Research

• Lack of benchmarks for agent harness reliability1 paper raises it · 1 addresses it

  Existing benchmarks focus on model capabilities rather than the reliability of the systems that support agent operation.

  • Trial by Trace: Benchmarking AI Agent Harnesses for Reliability and Failure Attributionraises · addresses
• Unclear net value of agent memory1 paper raises it · 1 addresses it

  Dedicated memory systems often underperform naive retrieval, making their benefits uncertain.

  • Memory Tax: Dissecting the Net Value of Agent Memoryraises · addresses
• Token collapse in graph soft prompting1 paper raises it · 0 address it

  Graph-specific tokens become indistinguishable, failing to convey intended relational information.

  • On the Token Collapse Phenomenon in Textual Graph Soft Promptingraises
• Poor trajectory supervision for tool-use agents1 paper raises it · 1 addresses it

  Successful trajectories may lack compositional coverage or decision grounding, limiting agent learning.

  • Trajectory2Gym: Decision-Grounded Compositional Interaction Synthesis for Tool-Use Agentsraises · addresses
• Uneven evaluation of coupled agent-world model systems1 paper raises it · 1 addresses it

  Separate evaluation of agents and world models fails to predict their joint performance or sustained learning.

  • Dual Exam: Controlled Evaluation of Multimodal Agents and World Modelsraises · addresses

Named by an AI from these papers' abstracts.

• Channel state information14+75.0%
• AI tutors13+18.2%
• Anomaly detectors190.0%
• Character recognition OCR100.0%
• Multimodal emotion7−12.5%
• Physics informed neural110−25.2%
• Machine learned interatomic32−31.9%
• Medical image segmentation119−35.0%

• arXiv cs.AIOct 8

  Defensive Writing in GPT Models

  Junchi Liao, Xiao Huang, Yucheng Liu, Jiawen Deng, Fuji Ren

  Researchers increasingly use ChatGPT to revise their papers, and when asked to make a paper hold up under review, recent GPT versions often narrow or even retract the authors' claims. We call such changes defensive writing when the given material does not support them, and we test two explanations: the model corrects the authors' overclaiming, or it writes for an anticipated reviewer. We ask GPT versions and models from other developers to rewrite paragraphs from papers written before ChatGPT, or to write from an evidence sheet that lists a paper's method and results. From GPT-5.5 on, GPT versions add clearly more ungrounded qualifications than the authors' originals contain, and GPT-6-astra adds the most and often retracts the authors' claims. Each explanation fits one part. Checked against the full papers, GPT-6-astra's retractions mostly correct overstated claims or explanations, but most added qualifications are redundant. The qualifications follow the request: they are absent under polishing, appear with any request to check the claims, and peak when the prompt names a reviewer's concerns. AI reviewers score such rewrites higher and score them lower once only the…

  PDF

  Summarize this
• arXiv cs.CLOct 8

  AdaptEvo: Learning under Imperfect Supervision through Confidence-Adaptive Training with Knowledge Evolution

  AllSpark Team

  Rule-governed contextual decision tasks require models to apply specified rules to case-specific context and evidence. Written rules can leave gaps in decision guidance and process evaluation, while reference judgments vary in their support from the rules and evidence. To address these challenges, we introduce AdaptEvo, a framework for learning under imperfect supervision that couples confidence-adaptive policy optimization with evolving decision knowledge and evaluation rubrics. Its Training module uses Confidence-Adaptive GRPO (CA-GRPO) to balance outcome and process rewards according to reference confidence. Its Evolution module synthesizes reusable decision knowledge from recurring failures across training cases and refines process rubrics to detect overlooked errors. To support empirical evaluation, we construct an industrial multimodal content moderation dataset comprising a training set and In-Period and Out-of-Period test sets, with the latter collected under changed rules. Using Qwen3.6-35B-A3B, AdaptEvo achieves 61.9% exact-label accuracy and 72.2% binary decision accuracy on In-Period, exceeding GRPO by 7.5 and 3.7 percentage points, respectively. On Out-of-Period, the…

  PDF

  Summarize this
• arXiv cs.AIOct 8

  RL-ARC: Calibrating Large Reasoning Models via Reasoning-guided Uncertainty

  Gukhyeon Lee, SangKeun Lee

  Language models (LMs) are commonly trained with Reinforcement Learning with Verifiable Rewards (RLVR) to enhance their reasoning capabilities. However, since RLVR does not explicitly account for calibration during training, it can lead to severe calibration degradation, including overconfidence. Recent calibration-aware training methods for LMs, which incorporate objectives for uncertainty estimation into training, improve calibration but still exhibit overconfidence under distribution shift, while sacrificing reasoning performance. To this end, we propose RL-ARC, a calibration-aware training framework that jointly leverages reasoning confidence and answer confidence. Specifically, RL-ARC leverages reasoning confidence as an auxiliary signal for calibrating answer confidence, applying it as reasoning-guided regularization for correct cases and as an overconfidence penalty for incorrect cases. Comprehensive results across ID and OOD settings show that, beyond improving calibration, RL-ARC enables reasoning models to adaptively estimate confidence based on the given question without substantially sacrificing reasoning performance, thereby highlighting the importance of reasoning…

  PDF

  Summarize this
• arXiv cs.CLOct 8

  Deception by Omission: Language Models Knowingly Hide Their Mistakes

  Lucas Florin, Amelie Knecht, Ulysse Schaller, Thilo Hagendorff

  Large language models (LLMs) increasingly act as agents with little human oversight, so potential mistakes they make can go unnoticed. Users then depend on the model to report what went wrong. An honest model discloses its mistakes, while a deceptive one conceals them. However, it is unclear how current LLMs behave in such situations. In this study, we prefill LLM trajectories with synthetic mistakes. The trajectories resemble real deployments in chat and agentic settings. Models fail to disclose their mistake in 36.4% of chat and 67.1% of agentic rollouts. In 2.4% and 5.3% of rollouts, respectively, they are aware of the mistake in their chain of thought but still deceptively conceal it. Rates vary by model: for instance, Gemini 3.5 Flash knowingly conceals mistakes in up to 19.9% of agentic rollouts. In 11.9% of chat and 51.8% of agentic rollouts, models show no awareness of mistakes, even though they reliably spot them when reviewing the same transcript as an outside observer. Our results show that, as agents take on more tasks with less oversight, users cannot rely on them to self-report possible mistakes. Developers should instead use independent monitors that review agent…

  PDF

  Summarize this
• arXiv cs.LGOct 8

  Sample-Efficient Generative Conformal Prediction

  Minxing Zheng, Shixiang Zhu

  Generative conformal prediction builds uncertainty sets from samples of a conditional generator, which are efficient only when the samples represent the response distribution well. This can require many samples, each of which can be costly, as in large diffusion models and scientific simulators, so the sampling budget must be used efficiently. Existing methods draw the same number of samples at every input, wasting samples where the response distribution is simple and undersampling where it is complex, which inflates sets and leaves those inputs under-covered. We propose CASA (Conformal Adaptive Sample Allocation), which characterizes the marginal value of an additional sample and allocates samples across inputs to minimize the expected set size subject to marginal coverage and an average sampling budget. Theoretical analysis shows that adaptive allocation yields smaller sets than a fixed count at the same budget: a missed mode forces a radius that spans the gap between modes, and even oracle radius cannot compensate for it. On synthetic and real tasks, CASA produces substantially smaller sets at the same budget, often improves conditional coverage, and complements existing…

  PDF

  Summarize this
• arXiv cs.CVOct 8

  EgoPhys: Estimating Peak Contact Force and Mechanical Work from Egocentric Manipulation Video

  Zhuo Dong, Jianhua Yang, Haohao Li, Yumeng Zhao, Keji He +2

  Physically grounded manipulation of articulated objects requires understanding both the maximum forces encountered during contact and the work performed as their parts move. Peak contact force and mechanical work quantify these complementary aspects, but estimating them from egocentric video is challenging because physical interaction cues are local and indirect. Moreover, peak force is associated with brief contact events, whereas mechanical work depends on force-motion coupling throughout the contact duration. To address these challenges, we propose EgoPhys, an RGB-only framework comprising Contact-Aware Spatial Aggregation (CASA) and Target-Specific Multi-Expert Temporal Routing (TMTR). CASA integrates appearance and geometry features to emphasize interaction-relevant cues, while TMTR models semantic, event, and motion cues with specialized temporal experts and routes them separately for force and work prediction. On the test split from Hoi! dataset, EgoPhys substantially improves predictions of peak force and mechanical work, achieving MAEs of \(5.205 \pm 0.584\) $N$ and $0.894 \pm 0.081$ $J$, respectively.

  PDF

  Summarize this
• arXiv cs.AIOct 8

  SynCo: Data Synthesis Co-Training for Self-Evolving LLMs via Multi-Agent Reinforcement Learning

  Wei Yang, Shawn Li, Yuehan Qin, Yawei Wang, Mingxi Wang +6

  Self-evolving LLM agents promise to improve autonomously through continual interaction and learning, reducing their dependence on manually curated supervision. Realizing this promise requires not only updating the agent, but also evolving its training experience as its capabilities change. However, most existing pipelines rely on static datasets or separately updated synthesis models, causing previously useful tasks to become trivial while overly difficult tasks remain uninformative. This growing mismatch between agent capability and training experience limits sustained self-improvement. To address this problem, we propose SynCo, an agentic data synthesis co-training framework for self-evolving LLMs based on multi-agent reinforcement learning. SynCo jointly optimizes two independently parameterized agents: a Synthesizer that constructs training tasks from the Reasoner's evolving capability state, and a Reasoner that learns from the resulting experience. Each synthesized task induces multiple Reasoner rollouts whose outcomes provide complementary rewards to both agents. Correctness feedback improves the Reasoner, while task quality, answer reliability, and outcome-grounded…

  PDF

  Summarize this
• arXiv cs.AIOct 8

  EvoSim: Learning to Model, Modeling to Learn

  Yun-Wei Song, Jinkai Tao, Jun-Dong Zhang, Rui Zhang, Yi-Min Wu +1

  Physics-based models connect scientific explanation with quantitative prediction. Constructing them requires selecting physical processes, defining states and governing equations, specifying couplings, and identifying parameters from experiments. Existing AI systems remain limited in making these model structure decisions autonomously. We introduce EvoSim, a self-evolving AI scientist for physical modeling. It uses experimental discrepancies to drive mechanism and equation revisions and held-out experimental data to test physical plausibility. Exploration traces make updates to knowledge, skills, and multi-agent orchestration. This co-evolution improves physics-based models and EvoSim's ability to select mechanisms, diagnose failures, and coordinate research. We evaluate EvoSim on two industrial battery modeling tasks. It predicts lithium-metal-plating onset from 25 to 45 degrees Celsius and 2 C to 6 C with a mean absolute error of 1.79% in state of charge. Dynamic voltage prediction under vehicle driving conditions achieves a root mean square error of 7.62 mV, surpassing the reported accuracy of models developed by human experts. Self-evolution reduces model and physics errors…

  PDF

  Summarize this
• arXiv cs.CVOct 8

  Point-Focused Attention Meets Context-Scan State Space: Robust Biological Visual Perception for Point Cloud Representation

  Kanglin Qu, Pan Gao, Qun Dai, Yuanhao Sun

  Synergistically capturing intricate local structures and global contextual dependencies has become a critical challenge in point cloud representation learning. To address this, we introduce PointLearner, a point cloud representation learning network that closely aligns with biological vision which employs an active, foveation-inspired processing strategy, thus enabling local geometric modeling and long-range dependency interactions simultaneously. Specifically, we first design a point-focused attention, which simulates foveal vision at the visual focus through a competitive normalized attention mechanism between local neighbors and spatially downsampled features. The spatially downsampled features are extracted by a pooling method based on learnable inducing points, which can flexibly adapt to the non-uniform distribution of point clouds as the number of inducing points is controlled and they interact directly with point clouds. Second, we propose a context-scan state space that mimics eye's saccade inference, which infers the overall semantic structure and spatial content in the scene through a scan path guided by the Hilbert curve for the bidirectional S6. With this…

  PDF

  Summarize this
• arXiv cs.FLOct 8

  Type-Checking for Pattern-Based Tree Transformations

  C. Aiswarya, Sahil Mhaskar, M. Praveen

  We introduce and study pattern-based tree transformations. As an illustrating example, consider a source pattern $(x \cdot y) + (x \cdot z)$ and a target pattern $x \cdot (y + z)$ as a pair. This source pattern matches any expression $e$ of the form $(e_1 \cdot e_2) + (e_1 \cdot e_3)$ (by substituting $x$ with $e_1$, $y$ with $e_2$, and $z$ with $e_3$) and the pair transforms it into the expression $e_1 \cdot (e_2 + e_3)$ as dictated by the target pattern. Note that in this example, the set of expressions that match the source pattern is not a regular tree language. We propose a model of tree transformations given by a finite representation of a (possibly infinite) set of such (source pattern, target pattern) pairs. The expressive power of this model comes at the cost of undecidability of checking equivalence. Nevertheless, we show that the type-checking problem is decidable for our model of pattern-based tree transformations. The type-checking problem asks whether applying a given transformation to trees having a given regular property (type) preserves the property. Our decision procedure is by a reduction to the emptiness problem of alternating tree automata.

  PDF

  Summarize this
• arXiv cs.AIOct 8

  ReCast: Attribution-Oriented Step Representation Learning for LLM-Based Agent Systems

  Weilin Jin, Mingyu Wang, Taiyu Zhu, Ziqi Zhou, Wenbo Li +5

  In LLM-based agent systems, failures can originate from early steps whose effects propagate through subsequent interactions, making their origins difficult to identify. To trace such failures back to their origin, failure attribution has been formulated as the task of identifying the earliest step responsible for the failure. Recent methods leverage LLM internal signals for failure attribution, typically using hidden states as step representations. We therefore conduct an empirical study to evaluate how effectively these representations distinguish root-cause steps from other steps and find limited separation. Motivated by this observation, we propose ReCast, a step representation learning method that transforms hidden states from a frozen LLM into attribution-oriented step representations. ReCast first selects attribution-relevant layers, then constructs complementary pattern and deviation features, and finally learns contextualized step representations through an encoder trained with contrastive and ranking objectives. We also introduce ReCast-2K, a training dataset for failure attribution. ReCast achieves the best Hit@1 across four benchmarks, surpassing the strongest baseline…

  PDF

  Summarize this
• arXiv cs.AIOct 8

  TokenBank: Financial Infrastructure for AI Services

  Cary Chang, Jialin Zhou

  AI services incur inference costs during execution, while revenue may arrive later. Changing API prices, limited upfront capital, and service failures can limit operators' ability to sustain or expand their services. Beyond reducing per-request costs, operators need to plan future spending, fund execution before revenue arrives, and obtain compensation for specified losses. This requires clear agreements across services with different pricing and execution conditions. These agreements must distinguish rights to consume services from rights to receive payments, define obligations under uncertain costs and income, and specify which failures qualify for compensation and how much can be paid. We present TokenBank, a financial infrastructure that represents these commitments through structured contracts. It supports service-consumption rights, agreements that settle API-price differences in cash (forwards), financing through limited rights to future service revenue, and protection claims for specified service failures. Contracts specify participants, covered services, validity, ownership, fulfillment conditions, and settlement rules. Evaluation combines replay of 899,441 API requests,…

  PDF

  Summarize this
• arXiv cs.CLOct 8

  ReCal: Calibrating Structured Pruning for On-Policy Distillation Recovery

  Houcheng Jiang, Mao Zheng, Mingyang Song, Qiyong Zhong, Jie Sun +2

  Structured pruning reduces the deployment cost of reasoning language models, but the resulting capability degradation can hinder subsequent on-policy distillation (OPD) recovery. Because OPD relies on student-generated trajectories, pruning damage that persists after offline distillation can limit its effectiveness. We propose RECAL, Recovery-Aware Calibration, a simple plug-and-play approach that improves OPD recovery by adjusting calibration before pruning. RECAL uses forward KL between an unpruned teacher and a pruned probe to identify teacher-supported predictions disrupted by pruning, then reweights calibration statistics to guide existing pruning criteria toward preserving these predictions. Across multiple models and pruning methods, RECAL consistently improves mathematical reasoning after OPD, achieving gains of up to 16.7 percentage points on AIME, alongside improvements in most code-generation comparisons. Further analysis shows that RECAL reduces residual damage at heavily affected tokens and establishes performance advantages that persist through recovery. These results demonstrate the value of recovery-aware calibration for improving on-policy distillation recovery…

  PDF

  Summarize this
• arXiv cs.CVOct 8

  PathLang: A Language-Centered Benchmark for Vision-Language Models in Computational Pathology

  Fanqi Cheng, Kuo Gong, Shangke Liu, Beidi Zhao, Junchao Zhu +13

  Pathology vision-language models (VLMs) have shown strong visual perception ability, but their robustness in the language domain remains poorly characterized. Existing pathology VLM benchmarks largely rely on canonical closed-set prompts or perturb only generic templates, treating language as a fixed evaluation component rather than a variable axis of model behavior. In clinical practice, however, diagnostic language varies across reports, institutions, and candidate diagnoses. We introduce PathLang, a language-centered and clinically grounded zero-shot benchmark. PathLang holds the underlying slides, ground-truth labels, and image-text evaluation direction fixed while systematically varying only the diagnostic language, so that performance differences reflect how a diagnosis is phrased rather than what is imaged. The language variation follows how pathologists actually rephrase diagnoses (terminology, specificity, and reporting style), and all prompts and candidate pools are validated by six board-certified pathologists. PathLang covers four task families: (1) zero-shot classification with image-text alignment analysis, (2) cross-modal retrieval, (3) paraphrase robustness,…

  PDF

  Summarize this
• arXiv cs.AIOct 8

  Mine Odyssey: Benchmarking Spatial Agentic Intelligence in the Wild

  Yuxuan Cao, Junlong Li, Hao Li, Junxian He

  Advances in foundation models are driving efforts to introduce agents to assist people in the physical world. Such agents require agentic spatial intelligence: exploring unfamiliar environments, updating spatial understanding through interaction, and adapting actions based on feedback to sustain progress toward a sequence of goals. Existing benchmarks cover only a limited range of spatial layouts, scales, and traversal requirements. We introduce Mine Odyssey, a benchmark for evaluating agentic spatial intelligence using Minecraft reconstructions of real-world locations. It comprises 180 tasks covering 30 such locations across 20 countries and regions on five continents, including 20 outdoor and 10 indoor settings. These settings span diverse spatial scales, layouts, terrains, and connectivity patterns, from Midtown Manhattan and rural Entrup to Santa Lucía Hill and Buckingham Palace. We select meaningful waypoints, such as landmarks, buildings, and rooms, and manually verify their accessibility. Each task provides a natural-language instruction specifying which waypoints to visit and in what order. Completing these tasks requires agents to find accessible routes and entrances,…

  PDF

  Summarize this
• arXiv cs.CVOct 8

  It's Always 10:10: Reference Images Break a Bias That Prompts Only Dent

  Luca Cazzaniga

  Text-to-image models appear to reproduce the habits of the photographs they learned from. Analog clocks are an extreme case: in advertising, watches almost always show 10:10, and generated clocks return to 10:10 even when another time is requested. We measure this bias and test three ways of overcoming it on 52 models available on the Magnific platform, with a replication on Higgsfield. Every image shows three identical clocks that must show 2:35, 6:50 and 11:20. The description of the object is fixed and only the request about the time changes: no time (A), the time in digits (B), the hand positions described by construction relative to the dial numerals (C), or the same description plus a drawn reference dial (D). Two AI readers read all 1,799 images blind from coded copies, with a third reader and the author settling disagreements (dial-level agreement 96.0% and 97.3%). With no time requested, 67% of the images have all three clocks at 10:10. On the 20 current models, all three clocks are correct in 34% of the images with digits, 30% with the hands described in words and 75% with the reference dial (D-B: +37 points, 95% CI +28 to +45); we found no evidence that describing the…

  PDF

  Summarize this
• arXiv cs.AIOct 8

  Finsler Flow Matching: Dynamics-Aware Geodesic Interpolation for Single-Snapshot Trajectory Inference

  Niklas Canova, Jonas Simon Fleck

  Single-cell snapshot data can resolve a continuum of cellular states but do not uniquely determine the dynamics governing transitions between them. However, additional dynamical information can often be encoded in a cell-cell Markov transition kernel. Existing generative approaches for single cell trajectory inference either infer transport only from population marginals, impose a symmetric geometry on the state space, or incorporate directionality through a single velocity vector at each observed state. We introduce Finsler Flow Matching (FFM), a framework for learning continuous stochastic dynamics from discrete Markov transition graphs. We use the first and second local moments to construct a Finsler structure motivated by the Freidlin--Wentzell action, where the second moment determines anisotropic accessibility and the first moment introduces a preferred direction of motion. We learn neural approximations of the resulting directed geodesics, use their Finsler cost to construct source-target couplings, and define geometry-aware stochastic conditional paths that can be distilled into a continuous generative process through simulation-free score and flow matching. Across…

  PDF

  Summarize this
• arXiv cs.AIOct 8

  DivMoE: Fine-Grained MoE Upcycling via Cross-Domain Expert Composition

  Yuxuan Lou, Kai Yang, Geng Zhang, Yong Liu, Yang You

  Mixture-of-Experts (MoE) architectures have become essential for scaling large language models, with recent work demonstrating the benefits of fine-grained expert designs. Training such models from scratch is expensive, and sparse upcycling from pre-trained dense models is an attractive alternative. However, we identify a structural pathology of fine-grained upcycling: when fine-grained experts are derived from a single source model, naive routing collapses and downstream accuracy drops to near-random (e.g., on Qwen3-1.7B, Drop-Upcycling-fine-grained reaches only 23.2% average accuracy across 15 benchmarks, essentially matching from-scratch training at 22.2%, while the same method's coarse-grained variant reaches 50.2%). We propose DivMoE, the first framework achieving fine-grained MoE Upcycling with structurally-balanced routing. DivMoE introduces domain-specialized fine-grained expert initialization, deriving experts from dense models that have undergone domain-adaptive continual pre-training, and diversity-constrained routing, a hard structural constraint guaranteeing that each token activates experts from distinct domain groups. Across two base models and 15 benchmarks,…

  PDF

  Summarize this
• arXiv cs.CLOct 8

  MetaEncoder: Exploring the Limit of Bi-Encoders for Multimodal System One Decision Making with Natural Language Interface

  Jianpeng Cheng, Guangyu Sun, Aashu Singh, Benyu Zhang, Haixing Dai +9

  System One models output constrained decisions and probability distributions rather than free-form text generation. While prevailing paradigms rely on structured schema objects to encode state, intent, and candidate choices, we revisit a fully natural language-based System One interface. In this framework, both the user request and each candidate option are expressed in natural language, supported by multimodal (image and video) auxiliary inputs. We introduce MetaEncoder, which fine-tunes a pre-trained Muse-Glimmer 30B decoder into an instruction-following decision-making encoder. To scale effectively across both small closed-set (< 256) and massive open-set (millions) candidate spaces, MetaEncoder employs a bi-encoder architecture trained via unidirectional contrastive learning for request-candidate alignment. We conduct extensive evaluations across 11 benchmark suites and 190 tasks spanning multimodal decision-making, understanding (closed-set) and retrieval (open-set), highlighting where MetaEncoder beats SOTA multimodal encoders, as well as its current limits on reasoning-intensive tasks.

  PDF

  Summarize this
• arXiv cs.LGOct 8

  Deflating the Hessian: Rank-4 W4A4 Quantization for Multimodal Diffusion Transformers

  Shiwen Wang, Pengxiang Zhao, Xiaoming Yuan

  In diffusion transformers, low-rank branches can mitigate 4-bit weight--activation (W4A4) post-training quantization (PTQ) loss by decomposing each weight into a low-bit residual and a high-precision low-rank component. Existing low-rank PTQ approaches, however, either optimize low-rank compensation and residual quantization separately, often requiring higher ranks, or rely on second-order weight updates without explicitly modeling activation quantization error, which becomes particularly pronounced under 4-bit quantization. To address these limitations, we present HSVDQuant, a unified framework modeling low-rank-assisted W4A4 PTQ as a coupled calibration problem and deriving optimization-based solvers from the joint objective. Eliminating the output-side low-rank factor yields a deflated Hessian that discounts residual errors already captured by the low-rank component, while an activation-noise surrogate is incorporated to suppress activation quantization error. Across five diffusion backbones, rank-4 HSVDQuant consistently outperforms rank-4 SVDQuant in PSNR and LPIPS. It further surpasses rank-32 SVDQuant on SANA-1.6B, FLUX.1-schnell, and FLUX.1-dev with an $8\times$ smaller…

  PDF

  Summarize this