Beyond pass@$k$, we introduce complexity-aware metrics -- including OPTT, OPTS, TRAPRATE, GAPT, and CONSENS -- to test whether a solution is not only functionally correct but also asymptotically suitable for the generated problem. Experiments across multiple LLMs and prompting strategies show that performance drops sharply on ALGOBENCH variants, retrieval can increase reuse of the old algorithm, and many correct-looking solutions fail to meet the required complexity. Error analysis shows that failures are mainly algorithmic rather than implementation-level, suggesting that ALGOBENCH evaluates adaptation beyond functional correctness.
Medical multiple-choice question answering requires parameter-efficient adaptation across heterogeneous knowledge domains and reasoning operations. A medication question, a diagnostic decision, a public-health item, and a nursing-action item may require different low-rank updates, while some recall items should preserve the base model's representation with only mild adapter intervention. We propose BiRG-LoRA, a single-adapter rank-gated LoRA method for medical question answering. BiRG-LoRA keeps one LoRA module per target layer but makes its rank dimension input-conditioned: for each question, a biaxial gate combines hidden semantic evidence with specialty/profession priors, clinical-operation priors, and their interaction to select a sparse top-$k$ subset of rank atoms. A scalar injection coefficient further controls the strength of the selected adapter update. Under a matched Qwen3-8B CMB-source protocol, BiRG-LoRA achieves the highest four-benchmark macro-average accuracy among trainable PEFT baselines and matched routing controls: 69.31% averaged over CMB, CMExam, MedQA, and MedMCQA. It improves over MoELoRA by 0.89 percentage points while using 28.1% fewer trainable parameters; a paired, benchmark-stratified bootstrap over final predictions gives a 95% confidence interval of [0.42, 1.37] for this macro-average gain. Basic controls show that BiRG-LoRA also improves over vanilla LoRA r16 and active-rank-matched LoRA r4 by 0.83 macro points, and an evaluation-time weak-axis perturbation check suggests that performance is not brittle to moderate tag noise. The results support a bounded claim: clinically structured rank allocation improves cross-benchmark medical QA under a matched single-seed protocol, while training-seed variance remains future work.
Graphical user interface (GUI) agents build on vision-language models to complete user tasks end-to-end in real applications through interface actions such as tapping, swiping, text entry, and navigation. However, existing GUI agents are trained and evaluated largely on offline trajectories, simulated environments, and standardized benchmarks. These differ substantially from real applications in interface layout, interaction logic, and abnormal-state distribution, and cannot faithfully characterize execution stability in real-world use, where account states, permission dialogs, payment authentication, and risk control continually reshape the state distribution and open a persistent gap between benchmark scores and real usability. To close this gap, we propose Xiaomi-GUI-0, a native multimodal GUI agent for real mobile environments, trained and evaluated within a real-device closed loop. At its core is a real-device-dominant hybrid infrastructure, where physical devices are the primary execution environment and sandboxes provide auxiliary support, so that data collection, training, rollout, and evaluation share an execution distribution close to real deployment. We construct multi-source training data spanning high-frequency head tasks, high-generalization data for long-tail intents, and capability-enhancement data for reflection and memory, and introduce an error-driven data flywheel that turns failure trajectories into corrected actions, reflective explanations, and recovery demonstrations. The model is trained through a progressive three-stage pipeline of supervised fine-tuning, step-level reinforcement learning, and agentic reinforcement learning. Evaluated on public benchmarks and our in-house RealMobile, Xiaomi-GUI-0 achieves 72.0% success on RealMobile and 78.9% on AndroidWorld, while substantially improving execution stability and abnormal-state recognition in real-world tasks.
Hierarchical Vision-Language-Action (VLA) models decouple high-level planning from low-level control to improve generalization in robot manipulation. Recent work in this paradigm uses 2D end-effector trajectories predicted by a Vision-Language Model (VLM) as explicit guidance for a downstream policy. However, state-of-the-art low-level policies operate in 3D metric space on point clouds, and feeding them 2D guidance that lacks depth forces each waypoint to be assigned the depth of whatever scene surface lies beneath it, producing geometrically distorted trajectories. We propose 3D HAMSTER, a hierarchical framework that closes this gap by having the planner directly output metrically reliable 3D trajectories. We augment a VLM with a dedicated depth encoder and a dense depth reconstruction objective to predict 3D waypoint sequences, which are directly integrated into a pointcloudbased low-level policy. Across 3D trajectory prediction, simulation, and real-world manipulation, 3D HAMSTER consistently outperforms proprietary VLMs and 2D-guided baselines, with the largest gains under appearance-altering shifts and unseen language, spatial, and visual conditions. The project page is available at https://davian-robotics.github.io/3D_HAMSTER/.