Artificial Intelligence
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
Monday 17 August
Safety alignment in open-weight language models is trivially removable: abliteration projects a refusal-mediating direction out of the weights in minutes, and no release-time defense we are aware of prevents it durably. What cannot be prevented can be deceived. Our defense, decoy hardening ("Fool's Gold"), concedes the refusal strip and poisons its payoff: once refusal is stripped, most answers to hazardous operational requests are confident, fluent decoys whose critical elements are falsified. Decoys are trained inside a differentiable simulation of the attack, expressing only in the attacked state; a refusal pin and benign leash hold clean-state behavior to the original. We instantiate it on seven models from five families (9B-122B, dense and mixture-of-experts). On the six models passing our pre-registered efficacy gate, 0.51-0.90 of attacked-state responses to held-out prompts are decoys, +0.27-0.84 attributable to the defense; all six stay within registered benign-behavior and capability budgets; the seventh (smaller) fails the gate (boundary case). Rates replicate on a frozen test split or untouched strata. The claim is epistemic: without independent ground truth, no observation surface we tested separates falsified answers from correct ones - on external red-team benchmarks' CBRNE-adjacent slice, the defended 122B is fatally wrong on 0.82-0.86 of matched-quality answers vs at most 0.10 undefended. Repeated sampling does not restore trust: element-wise consensus at K=64 reconstructs a fully usable procedure on 0.083-0.625 of prompts where the instrument validates, vs 0.58-0.96 undefended, with no label-free way to tell the regimes apart; on the weakest such model the claim is per-draw only. We evaluate chemical and biological hazards; the defense does not address in-context jailbreaks and protects only the initially released defended weights.
Information Theory · Cross-listed from quant-ph
One-at-a-Time Quantum Guessing: Multipartite Entanglement Beyond MoE Games
Multipartite entanglement remains a challenging and not fully understood aspect of quantum information. Monogamy-of-Entanglement (MoE) games have been highly effective for studying limitations on the usefulness of entanglement imposed by monogamy constraints. To better reveal the extent to which multipartite entanglement can be useful, we introduce a class of quantum guessing games, termed One-at-a-Time Guessing (OTG) games. In these games, quantum players individually guess the outcomes of random measurements performed by a referee on a pre-shared entangled state. Unlike MoE games, OTG games select players individually at random according to a specified probability distribution, thereby probing each player's correlation with the referee. We show that, despite monogamy constraints, players sharing certain entangled states can moderately outperform those relying only on classical uncertainty. This advantage arises even in simple OTG games involving only Pauli measurements on qubits, where optimal entanglement increases the winning probability by at least 4%. This contrasts with MoE games, where shared entanglement has been shown in several settings to provide only limited (if any) advantage over classical strategies. We further establish a majorization property: the value of an OTG game respects the majorization ordering of the player-selection probability distribution. We also analyze in detail a two-player OTG game in which the referee measures one of the three Pauli observables on a qubit, and show that it is optimally played using a specific parameterized family of three-qubit $W$-like states. These results suggest that OTG games provide a useful framework for investigating the usefulness of multipartite entanglement in multiparty quantum correlations.
Read the full abstract
Numerical Analysis · Cross-listed from math-ph
A Projected Semiexplicit Integrator for Dissipative Systems with Configuration-Dependent Kinetic Energy: Contact-Herglotz Formulation and Benchmarks
Contact Hamiltonian dynamics gives dissipative mechanics an intrinsic action variable, but explicit contact splittings reach only kinetic energies whose terms are exactly integrable: frozen-coordinate diagonal metrics (the spherical pendulum, a torus particle) are included, while dense metrics with momentum cross terms, with the double pendulum as flagship, are not. We introduce a projected Pihajoki-contact integrator for this non-separable setting, combining phase-space duplication, symmetric projection onto the physical diagonal, and constant-friction damping half-steps, with the action factor carried by an exact Herglotz update. As in the projected extended-phase-space framework it builds on, the construction needs no binding parameter, returns the copies to the diagonal at every step, and confines the nonlinear solve to the $2n$ projection variables. For constant friction the step rescales $\omega=d\eta$ by the exact factor $e^{-\gamma\tau}$ when the projection is solved exactly (a classical conformally symplectic identity, realized here for this class), while time-symmetry, consistency, and smoothness yield an $O(\tau^3)$ one-step contact-form residual, a bound not specific to the contact form. On the damped double pendulum, spherical pendulum, and torus particle the method is second-order accurate, reproduces the contact decay law, and controls long-time energy and contact drift in coarse or stiff regimes where the Tao baseline and the unprojected average lose the solution. A head-to-head with exact-contactomorphism splittings delimits the niche: where a frozen-coordinate splitting exists it preserves the contact form exactly and wins at matched cost; for the dense double-pendulum metric the realizable alternative is first-order with a prohibitive constant and the projected method prevails. The contact-form estimate is local, one-step, and constant-friction.
Read the full abstract
Software Engineering
Graphectory Viewer: A Tool for Process-Centric Analysis of Agentic Software Trajectories
We present Graphectory Viewer, a web-based tool for interactive, process-centric analysis of software-agent trajectories. Building on the Graphectory representation introduced in our previous work, Graphectory Viewer transforms heterogeneous raw trajectories into phase-aware graphs that connect low-level execution details with higher-level behavioral structures. The tool supports trajectories from multiple agent frameworks and provides interactive graph construction; node-level inspection of thoughts, actions, and observations; search and filtering over large trajectory collections; and Sankey-style summaries of problem-solving phase transitions. These capabilities enable researchers and practitioners to inspect individual executions, identify recurring behavioral patterns, compare successful and failed runs, and analyze large trajectory corpora beyond final task outcomes. To support reproducibility and further research, we release Graphectory Viewer as an open-source artifact together with documentation, precomputed graphs, and the large-scale trajectory corpus.
Read the full abstract
Machine Learning
How smoothing the affinity matrix affects neighborhood preservation in t-SNE
Dimensionality reduction methods are instrumental to visualize high-dimensional data, and t-SNE stands as one of the most widely used methods due to its emphasis on local neighborhood preservation. A central component of t-SNE is the affinity matrix, which expresses pairwise similarities in the form of symmetrized probabilities, over which the optimization problem of t-SNE is defined. We study how the sharpness of this probability distribution affects neighborhood preservation at different scales. We introduce a row-wise power transform controlled by a parameter gamma that can smooth or sharpen each row of the affinity matrix while preserving sparsity and rank order. We show that this transform is equivalent to rescaling the Gaussian bandwidth and thus to changing the perplexity. However, as the sharpness of the probability distribution varies per point, a fixed gamma leads to point-dependent effective perplexities, making it distinct from changing the global perplexity. Empirically, we find that sharpening improves preservation of the very nearest neighbors, while smoothing improves preservation of broader local neighborhoods, outperforming alternative affinity constructions including multiscale methods in the mid-local range.
Read the full abstract
Computation and Language
Token Optimization and Context Window Management in Multi-Agent AI Workflows
Multi-agent AI workflows are limited not only by model quality but by token cost, latency, and context-window quality. This paper presents a practitioner framework for token optimization and context-window management, grounded in an internal production dashboard that extracts structured work items from meetings, email, and chat with LLMs and routes summaries across workstreams. Six patterns are described: context stratification, fetch-once/process-locally architecture, schema-contracted prompts, token-aware fallback chains, semantic caching, and inter-agent communication compression. In production they cut measured cold-load latency to 61-116 seconds (six timed runs) from an operational baseline of roughly 3.5-10.5 minutes, with an estimated 60-70% token reduction. It also reports a controlled context-composition study: 2,420 confirmatory trials across 11 model configurations, using 661 anonymized workplace items scored for relevance. Holding the prompt at a fixed ten items, replacing some high-relevance items with same-domain low-relevance items improves the model's relevance-score concordance on the target items, versus high-relevance items only; we call this relevance-contrast context. In the all-11 paired analysis, the 50:50 signal/noise condition improved relevance accuracy by +0.077 over the 100% condition (naive 95% CI [+0.056, +0.098], Cohen's d = 0.49, Holm-adjusted p < .001, n = 220). These cells are not independent; by the nine model families the effect is +0.084 (95% interval [+0.064, +0.103]), reported as a within-corpus descriptive comparison, not a population inference. A Fusion-of-N follow-up found that learned synthesis did not beat the mechanical set union of item IDs. The contribution is a measured engineering layer between model research and production agent practice: repeatable patterns and evaluation methods for faster, cheaper, more reliable workflows.
Read the full abstract
Computation and Language
AISA: AI Safety Assistant Framework for Continuous Improvement of Highway Construction
Job Safety Analysis (JSA) and pre-task planning can benefit from prior incident records, yet historical accident data is often stored as unstructured narratives that are difficult to consult at the point of planning. A novel framework centered on large language models (LLMs) for highway construction safety reporting and planning is proposed as a foundation for future agentic applications, prioritizing deterministic, local inferencing. The first aim is to enable classification and quality scoring of incident narratives for existing and future reporting purposes. The second is to evaluate retrieval of relevant historical accidents, related imagery, and trusted industry documents for incorporation into daily safety plans. Neural probes were trained to classify incidents along four multiclass and two binary Occupational Injury and Illness Classification System (OIICS) fields and to derive an overall quality score, evaluated on a test set of over 15,000 narratives and a held-out set of 100 author-labeled records, benchmarked against a majority-vote LLM ensemble. The retrieval of historical accidents, reference imagery, and industry documents was benchmarked across embedding models using standard information retrieval metrics. OIICS classification reached 75% held-out accuracy, though the two binary flags were degenerate. The quality score, while meaningful on one database, was distorted on out-of-distribution fatalities in the held-out dataset. Accident retrieval recovered relevant incidents far above chance, performing best on lexically distinct construction activities. On document question answering, an open-weight decoder embedding model surpassed proprietary models. Overall, this work provides a new framework rooted in local inferencing and text embedding models for future agentic applications, with emphasis on bridging external data to JSA reports.
Read the full abstract
Artificial Intelligence
Benchmarking the Benchmarks: Evaluating Automated Safety Benchmarks for Small Language Models
Small Language Models (SLMs) are increasingly deployed in resource-constrained, privacy-sensitive settings, where safety and bias failures can cause security and societal risks. However, existing AI safety\slash security\slash compliance benchmarks are designed for large language models that may not transfer reliably to SLMs. We therefore ask: Can these benchmarks effectively and reliably evaluate SLMs? To answer this question, we conduct a large-scale assessment of the effectiveness and robustness of these automated pipelines by evaluating five widely used benchmark suites across 26 open-source SLMs under a unified judging rubric, which assigns a score of 0, 1, or 0.5 to harmful, safe, or ambiguous/irrelevant responses, respectively. Across the benchmarks, ambiguous judgments dominate and correlate with prompt complexity and model architecture, indicating that LLM-centric safety benchmarks are insufficient as standalone evidence for SLM safety assessment. In general, the ambiguity rate increases with lexical density, output perplexity, and output length and decreases with lexical sophistication, self-coherence, and reply-prompt similarity. This reveals a capability-safety confound that mixes model capability with apparent safety. Since ambiguity is prevalent, aggregate mean-score leaderboards are mathematically brittle: model rankings change significantly under reasonable ambiguity treatments, even when the underlying outputs remain unchanged.
Read the full abstract
Graphics
RADmesh: Remesh-Aware Mesh Deformation
We propose a remeshing-enhanced method for generatively deforming shapes with visual losses. It is intuitive that sufficiently drastic deformations of a mesh without changing its triangulation can easily compromise element quality, even if such large geometry changes may be semantically desired. Shape deformation methods could thus benefit from changing the triangulation; however, this is not done by most generative, text-based, visually-supervised mesh deformation methods. Remeshing is a discrete operation, proven to be especially challenging to couple with the notoriously noisy supervision signal provided by visual losses. We propose a vertex-based deformation optimization quantity capable of large deformations and robustness to such noise; we periodically remesh using an isotropic remesher that interpolates and carries forward the deformation optimization state. This enables continuous, geometry-informed progress in coarse-to-fine addition of resolution. The resulting shapes' triangulations fit their optimized geometry and have neat isotropic elements. Further, our method is localizable, able to grow new features on a base shape with expressive detail, leaving the rest unchanged. We showcase the effectiveness of our method on a variety of shapes and prompts, both local and global deformations, and demonstrate its superior visual quality and triangle efficiency. Our project page is at https://threedle.github.io/radmesh.
Read the full abstract
Machine Learning
Reinforcement Learning as (Discrete) Potential Theory
Reinforcement learning (RL) theory fundamentally depends on probability theory through the Markov chain. There is a deep connection between probability theory and potential theory. This paper reviews that connection and explores the potential-theoretic viewpoint for core reinforcement learning representations and algorithms under a fixed-policy assumption. This viewpoint may offer a path for improved sample efficiency and formal constraints that can be applied to RL. When the fixed-policy assumption is relaxed, the linear potential theory framework can be naturally extended to the nonlinear case.
Read the full abstract
Machine Learning
Task Specialization Fine-Tuning for Contextual Reinforcement Learning
Contextual Reinforcement Learning (CRL) seeks to generalize classical RL by maximizing task coverage across a context space of related tasks. While prior works often train from scratch and rely on either multi-task learning for a single policy or strategically training multiple policies, we advocate for a unified alternative: pretraining a single policy with good initial performance, followed by fine-tuning multiple policies for task specialization. This new paradigm, however, introduces unique challenges, such as heterogeneous marginal returns and sample inefficiency. This raises a critical research question: given a pretrained policy and a constrained budget, how much fine-tuning should each task region receive to enable sample-efficient CRL? To this end, we propose Task Specialization Fine-Tuning (TSFT), an online framework that predicts fine-tuning performance with a simple parametric model and exactly solves the resulting discrete budget allocation problem via integer linear programming. Extensive experiments across diverse decision domains, including combinatorial optimization, continuous control, and LLM fine-tuning, demonstrate that TSFT significantly outperforms baselines in task coverage and approaches oracle performance. Our work charts a new direction for model-based CRL, aligning with the modern pretrain-finetune era.
Read the full abstract
Computer Vision
Mask What Matters: Saliency-Guided Video Self-Supervised Learning for Autonomous Driving
Video self-supervised learning through masked spatiotemporal prediction has emerged as a promising paradigm for learning feature representations from unlabeled data. However, existing methods typically rely on random masking, which indiscriminately removes regions irrespective of their semantic or temporal relevance. In ego-centric driving videos, this can weaken the pretext signal since safety-critical cues such as pedestrians, vehicles, lane boundaries, and dynamic interactions often occupy only a small portion of the frame, yet are central to downstream perception. We introduce V-JEPA4A, a domain-specialized variant of V-JEPA for autonomous driving that is pre-trained on publicly available driving videos with a novel saliency-driven masking policy. It accounts for semantically and temporally relevant context. The proposed policy preserves and predicts context according to semantic importance and temporal relevance, yielding more informative representation learning while retaining the efficiency of masked prediction. We evaluate the resulting encoders on four driving benchmarks spanning tracking, semantic segmentation, and depth estimation. The results demonstrate that V-JEPA4A reduces identity switches on BDD100k MOT by 25% over V-JEPA with random masking, achieves 73.2 mIoU on Cityscapes, and 3.75 RMSE on KITTI-2015 depth, while incurring only ~14% additional pre-training iteration overhead.
Read the full abstract
Software Engineering
Grounding AI Agents in Contracts: An Empirical Evaluation of Spec-Driven Test Generation
LLM-based agents are increasingly used for coding tasks, where they have outperformed many classical approaches and scaled to repository-level tasks, such as test generation. However, when directly prompted to generate tests, these agents can fail to reason about the code and its underlying contracts, thereby missing edge cases and behavioral boundaries that affect test quality. To address this limitation, we propose Spec-Driven Test Generation, where we instruct an agent to first reason about – and explicitly document – code pre-conditions, post-conditions, and undefined behaviors. This intermediate semi-formal specification acts as a cognitive scaffold to guide subsequent test generation. Our evaluation on production bugs from Google shows that the spec-driven agent can deliver a 9.8 percentage points ($p = 0.0352$) improvement in bug detection rate and a 2.5 percentage point ($p = 0.0034$) improvement in branch coverage, compared to a traditional test generation agent baseline. Using LLM-as-a-Judge, we further show that test suites generated by the spec-driven agent are superior to the baseline and human-authored tests in 77.8% and 56.7% of the cases, respectively, and demonstrated improvements on following best practices, readability, and edge-case coverage.
Read the full abstract
Cryptography and Security
The Acknowledgment Point Is the System: Durable Policy-Decision Receipts for AI Audit Evidence
An AI audit record is useful only if its durability and trust boundary are explicit. Returning a guarded decision before any durable write minimizes latency, but it cannot guarantee that evidence survives an immediate crash. We rebuild RuntimeGuard-AI around this constraint. The resulting research prototype binds each deterministic policy decision to the exact policy source, commits a privacy-minimizing record at a caller-selected synchronization boundary, and returns an Ed25519-signed receipt that states whether that boundary completed. After restart, the engine validates framed records, manifests, shard placement, sequence continuity, and replay identity. A separate attestation path groups committed records into chained, signed Merkle epochs that an auditor verifies with an externally obtained key. On an Apple M4 Pro at four worker threads and 2,048-byte prompts, buffered signed evidence reaches 27,193 requests/s with 141.9 microseconds median latency. Per-record data and full synchronization reduce throughput to approximately 242 requests/s and raise median latency to 16.0 ms. Sealing a 100,000-record signed epoch takes 97.0 ms. The result is a measured durability-latency trade-off, not a "free" asynchronous audit path. The prototype does not prove model execution, prevent a compromised signer from forking history, or establish legal conformity.
Read the full abstract
Human-Computer Interaction
Balancing Safety and Autonomy: Accessibility-Oriented Interventions in Generative AI for Cognitive Impairment
Generative AI systems are increasingly used by older adults with cognitive impairment for everyday tasks such as information seeking, health management, and communication. While these systems provide flexible, language-based support, their open-ended outputs introduce risks of over-reliance, misinterpretation, and inappropriate decision-making. Prior work has focused on usability and adoption, with limited attention to how system design shapes users' participation in decision-making and the distribution of agency in care contexts. We present a qualitative study of 45 individuals with cognitive impairment and their caregivers. We identify five accessibility-oriented mechanisms: AI Capability Constraint, Human Oversight Embedding, Cognitive Engagement Maintenance, Human-AI Relationship Regulation, and Risk Transparency and Control, through which systems structure interaction. These mechanisms both support and constrain users by redistributing decision-making across users and caregivers. We show that their effects vary by impairment level: while protective mechanisms support users with severe impairment, they can restrict autonomy for those with mild impairment. As impairment progresses, tensions become less visible as user participation diminishes. Our findings highlight the need for dynamic designs that balance safety and autonomy in AI-supported care.
Read the full abstract
Machine Learning
Population Health-Based Machine Learning Reveals Associations Between Psychosocial Factors and Chronic Kidney Disease
Chronic kidney disease (CKD) progresses silently and severely undermines quality of life, making early detection critical for improving patient outcomes. We present a two-part study that combines large-scale telehealth data with advanced machine learning to both classify self-reported CKD status and identify key drivers of disease. Using selected features from the Behavioral Risk Factor Surveillance System (BRFSS 2021: 438,693 samples; BRFSS 2019: 418,268 samples) and the National Health Interview Survey (NHIS 2021: 29,482 samples; NHIS 2020: 31,568 samples), we addressed missing data with nine state-of-the-art imputation methods and mitigated class imbalance via sampling strategies. Our customized stacked ensemble model achieved balanced accuracy of 72.56-76.12%, with corresponding AUROC scores of 79.59-82.29%. SHapley Additive exPlanations (SHAP) analysis, followed by clinical review, highlighted critical predictors, including regular medical check-ups, age, blood pressure, and indicators of mental health stress. These findings deliver a robust and interpretable framework for CKD risk stratification and provide actionable insights into its associated factors.
Read the full abstract
Machine Learning · Cross-listed from stat.ML
Policy Optimization and Statistical Inference for Online Contextual Matrix Games
Online decision making often requires navigating a landscape shaped by both dynamic contexts and strategic interactions. In competitive pricing, for example, hotels must account for both dynamic contextual factors and rivals' strategic responses. Existing approaches address only part of this challenge: contextual bandits optimize single-agent decisions using observable features but ignore multi-player interactions, while online matrix games capture strategic behavior through Nash equilibrium but assume fixed payoffs, ignoring contextual information. How should agents act then when strategic payoffs evolve with contextual signals? We introduce online contextual matrix games to integrate contextual information into multi-player online games. We further propose OnGameLearn, an online learning algorithm that efficiently balances exploration and exploitation across both player actions and contexts. This approach comes with statistical guarantees: tail bounds for the estimated payoff matrix, the convergence of the estimated Nash equilibrium, the asymptotic normality of the parameter estimators, and the sublinear regret bound. We also develop the notion of policy value in matrix games and develop a doubly robust, $\sqrt{T}$-consistent estimator for it. Across simulated studies and a real-world hotel pricing application, we find that OnGameLearn effectively navigates the intertwined challenges of strategic and contextual decision-making.
Read the full abstract
Neural and Evolutionary Computing
Automating Parent Selection Configuration in Genetic Programming with Agentic AI
We investigate whether agentic artificial intelligence can automate parts of the process of designing genetic programming systems by introducing an agentic framework that identifies and implements parent selection algorithms using large language model (LLM) reasoning and retrieval-augmented generation. Using symbolic regression as a test bed, we first conduct an ablation study across four LLM types to evaluate the effects of agentic reasoning and retrieval on generated algorithm categories, validity, implementation similarity, and downstream performance. Results show that these components substantially influence the types of algorithms generated, but their downstream performance largely depends on the underlying LLM. The strongest configuration, the full agentic setup with 5 mini (5 mini–AR), consistently generated established $\epsilon$-lexicase implementations while maintaining competitive downstream performance. We then benchmark this configuration against fixed implementations of tournament selection and semi-dynamic MAD $\epsilon$-lexicase. Across six symbolic regression problems, 5 mini–AR performed similarly to $\epsilon$-lexicase while generally outperforming tournament selection. These findings demonstrate the potential of agentic AI to translate domain knowledge into generating executable components, providing a step toward automated configuration and design of evolutionary systems.
Read the full abstract
Computation and Language
Polaris: Learning to Generate Table Descriptions from Retrieval Feedback
Many table-centric NLP tasks such as NL2SQL first retrieve relevant tables from large collections using keyword search. Recent work uses LLMs to generate natural-language table descriptions to improve retrieval, but they are typically optimized for fluency rather than retrieval effectiveness. We present Polaris, a system that trains an LLM to generate table descriptions directly from retrieval feedback. Our key insight is that existing table retrieval benchmarks already contain the supervision needed for this task: given query-table relevance judgments, we generate multiple candidate descriptions for each table, rank them by their BM25 retrieval effectiveness, and use the resulting preference pairs to fine-tune the LLM with Direct Preference Optimization (DPO). Polaris further expands abbreviated table and column names before generation to reduce vocabulary mismatch. Extensive experiments show that Polaris outperforms the state-of-the-art AutoDDG solution, often by a significant margin. More broadly, our results demonstrate that retrieval benchmarks can be repurposed as supervision for training LLMs to generate retrieval-oriented metadata.
Read the full abstract
Artificial Intelligence
Synthesizing Feature Extractors: An Agentic Approach for Algorithm Selection
Algorithm selection for constraint satisfaction problems requires extracting features that capture problem structure. Manually designing feature extractors demands deep domain expertise and quickly becomes a bottleneck when new problem classes appear. We present an automated approach that uses Large Language Models (LLMs) in an agentic check–fix–verify loop to synthesize executable Python scripts that act as interpretable, problem-specific feature extractors. Given a high-level MiniZinc model and an instance, the LLM agent generates code that constructs a typed graph representation and computes structural properties such as graph density, variable clustering, and constraint tightness. We evaluate our approach on three combinatorial problems (vehicle routing, car sequencing, fixed-length error-correcting codes) with a portfolio of five state-of-the-art solvers. The synthesized extractors yield algorithm selectors that consistently outperform both expert-curated mzn2feat features (up to $8.3$ percentage points (pp) test-set accuracy on FLECC) and the best transformer-based trans2feat variants. In the meanwhile, the synthesized feature extractors remain inspectable.
Read the full abstract
Computation and Language
Can LLMs Reason in a Legally Meaningful Manner? A Small-scale Study on European Court of Human Rights Cases
Reasoning has become a standard technique and feature for contemporary LLMs; however, its application and quality in the context of demanding legal-oriented tasks, such as legal case forecasting, remain under explored. We investigate how LLMs reason in the context of legal case forecasting, using legal cases from the European Court of Human Rights (ECtHR) as a testbed. We evaluate OpenAI GPT 5.4, a recent top-tier LLM, by exploring alternative prompting strategies that are more or less suggestive of what counts as legally meaningful reasoning in the context of ECtHR jurisprudence. We present our findings derived from assessing the model's responses with both human and LLM evaluation. We find that the examined model scores far from ideal in legal reasoning, the model produces structurally complete but substantively shallow analyses, and that LLM-as-a-Judge evaluators are internally consistent yet align only weakly with our trained annotators, i.e., reliable but not a valid substitute for human evaluation. Overall, the expert-curated prompt leads to more comprehensive reasoning, which does not result in more accurate predictions compared to the other examined settings. Based on our findings, we urge the community not to rely solely on automated LLM-based evaluation and to avoid using task accuracy as an appropriate proxy for reasoning quality.
Read the full abstract
Information Theory
Expected free energy as an information constraint on the Bethe Lagrangian
Active inference selects actions by minimising an expected free energy functional over predicted futures. However, adding an expectation over yet-unobserved outcomes means the free energy functional no longer has a Kullback-Leibler structure, which hinders message passing treatments of inference procedures. We propose an alternative formulation based on a Bethe free energy functional, fully supporting inference by message passing. The epistemic drive is maintained by imposing an information constraint, next to normalisation, marginalisation and form constraints, insisting that the mutual information between future observations, states and parameters given actions must be at least as large as the entropy of the goal prior. For a specific value of the corresponding Karush-Kuhn-Tucker multiplier, the stationary point of this constrained Bethe Lagrangian recovers the expected free energy solution. We show that, as the information demand is varied, the solved multiplier moves through its inactive, interior, and saturated regimes. In the inactive regime the agent's epistemic drive switches off entirely, while in the saturated regime it is maximal. We compare the performance of the constrained Bethe agent on three tasks against EFE and Q-MDP.
Read the full abstract
Computer Vision
Rapid Debris-Volume Estimation from Post-Hurricane Aerial Imagery
Hurricane debris removal is planned, contracted, and federally reimbursed on the basis of volume estimates, yet operational practice still relies on parametric forecasts with 41-90% documented over-estimation or on truck-load tallies that arrive only after hauling begins. We present DebrisHeightNet, a segmentation-conditioned monocular debris-height network that estimates spatially explicit debris volume from a single pass of post-event aerial RGB imagery, the kind of survey routinely flown within days of a hurricane landfall. We train only a lightweight 1.08 M-parameter head on top of two frozen vision foundation models. This head regresses height from a Depth Anything V2 backbone, conditioned on the debris segmentation of CLIPSeg-debris from our prior work. Because no post-hurricane debris-height ground truth exists, we synthesize the training target by confidence-weighted LiDAR-monocular fusion (CW-LMF), designed to suppress non-debris LiDAR returns. This fused target is a constructed supervision signal rather than ground truth, so we corroborate it against external references rather than claiming it as truth. A region-level power-law calibration, driven by each region's low-density debris fraction, converts model volume into an estimate of the reported hauled debris with quantified uncertainty. Across ten regions spanning five hurricanes and three states, the uncalibrated model agrees with an independent uncrewed-aerial-vehicle (UAV) survey of the training region at Spearman $\rho = 0.87$ and lands within 30% of the reported record where the Hazus and FEMA-hybrid parametric forecasts over-predict it by 2.7-4.8$\times$. Deployment requires no LiDAR, no ground access, and no second flight, so the method can produce spatially explicit volume estimates wherever single-pass post-event imagery is flown.
Read the full abstract
Machine Learning
SCENARIODIFF: A Scenario-level Guidance Framework for Multimodal Time Series Forecasting–Extended Version
Textual context such as news, reports, and logs can provide valuable signals for time series forecasting, especially when future dynamics are driven by external events that are not yet visible in historical values. Existing multimodal forecasting methods often either ask large language models (LLMs) to predict numerical values directly or fuse text and time series implicitly, making contextual influence difficult to interpret and control. We propose SCENARIODIFF, a hierarchical contextual reasoning framework for multimodal time series forecasting under noisy and weakly aligned documents. SCENARIODIFF organizes contextual information into three levels: a Historical Context Agent extracts stepwise evidence from raw documents, a Scenario Agent produces a qualitative scenario description for the forecast horizon, and an Anchor Guidance Agent generates sparse anchor points for event-relevant future regions. These structured signals condition a Multimodal Diffusion Transformer, while Anchor Blended Sampling locally refines generated trajectories without retraining. Experiments on the Time-MMD benchmark show that SCENARIODIFF is especially effective in event-driven domains, demonstrating the value of explicit hierarchical scenario guidance for multimodal time series forecasting. Our full implementation is available at https://anonymous.4open.science/r/ScenarioDiff_ICDM-2C4C
Read the full abstract
Machine Learning
Q-Learning With World Models
Off-policy reinforcement learning (RL) has become increasingly sample-efficient, enabling applications such as RL fine-tuning of Vision-Language-Action models into reliable, high-performing policies. World models offer a further lever for sample efficiency, as they predict state changes rather than actions alone, but their success has largely been confined to supervised policy learning. Prior model-based RL methods often optimize the policy or value function directly on imagined rollouts, which is prone to compounding bias and struggles to scale to large, high-dimensional problems such as real-world robotics, a problem that worsens with task horizon and visual complexity. In this work, we instead ask whether we can leverage world models directly on top of standard Q-learning to improve performance, while remaining trained and grounded in the real, online setting. We propose QWM, a framework that leverages world models to perform test-time search over imagined trajectories on top of Q-learning to select high-value actions during both online rollouts and evaluation. Since the policy and value function are trained only on real transitions, QWM avoids compounding model bias while still gaining the sample-efficiency benefits of predictive search. On challenging manipulation benchmarks Robomimic and LIBERO, QWM significantly outperforms strong prior state-of-the-art methods on both sample efficiency and performance.
Read the full abstract
Machine Learning
OraclePhys: A Systematic Framework for LLM Fine-Tuning on Structural Mechanics
What fine-tuning teaches a model is usually diagnosed after the fact. We make it an experimental variable. OraclePhys has three parts. OraclePhys-Bench is a structural-mechanics benchmark whose finite-element oracle grades every answer and counterfactual edit. OraclePhys-30K holds seven answer forms over byte-identical structure descriptions. A controlled training study compares the forms and verifier roles. Two findings follow. First, within this task family the answer form of the label determines what fine-tuning teaches. Ranking supervision lifts held-out localization of the most-deforming story on taller structures from 0.25 to 0.86, a peak-value label yields at best a partial predictor, and a pass/fail label none. This ranking-versus-peak gap replicates on a second domain, a second model family and a paraphrased surface. Second, the gap between our standard SFT and GRPO runs is mostly a recipe effect. With learning rate and step budget matched, the two reach held-out accuracies within 0.03 of each other. At higher rates, GRPO falls below its start while SFT does not. Both also teach the model to apply edits given as instructions. The trained 8B matches a specialist and exceeds a frontier LLM. Both domains are linear, steady-state systems with synthetic text inputs.
Read the full abstract
Software Engineering
A Multi-Surface Consistency Audit of Software Citation Metadata
Research software projects describe themselves in many places at once: citation files in the repository, archive deposits, DOI registry records, package registries, and README text. We treat the software as the underlying object and these machine-readable self-descriptions as its surfaces: the points where people and automated systems read what the project declares about the software. Citation guidance, indexing services, and automated agents may read a different subset of these surfaces, so disagreement between them can silently fragment credit and provenance. This paper asks a simple question that has not been measured directly: when a project's own metadata surfaces are compared with each other, how often do they agree? We audited 117 open-source research software projects, comprising an 87-project high-performance computing and quantum computing corpus and a 30-project registered baseline drawn from the JOSS and pyOpenSci accepted-package lists, across up to seven machine-readable surfaces per project. Using a four-level verdict rubric across six metadata fields, with 98.5% hand-verified verdict precision on a 338-row stratified sample, we found that 52 of the 62 projects exposing at least two comparable surfaces (83.9%) contain at least one core-field conflict, a result that is insensitive to the fuzzy-matching threshold. Half of hand-adjudicated cross-surface conflicts trace to a single mechanism: surfaces describing the software's paper rather than the software itself. Among projects whose CITATION.cff includes a preferred citation, 28 of 32 route citations to a record that disagrees with the software's own metadata. The author lists and titles disagree the most, and the registry surfaces are the least aligned. We release the audit pipeline as an importable library, the corpus, the registered sampling protocol, all raw snapshots, and the complete verification log.
Read the full abstract
Numerical Analysis
Robust High-Order Projector-Splitting Integrators
We develop and analyze an increment-based $K$-$L$ integrator for dynamical low-rank approximation. This formulation eliminates the separate backward $S$-step and is algebraically equivalent to the fixed-increment projector-splitting integrator of Lubich and Oseledets. We establish a relaxed-exactness estimate that bounds the error in the low-rank approximation in terms of the prescribed increment error, with constants independent of small singular values. This estimate provides a framework for achieving robust high-order convergence through sufficiently accurate increments. We then construct low-rank Runge–Kutta methods in which both the stage and final approximations are computed using the $K$-$L$ integrator, without basis augmentation or rank truncation. Since classical Runge–Kutta order conditions alone do not guarantee robust high-order convergence for this construction, we derive sufficient conditions on the coefficients that ensure the required stage accuracy and establish robust convergence of order $p$. We identify explicit Runge–Kutta schemes of orders $p=2,3,4$ satisfying these conditions, and numerical experiments confirm the theoretical results.
Read the full abstract
Cryptography and Security
Beyond the Hype: Evaluating LLM Integration and Practical Limitations in Security Operation Centers
Large Language Models (LLMs) are increasingly being explored within Security Operation Centers (SOCs) to support text-heavy analytical work such as alert contextualization, incident summarization, and drafting investigative artifacts. Despite this interest, practitioners describe critical operational concerns, most notably hallucinations (plausible but incorrect outputs), opaque reasoning, and the verification effort required to safely use model-generated content in security workflows. In this paper, we present findings from semi-structured interviews with 20 SOC practitioners spanning frontline analysts, SOC managers, and tool developers. Participants report perceived time savings for low-stakes tasks that are quickly verifiable (e.g., summarizing logs or drafting initial investigative leads), but they consistently frame LLM outputs as preliminary drafts and suggestions rather than decision-grade conclusions. Participants also describe limited trust in LLMs for high-stakes security decisions due to unreliable outputs and unclear model reasoning, and they report relying primarily on ad-hoc verification norms and continuous human oversight rather than standardized mitigation procedures. Based on these interview-grounded accounts, we introduce a maturity rubric to characterize readiness for LLM integration and outline a research agenda emphasizing auditability and transparent explanation mechanisms to support safer adoption in SOC workflows.
Read the full abstract
Computation and Language
When Detection Does Not Guarantee Resistance: Reasoning and Poisoned Context in RAG
Retrieval-Augmented Generation (RAG) exposes large language models to knowledge-poisoning attacks, where misinformation injected into retrieved documents can influence model outputs. Prior work has shown that models may detect contradictory evidence yet still allow it to influence their responses, revealing a gap between monitoring and control. We investigate whether deliberative reasoning changes this relationship. Because attack success and poison detection alone do not reveal whether detected poison continues to influence the final answer, we use two complementary measures: Cordon Rate, the probability that a model detects poison and nevertheless produces a poison-aligned answer, conditioned on a non-poison-aligned no-RAG response; and Leakage Rate, the probability that poisoned context influences the answer despite an explicit instruction to ignore retrieved documents. Across 200 SciFact questions, we compare reasoning-disabled and reasoning-enabled configurations of DeepSeek-V4-Flash and Qwen3.6-Plus. For DeepSeek-V4-Flash, reasoning reduces Cordon Rate from 0.205 to 0.075 and Leakage Rate from 0.235 to 0.140, despite increasing Attack Success Rate from 0.233 to 0.298 and decreasing Poison Detection Rate from 0.965 to 0.665. Qwen3.6-Plus shows the same qualitative pattern. These results demonstrate that poison detection and downstream influence capture distinct aspects of poisoning behavior.