Showing 1–50 of 3,435 results for author: Zha, W

Contrastive Mixed Prompt Learning for Incomplete Multimodal Sentiment Analysis with Unseen Modality Combination

Authors: Kaixin Xu, NaiJin Liu, Yulin Kang, Tangyue Jin, Zixuan Yu, Wenxi Zhao, Yibei Liu, Qianle Zhang, Yangyang Wu, Mengying Zhu, Meng Xi

Abstract: Incomplete multimodal sentiment analysis has garnered significant attention in recent years. Existing approaches typically assume that data is missing at random or are designed specifically for certain missing patterns, ignoring the modality combination inconsistency between training and testing phases. However, in real-world scenarios, the testing phase often encounters modal combinations that we… ▽ More Incomplete multimodal sentiment analysis has garnered significant attention in recent years. Existing approaches typically assume that data is missing at random or are designed specifically for certain missing patterns, ignoring the modality combination inconsistency between training and testing phases. However, in real-world scenarios, the testing phase often encounters modal combinations that were not present during the training phase, which leads to insufficient generalization capabilities and unstable performance. In this paper, we introduce the problem of Incomplete Multimodal Sentiment Analysis with Unseen Modality Combinations (IMSAUMC), aiming to enhance model generalization for unseen modality combinations. To address this challenge, we propose the model named $\textbf{C}$ontrastive $\textbf{M}$ixed $\textbf{P}$rompt $\textbf{L}$earning ($\textsf{CMPL}$) for IMSAUMC. It introduces a label-guided contrastive feature learning mechanism to learn robust and discriminative cross-modal representations. Additionally, we design modality-combination prompts with a soft router to facilitate better learning of various modality combinations. Furthermore, we introduce three prompt contrastive learning strategies, which enable effective learning of prompts corresponding to unseen modality combinations, thereby significantly strengthening the model's generalization capabilities in diverse testing scenarios. Extensive experiments on three widely used datasets demonstrate that $\textsf{CMPL}$ achieves more than a 5% improvement in accuracy compared to state-of-the-art approaches. △ Less

Submitted 20 August, 2026; originally announced August 2026.

LabDex: A Hierarchical Benchmark for Dexterous Manipulation in Laboratories

Authors: Zhipeng Tang, Sihang Chen, Sha Zhang, Peihao Yang, Yan Liu, Wentao Zhao, Xinrui Liu, Rui Huang, Wensheng Du, Yuting Huang, Jiajun Deng, Lidian Wang, Yuan Zhang, Yanyong Zhang

Abstract: Autonomous laboratories hold great promise for accelerating scientific discovery. To achieve this vision, robots are supposed to dexterously manipulate diverse labware and instruments and execute long-horizon, state-dependent experimental procedures. Yet existing benchmarks do not jointly capture dexterous hand use, real-world laboratory interactions, and multi-stage experimental procedures, limit… ▽ More Autonomous laboratories hold great promise for accelerating scientific discovery. To achieve this vision, robots are supposed to dexterously manipulate diverse labware and instruments and execute long-horizon, state-dependent experimental procedures. Yet existing benchmarks do not jointly capture dexterous hand use, real-world laboratory interactions, and multi-stage experimental procedures, limiting systematic training and evaluation. To bridge this gap, we introduce LabDex, a large-scale real-world dataset and benchmark for dexterous manipulation in chemistry laboratories, organized around a hierarchical task taxonomy spanning atomic skills, compositional tasks, and long-horizon experiments. First, LabDex is cross-platform and, for the first time, unifies real-world and simulation platforms under a common framework, providing standardized task definitions, demonstrations, and evaluation protocols. Second, LabDex is large-scale and systematically organizes chemistry laboratory operations into three interconnected levels: Atomic Skills, which characterize fundamental dexterous manipulation capabilities; Compositional Skills; and Long-Horizon Laboratory Workflows. This hierarchical design not only supports the evaluation of end-task performance, but also enables the analysis of how fundamental dexterous skills compose and influence more complex laboratory operations. We conduct cross-level evaluations of representative robot learning methods in both real-world and simulation environments. The experimental results validate the effectiveness of the LabDex task design and demonstration data, and show that the benchmark supports the training and systematic evaluation of existing robotic policies across laboratory dexterous manipulation tasks at different levels, providing a foundation for further research and development of autonomous laboratory robots. △ Less

Submitted 19 August, 2026; originally announced August 2026.

Comments: 13 pages, 3 figures

Fourier Transforms of Color Glass Condensate Multi-Wilson-Line Correlators via Filon Quadrature

Authors: Haowu Duan, Si-Wei Dai, Cong Yi, Wenbin Zhao

Abstract: Calculating cross sections in the Color Glass Condensate effective theory requires Fourier transforms of multi-Wilson-line correlators from transverse coordinate space to transverse momentum space. Under the common assumption of impact-parameter independence, each transform reduces to a set of Hankel transforms whose Bessel-function kernels oscillate rapidly at phenomenologically relevant momenta,… ▽ More Calculating cross sections in the Color Glass Condensate effective theory requires Fourier transforms of multi-Wilson-line correlators from transverse coordinate space to transverse momentum space. Under the common assumption of impact-parameter independence, each transform reduces to a set of Hankel transforms whose Bessel-function kernels oscillate rapidly at phenomenologically relevant momenta, making direct quadrature prohibitively expensive. We present a Filon-type quadrature, applicable to any integrand, that integrates these oscillatory factors in closed form on the stored coordinate grid, reducing each Hankel transform to a precomputed weight vector and the full nested transform chain to a sequence of matrix products. We develop and validate the method on the deep inelastic scattering dijet cross section beyond the correlation-limit approximation, where an exprel-based reformulation of the quadrupole Wilson-line correlator removes a numerical $0/0$ instability inherent to its standard parametrization. Porting the calculation to the Graphics Processing Unit (GPU), with custom CUDA kernels that fuse the momentum-space contraction directly into the correlator evaluation, brings the runtime for one dipole input down to about two minutes on a single NVIDIA A800, from several hours on a multi-core Central Processing Unit (CPU). We further generalize the algorithm to three sequential Hankel transforms and validate the resulting six-dimensional transform against an analytic Gaussian integrand family with closed-form results at every stage. This general, process-independent algorithm is directly applicable to next-to-leading-order proton-nucleus and electron-ion scattering cross-section calculations performed without the correlation-limit approximation. The code is publicly available at https://github.com/CCNU-CGC-py/FFT_filon. △ Less

Submitted 19 August, 2026; originally announced August 2026.

Comments: 42 pages, 14 figures, 5 tables

ClawGym II: Exploring Black-Box RL on Agent Harness

Authors: Huatong Song, Fei Bai, Ming Yang, Renyuan Li, Jia Deng, Jujie He, Zhange Zhang, Daixuan Cheng, Yan Xing, Qi Yun, Xuxing Chen, Danyang Li, Feng Chang, Chuan Hao, Ran Tao, Jian Yang, Bryan Dai, Wayne Xin Zhao, Mingjie Tang, Ji-Rong Wen

Abstract: Agent harnesses have substantially improved performance on long-horizon tasks by coordinating agent interactions with the environment. However, reinforcement learning through complex harnesses remains largely unexplored, as scaling such training to long-horizon agent tasks introduces fundamental challenges. In this work, we present a unified black-box RL framework for stable and scalable optimizat… ▽ More Agent harnesses have substantially improved performance on long-horizon tasks by coordinating agent interactions with the environment. However, reinforcement learning through complex harnesses remains largely unexplored, as scaling such training to long-horizon agent tasks introduces fundamental challenges. In this work, we present a unified black-box RL framework for stable and scalable optimization of general agents through complex harnesses. Concretely, we first build a sandbox-based execution infrastructure that isolates task environments and harnesses within temporary sandboxes for large-scale concurrent rollouts. We then decouple policy optimization from opaque harness execution and place a serving proxy at the model boundary to capture model calls. To reconstruct multi-turn trajectories and improve training efficiency, we organize the captured calls into prefix trees and further adapt both critic-based PPO and critic-free GRPO to optimize over the recovered tree structure. Meanwhile, we maintain training-inference consistency throughout the optimization process. Finally, we introduce mix-harness training, allowing a single model to be jointly optimized by heterogeneous harnesses. With Qwen3-30A3B, black-box RL improves Pass@1 on ClawGym-Bench by 9.98 and 14.81 points through OpenClaw and Claude Code, respectively, while remaining stable over 200-400 optimization steps. Moreover, the framework yields consistent gains on more challenging tasks such as JobBench and OfficeQA. Overall, our framework enables effective, stable, and scalable optimization of general agents through black-box harnesses, supporting unified training across heterogeneous execution systems. △ Less

Submitted 17 August, 2026; originally announced August 2026.

Physics of Agents: Statistical Mechanics Predicts Collective Behavior of AI Agents

Authors: Batu El, Jinhee Paeng, Fatih Dinc, Shiye Su, Mete Erdogan, Aneesh Pappu, Haotian Ye, Wanjia Zhao, Surya Ganguli, James Zou

Abstract: AI agents increasingly operate as part of interacting systems rather than in isolation. As agents exchange information and jointly make decisions, their interactions can improve collective reasoning but may also produce herding, polarization, or amplify shared biases. Understanding and predicting these collective dynamics is therefore important for designing effective and aligned multi-agent syste… ▽ More AI agents increasingly operate as part of interacting systems rather than in isolation. As agents exchange information and jointly make decisions, their interactions can improve collective reasoning but may also produce herding, polarization, or amplify shared biases. Understanding and predicting these collective dynamics is therefore important for designing effective and aligned multi-agent systems. Here, we study over 10,000 communities of language-model agents that repeatedly exchange messages and revise their opinions across objective mathematics questions and subjective political statements. Despite substantial diversity in possible behavior, the individual and group dynamics can be represented by three characteristic regimes: indifference, polarization, and consensus. AI agents start indifferent and build conviction as they interact. On objective questions, communication improves collective accuracy, while on subjective questions it often drifts group opinions toward the right in the political spectrum. We explain these observations with a statistical-mechanics formalism in which agents stochastically favor lower social pressure. Given only initial opinions, our model predicts individual trajectories, outperforms all standard baselines, generalizes to unseen community graphs, and reproduces the observed group archetype distributions. Our fitted model parameters reveal the mechanics underlying our key observations: i) communities operate below the critical social temperature, which explains conviction buildup; ii) attractive ties outweigh repulsive ones, which favors consensus; and iii) agents holding the correct answer exert the strongest pull, which drives truth-seeking. Overall, our results demonstrate that collective behavior of AI agents, like that of other complex systems, follows compact and predictive dynamical laws. △ Less

Submitted 17 August, 2026; originally announced August 2026.

Comments: 51 pages, 20 figures, 9 tables

Tripwire: Triggering Aligned Refusal via Statistically Certified Safety Neurons

Authors: Wei Zhao, Zhe Li, Peixin Zhang, Jun Sun

Abstract: Neuron- and path-level interventions offer the finest-grained route to defending large language models (LLMs) against jailbreak attacks, yet existing methods fall short of this promise, i.e., they often compromise model utility significantly. Specifically, one line of work suppresses toxic neurons to erase harmful semantics, but since such semantics are distributed across the network, blocking eve… ▽ More Neuron- and path-level interventions offer the finest-grained route to defending large language models (LLMs) against jailbreak attacks, yet existing methods fall short of this promise, i.e., they often compromise model utility significantly. Specifically, one line of work suppresses toxic neurons to erase harmful semantics, but since such semantics are distributed across the network, blocking every pathway forces a large intervention footprint. An alternative line of research focus on identify safety neurons using external classifiers. While promising, the existing approaches suffer from compromising neurons that are important for the model utility as well. Moreover, both approaches remain always on and thus perturb every benign request even when no attack is present. To address these limitations, we present \ours{}, a training-free defense that first identifies safety-specific neurons through per-neuron hypothesis tests under false-discovery-rate control together with a utility-specificity filter. Based on this identification, a trigger-style clamp holds the selected neurons at their harmful-conditional mean activations, injecting an internal harmful-input signal that triggers the refusal behavior learned during alignment. The clamp is then realized by two provably equivalent deployment modes, namely a detector-gated inference-time intervention and an offline bias-patch weight edit. Extensive experiments across four safety-aligned LLMs and four representative attacks demonstrate that \ours{} reduces the average attack success rate to at most 2.0\% while incurring a utility drop of only 0.5\% to 5.3\% on MT-Bench, the smallest among all defenses. Code is available at https://anonymous.4open.science/r/Tripwire-65C4. △ Less

Submitted 14 August, 2026; originally announced August 2026.

Can We Defend Against AI-Generated Video Attacks on Real-World Crisis Events? A Systematic Evaluation of Detectors, Generators and Social Dissemination

Authors: Shuo Liang, Yixing Ma, Pengfei Zhou, Zhenglin Wan, Xingyan Chen, Zihan Mei, Manting Li, Feihan Chen, Zhiwen Wang, Bin Xu, Haotian Zhang, Jiajun Song, Shiya Su, Run Liu, Zhenghang Ni, Yifa Yu, Jintao Hong, Bolong Feng, Yifei Liu, Zirui Zhang, Jingxuan Zhang, Songlin Zhao, Yifan Bai, Kang Tan, Yizhe Liu , et al. (11 additional authors not shown)

Abstract: Recent video generators can fabricate realistic depictions of wars, disasters, public emergencies, and other real-world crises, creating substantial risks of misinformation. Existing benchmarks, however, provide limited evidence on detector and generator behavior in such settings, including how detectability varies with generation conditions, how people perceive generated videos, and whether detec… ▽ More Recent video generators can fabricate realistic depictions of wars, disasters, public emergencies, and other real-world crises, creating substantial risks of misinformation. Existing benchmarks, however, provide limited evidence on detector and generator behavior in such settings, including how detectability varies with generation conditions, how people perceive generated videos, and whether detectors remain reliable during social dissemination. To address this gap, we introduce RA-Bench, a benchmark for AI-generated video detection that uses Real videos as Anchors. RA-Bench contains 17,886 videos, comprising 1,830 real-video anchors across 10 social-risk categories and 16,056 generated clips from four open-source and five closed-source generators. Based on RA-Bench, we organize our evaluation along three dimensions. We first assess detector generalization across seven traditional detectors, ten zero-shot multimodal models under three review settings, and two MLLMs specifically fine-tuned on AI-generated video detection. Across these methods, none of the three detector families generalizes consistently across RA-Bench instances. We then examine how detectability varies with generation quality, conditioning information, and sampling seeds. These analyses show that generation properties affect detector families differently, while source-level detection patterns remain stable across seeds. Finally, we study human authenticity judgments and detector reliability during social dissemination. We find that videos that mislead people are also difficult for current detectors, and that social dissemination makes detection harder. Together, these findings show that current methods struggle to detect realistic AI-generated videos, highlighting the need for detectors robust to evolving video generators. △ Less

Submitted 16 August, 2026; v1 submitted 14 August, 2026; originally announced August 2026.

Comments: 63 pages, 20 figures, 32 tables

How Do VLMs Behave When Blind or Misled? Behavioral Evaluation of VLMs on Scientific Figures

Authors: Paul Osemudiame Oamen, Owusu-Banahene Osei, Ananya Mukherjee, Christian Greisinger, Steffen Eger, Pius Onobhayedo, Wei Zhao

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

Submitted 13 August, 2026; originally announced August 2026.

Comments: 25 pages including appendix. Project website: https://scifigbench.nlp4sci.com

ACM Class: I.2.7; I.2.10

RbFT-Net: Rectify-Before-Fuse Temporal Radar Anchors for 4D Radar-Camera Depth Completion

Authors: Wentao Zhao, Shouxuan Wu, Yongtao Cen, Tianchen Deng, Yuyang Zhang, Jingchuan Wang

Abstract: Dense metric depth prediction from cameras and millimeter-wave radar offers a cost-effective sensing solution for autonomous systems. However, radar measurements are inherently sparse and susceptible to clutter, multipath reflections, and projection errors. While aggregating multiple radar frames provides denser metric cues, it also introduces temporal misalignment and dynamic-object interference.… ▽ More Dense metric depth prediction from cameras and millimeter-wave radar offers a cost-effective sensing solution for autonomous systems. However, radar measurements are inherently sparse and susceptible to clutter, multipath reflections, and projection errors. While aggregating multiple radar frames provides denser metric cues, it also introduces temporal misalignment and dynamic-object interference. Directly propagating such unreliable measurements can therefore corrupt large regions of the predicted depth map. To address this issue, we propose RbFT-Net, an end-to-end rectify-before-fuse framework for multi-frame 4D radar-camera depth completion. Rather than assuming accumulated radar returns to be accurate, RbFT-Net treats them as noisy temporal anchor candidates. An image-conditioned rectification module jointly corrects their image-plane locations and metric depths while estimating pointwise reliability. The rectified anchors are then selectively propagated before high-level multi-modal fusion, suppressing the influence of unreliable measurements. Experiments on ZJU-4DRadarCam and a newly collected 4D radar-camera-LiDAR dataset show that RbFT-Net consistently outperforms the evaluated independent radar-camera methods and remains competitive with plug-in pipelines using auxiliary monocular depth models. Cross-platform evaluation and component analyses further support the effectiveness of the proposed rectification and reliability-aware propagation strategy. △ Less

Submitted 13 August, 2026; originally announced August 2026.

AI4AI at Test-Time: Strong-to-Weak Capability Transfer via Harnesses

Authors: Cheng Qian, Wenting Zhao, Liangwei Yang, Heng Wang, Jielin Qiu, Heng Ji, Silvio Savarese, Huan Wang, Shelby Heinecke

Abstract: Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-time methods. In this paper, we ask whether such transfer can instead occur at test time. We study strong-to-weak scaffolding: whether a stronger builder model can construct inference-time harnesses t… ▽ More Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-time methods. In this paper, we ask whether such transfer can instead occur at test time. We study strong-to-weak scaffolding: whether a stronger builder model can construct inference-time harnesses that help a weaker target model solve tasks more reliably without any parameter updates. Using four representative Theory-of-Mind benchmarks, each builder model uses 5% of the data as a validation set to iteratively refine its harness over multiple rounds, after which the finalized harness is evaluated on the full test set. Empirically, this form of test-time capability transfer is highly effective, nearly doubling average target-model performance from 0.49 to 0.91. Our analysis shows that the gains come primarily from offloading unstable model reasoning into deterministic code, benchmark-specific routing, and strict answer-format enforcement, rather than from encouraging the target model to reason more extensively or sample more broadly. We further find that builder-model reasoning effort improves harness quality monotonically, platform effects are modest relative to the builder model's own capability, and weaker target models receive the largest gains. These results suggest that inference-time harness design is an important complement to conventional training-time distillation, enabling strong models to transfer cognitive structure to weaker models without retraining. △ Less

Submitted 12 August, 2026; originally announced August 2026.

Comments: 23 Pages, 12 Figures, 6 Tables

SCOPE-Router: Cost-Aware Open-Set VLM Routing for Execution-Oriented Tasks

Authors: Tao Yu, Yifei Qu, Zhiqing Cui, Pengfei Zhou, Zhongtian Luo, Yujia Yang, Shenghua Chai, Haopeng Jin, Zhenghao Zhang, Xinming Wang, Hongzhu Yi, Wangbo Zhao, Zhenglin Wan, Yan Huang, Yeshani, Jinwen Luo, Yang You

Abstract: Model routing aims to select the most suitable model from a candidate pool for each query, balancing quality and cost. Existing VLM routing research is limited to traditional VQA evaluation, lacks systematic calibration optimization for open-set scenarios, and employs training objectives that dilute multi-positive signals via softmax normalization without incorporating cost. We address these limit… ▽ More Model routing aims to select the most suitable model from a candidate pool for each query, balancing quality and cost. Existing VLM routing research is limited to traditional VQA evaluation, lacks systematic calibration optimization for open-set scenarios, and employs training objectives that dilute multi-positive signals via softmax normalization without incorporating cost. We address these limitations with three contributions: (1)VLM-ExecRouterBench, the first execution-oriented VLM routing benchmark covering Code, Agentic, and Search domains with 11 candidate models spanning nearly two orders of magnitude in pricing; (2)SCOPE-Router, a dual-tower router that matches queries to model behavior profiles constructed via hybrid calibration (random/diagnostic/diversity sampling), enabling new models to join routing without retraining; (3)CRM+RCCR, an architecture-agnostic cost-aware objective that encodes cost preference into continuous relevance targets through per-pair independent scoring, eliminating multi-positive dilution while regularizing queries with similar routing preferences to be closer in the routing space. Empirically, SCOPE-Router achieves the best Rank Score on all three benchmarks, surpassing the runner-up by 1.84 points under OOD settings and by 6.75 points under doubly OOD open-set evaluation. When applied to four diverse routers, CRM+RCCR improves Rank Score by 1.25--6.21 points. △ Less

Submitted 19 August, 2026; v1 submitted 12 August, 2026; originally announced August 2026.

Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA

Authors: Mind Lab, :, Vin Bo, Asher Cai, Jingwei Cao, Song Cao, Vic Cao, Amelia Chen, Andrew Chen, Kaijie Chen, Cleon Cheng, Steven Chiang, Kaixuan Fan, Hera Feng, Huan Feng, Arthur Fu, Jun Gao, Pyke Han, Nolan Ho, Ori Hong, Hailee Hou, Piers Hua, Charles Huang, Miles Jiang, Nora Jiang , et al. (52 additional authors not shown)

Abstract: Macaron-V1 is an open agent-model family for experiential intelligence: learning from experience in real environments and continuing to learn after deployment. It is organized around two system goals. Adaptation is pursued through recursive improvement of versioned model-harness pairs, where experience from one configuration is evaluated under an external contract and used to construct its success… ▽ More Macaron-V1 is an open agent-model family for experiential intelligence: learning from experience in real environments and continuing to learn after deployment. It is organized around two system goals. Adaptation is pursued through recursive improvement of versioned model-harness pairs, where experience from one configuration is evaluated under an external contract and used to construct its successor. Collaboration is pursued via the Mixture-of-LoRA (MoL) architecture that freezes a base model, composes specialist LoRA adapters, and selects one LoRA per user turn. The flagship Macaron-V1-Venti combines a 744B GLM-5.2 base with four LoRAs for chat, agent, coding, and GenUI; the Qwen3.6-based Macaron-V1-Tall (50B) uses the same design for local deployment. This report presents Macaron-V1 as a co-designed system spanning architecture, algorithms, and infrastructure. The MoL architecture supports continual learning through extensible LoRA specialists. The algorithm combines Model-Harness Co-design and recursive self-improvement loop, including the UI4A component-native GenUI harness, a stateful action substrate, versioned HCP contract, and the agentic RL framework MindForge. The supporting infrastructure includes the post-training platform MinT, the long-context RL method LongStraw, and stability techniques for sparse MoE and DSA base models. We evaluate Macaron-V1 on Personal Intelligence, GenUI, and general capability benchmarks against frontier baselines. Our results validate the current system, while compounding gains from continual learning and collective intelligence remain open questions. △ Less

Submitted 10 August, 2026; originally announced August 2026.

Comments: 49 pages, technical report

Se-DPO: Self-Evolving Token Credit for Direct Preference Optimization

Authors: Wenxiao Zhao, Shu Wang, Ying Nian Wu

Abstract: Direct Preference Optimization (DPO) aggregates token-level log-probability ratios via uniform summation, implicitly treating all tokens as contributing equally to the preference signal. However, the contribution of individual tokens to the preference signal varies. We introduce token credit, which modulates each token's KL regularization based on its contribution to the preference outcome. We der… ▽ More Direct Preference Optimization (DPO) aggregates token-level log-probability ratios via uniform summation, implicitly treating all tokens as contributing equally to the preference signal. However, the contribution of individual tokens to the preference signal varies. We introduce token credit, which modulates each token's KL regularization based on its contribution to the preference outcome. We derive that effective token credit is proportional to the magnitude of each token's implicit reward, and observe that this quantity evolves substantially during training. This implies that static token credit becomes increasingly misaligned as training progresses. In this work, we propose Se-DPO (Self-Evolving Token Credit for DPO), a live mechanism that derives token credit from the model's own evolving internal signals during DPO training. Since the reward signal varies in reliability across positions, Se-DPO calibrates token credit based on both the strength and the confidence of each token's contribution. Se-DPO requires no external models, adding only a lightweight calibration network with minimal computational overhead. Experiments show that Se-DPO improves over DPO by up to 9.8 points on AlpacaEval~2 and 12.2 points on Arena-Hard. △ Less

Submitted 10 August, 2026; originally announced August 2026.

Comments: 16 pages, 2 figures, COLM2026

Is the ACL Responsible NLP Checklist a Box-Ticking Exercise? A Large-Scale Analysis of EMNLP 2025

Authors: Nusrath Jinnath, Wei Zhao

Abstract: Responsible NLP practice includes a) transparency, b) ethics, and c) societal impacts. The Responsible NLP Checklist aims to push these goals, and promote responsible practice. Recently, ACL released the EMNLP 2025 Checklists to aid transparency on the current research practice, which we focus on. We curate and release the first two datasets of: a) all the checklist responses and justifications fr… ▽ More Responsible NLP practice includes a) transparency, b) ethics, and c) societal impacts. The Responsible NLP Checklist aims to push these goals, and promote responsible practice. Recently, ACL released the EMNLP 2025 Checklists to aid transparency on the current research practice, which we focus on. We curate and release the first two datasets of: a) all the checklist responses and justifications from the EMNLP 2025 Main and Finding tracks; b) checklist reference linking to paper sections. We also provide the first analysis of recent EMNLP Checklists, by examining $73,922$ responses and justifications to them. For the Main track, we find that authors isolate ethics questions of the Checklist from the paper's bulk, mimicking the trend of ethics being an afterthought. We then examine \texttt{NO} responses. We find $44.9\%$ of justifications are poor or bad-faith, being brief or empty. Then, we find significant issues with the checklist design and effort of authors, namely that $6\%$ of all checklists contained logical contradictions between parent and child responses. We also find evidence of surface compliance for responsible ethics, with $53\%$ authors dismissing potential risks or social impacts of their work, for which there should be none. We compare this to the Findings track, noticing a similar trend in both tracks. Lastly, we discuss the implications of the checklist design and provide recommendations for future checklist iterations. Including: a) enforcing a minimum word count, b) enforcing more scrutiny on the risks of appliances. △ Less

Submitted 10 August, 2026; originally announced August 2026.

Evo-Bench: Can Language Models Improve Agent Harness?

Authors: Lisheng Huang, Chen Yang, Hao Zhou, Huatong Song, Zongchao Chen, Ran Le, Yang Song, Wayne Xin Zhao, Tao Zhang

Abstract: Large Language Models (LLMs) have driven rapid progress in autonomous agents, yet standard evaluations remain confined to static task solving. An emerging frontier is harness evolution---the agent's capacity to autonomously optimize its own operating harness. However, systematically benchmarking this capability remains challenging, as existing evaluations fail to isolate harness improvements from… ▽ More Large Language Models (LLMs) have driven rapid progress in autonomous agents, yet standard evaluations remain confined to static task solving. An emerging frontier is harness evolution---the agent's capacity to autonomously optimize its own operating harness. However, systematically benchmarking this capability remains challenging, as existing evaluations fail to isolate harness improvements from base model strength, prevent task-specific overfitting, or capture long-horizon iterative research. To address these challenges, we introduce Evo-Bench, the first benchmark designed to evaluate models' intrinsic harness-evolving capabilities across Search, Office, and General agent domains. To rigorously isolate this capability, Evo-Bench employs a novel harness-guided construction framework: it leverages auxiliary-task evolution to identify tasks genuinely sensitive to framework improvements, followed by sensitivity-aware stratified splitting to ensure robust cross-suite generalization. Extensive evaluations across nine frontier and open-weight models reveal that top models achieve massive absolute gains reaching 16.6 points, closely approaching state-of-the-art human-engineered baselines. Crucially, while autonomous evolution outpeforms artificial harness in General tasks and excels in Search tasks, it struggles in Office tasks that demand highly specific processing workflows. Furthermore, our analysis exposes critical temporal anomalies like early saturation, while demonstrating that the synthesized harnesses act as highly transferable reasoning structures, consistently boosting diverse policy models. △ Less

Submitted 10 August, 2026; v1 submitted 9 August, 2026; originally announced August 2026.

Improving Generalization Robustness of Multimodal RLVR

Authors: Pengfei Zhou, Zhiwei Tang, Xiaopeng Peng, Chenrui Zhou, Lama Moukheiber, Yixing Ma, Bin Xu, Jiajun Song, Zhenglin Wan, Wangbo Zhao, Jiasheng Tang, Bohan Zhuang, Fan Wang, Yang You

Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) makes Multimodal Large Language Models more accurate, but the gains are brittle: simply paraphrasing a question or changing the prompt template can degrade them, which challenges reliable deployment in high-stakes scenarios like medical VQA. We trace this to two issues of the standard RL objective. First, the binary verifier conflates format wi… ▽ More Reinforcement Learning with Verifiable Rewards (RLVR) makes Multimodal Large Language Models more accurate, but the gains are brittle: simply paraphrasing a question or changing the prompt template can degrade them, which challenges reliable deployment in high-stakes scenarios like medical VQA. We trace this to two issues of the standard RL objective. First, the binary verifier conflates format with content, so the reward signal cannot tell a wrong answer apart from a misformatted one. Second, the training distribution covers only a thin slice of the real-world prompts that the model might meet at deployment, so policies that perform well on the training distribution can behave differently under unseen prompts during test. Both failures call for a robust post-training method that helps the policy cover a broader distribution of semantically equivalent prompts, and we identify two measures that help achieve this objective: separating format from semantics in the reward, and applying policy invariance across perturbed prompts with equivalent semantics. We therefore propose Prompt-Invariant RLVR (PIRL), consisting of a dynamic trinary reward and a consistency regularizer based on an embedding-space adversary. Under stress testing, PIRL's average accuracy on benchmarks drops by only $\le 1\%$, where GRPO drops ~3%. On dynamic evaluation, PIRL also achieves the smallest performance drop. △ Less

Submitted 14 August, 2026; v1 submitted 9 August, 2026; originally announced August 2026.

Comments: 32 pages, 5 figures

BOUND: Brief-Guided Corrective Preference Distillation at Search-Control Boundaries

Authors: Qingying Niu, Ruiyang Ren, Wayne Xin Zhao, Yaliang Li

Abstract: Large language model (LLM)-based deep search agents solve tasks through iterative retrieval and reasoning, but locally relevant evidence can cause persistent wrong-anchor drift, constraint drift, or local-topic drift. Existing methods supervise trajectories, outcomes, or steps, but rarely distinguish task-aligned continuations from locally plausible ones that reinforce drift. We propose BOUND, a b… ▽ More Large language model (LLM)-based deep search agents solve tasks through iterative retrieval and reasoning, but locally relevant evidence can cause persistent wrong-anchor drift, constraint drift, or local-topic drift. Existing methods supervise trajectories, outcomes, or steps, but rarely distinguish task-aligned continuations from locally plausible ones that reinforce drift. We propose BOUND, a brief-guided corrective preference distillation framework for persistent search drift. For each student-induced decision-time state, BOUND constructs a teacher-side search-state brief that preserves the original search target and key constraints while summarizing confirmed evidence, missing information, and drift status. Guided by the brief, the teacher determines whether the student's continuation contains a correctable local search-control error likely to affect subsequent decisions. Together with the rollout outcome, this assessment determines whether to construct a corrective contrast between a student-specific correction and the original continuation, or a termination contrast between a supported answer and an unnecessary retrieval continuation. Each validated state-matched preference pair operationalizes a search-control boundary. Direct preference optimization (DPO) distills these preferences into the student, while the brief and teacher-side computation remain confined to training. We evaluate BOUND on four multi-hop QA benchmarks and three deep-search benchmarks. Across the six benchmarks for which we reran baselines, BOUND leads on five datasets and 12 of 14 metrics. Under the same search-control interface and matched settings, BOUND outperforms Trajectory SFT by 5.6 EM points on Bamboogle and 4.8 accuracy points on BrowseComp-Plus. Code is available at https://github.com/RUCAIBox/BOUND. △ Less

Submitted 9 August, 2026; originally announced August 2026.

Comments: 15 pages

Goal-oriented Navigation Instruction Generation with Tour Video Priors

Authors: Fangdi Li, Juncheng Liao, Changxu Cheng, Jiazhi Wang, Senda Chen, Tao Wang, Wuyue Zhao

Abstract: Navigation Instruction Generation (NIG) aims to produce step-by-step natural language instructions for navigation guidance. Existing studies primarily treat NIG as an auxiliary task for vision-andlanguage navigation (VLN), focusing on data augmentation or multi-task learning. However, generating navigation instructions from compact environmental priors requires meticulous spatial reasoning, especi… ▽ More Navigation Instruction Generation (NIG) aims to produce step-by-step natural language instructions for navigation guidance. Existing studies primarily treat NIG as an auxiliary task for vision-andlanguage navigation (VLN), focusing on data augmentation or multi-task learning. However, generating navigation instructions from compact environmental priors requires meticulous spatial reasoning, especially when the target route does not simply follow the demonstrated tour, and remains challenging for current multimodal models. In this work, we introduce VideoNIG, a goal-oriented video-grounded NIG task that generates navigation instructions from ego-centric tour videos, an initial observation, and a textual or visual goal, without relying on intermediate representations such as graphs and maps. We instantiate VideoNIG in a controlled simulator benchmark with 60K tour videos across continuous indoor environments and 37K multimodal prompts with progressive difficulty levels. We further introduce a diagnostic evaluation protocol that combines text similarity, choice-based spatial consistency tests, and downstream navigation execution. To address this task, we propose a two-stage Curriculum Learning framework that decomposes the learning into foundational motion perception and long-horizon navigation reasoning. Specifically, we first employ Action Warmup for spatial action-view alignment, followed by Complexity Progression using trajectories with increasing exploratory difficulty. Extensive experiments show that existing MLLMs struggle with VideoNIG, while our approach significantly improves instruction quality across complementary diagnostic metrics. Finally, integrating VideoNIG-generated instructions with a VLN agent demonstrates the executability of this task formulation for end-to-end navigation. △ Less

Submitted 9 August, 2026; originally announced August 2026.

A Note on Market Segmentation and Bertrand Competition

Authors: Zhang Xu, Mingsheng Zhang, Wei Zhao

Abstract: In this note, we show that equilibrium profit is zero in Bertrand competition with a finite number of firms and consumers whose willingness to pay are bounded, under any market segmentation profile. In this note, we show that equilibrium profit is zero in Bertrand competition with a finite number of firms and consumers whose willingness to pay are bounded, under any market segmentation profile. △ Less

Submitted 8 August, 2026; originally announced August 2026.

Unified Hallucination Fuzzing for Multimodal Large Language Models

Authors: Pengfei Zhou, Jiajun Song, Zhiwei Tang, Yixing Ma, Xiaopeng Peng, Donghui Si, Yuhang Xu, Huiqi Song, Yiyuan Miao, Yichen Qian, Weihua Chen, Wangbo Zhao, Bohan Zhuang, Jiasheng Tang, Yang You

Abstract: Hallucination remains a persistent challenge for Multimodal Large Language Models (MLLMs), severely limiting their reliability in high-stakes applications. Existing evaluations, predominantly based on static benchmarks, suffer from narrow taxonomical coverage and rapid performance saturation, failing to reflect model robustness in evolving real-world scenarios. To bridge this gap, we present a sys… ▽ More Hallucination remains a persistent challenge for Multimodal Large Language Models (MLLMs), severely limiting their reliability in high-stakes applications. Existing evaluations, predominantly based on static benchmarks, suffer from narrow taxonomical coverage and rapid performance saturation, failing to reflect model robustness in evolving real-world scenarios. To bridge this gap, we present a systematic evaluation framework integrating a comprehensive benchmark with self-evolving stress testing. First, we introduce UniHall, a fine-grained dataset grounded in a unified taxonomy spanning Object, Instruction, and Knowledge dimensions. Second, to address benchmark saturation, we propose Self-Adaptive Multimodal Fuzzing (SAMF), a self-adaptive framework that employs evolutionary mutation strategies to explore the boundaries of model hallucinations. Crucially, to ensure reliable assessment of dynamic inputs, SAMF incorporates a structured metric suite driven by an ensemble of multi-modal oracles. Our extensive experiments reveal that state-of-the-art MLLMs exhibit significant performance degradation under fuzzing compared to conventional settings, exposing a dissociation between reasoning capabilities and factual grounding. Furthermore, we identify a helpfulness-hallucination trade-off, where reinforcement learning alignment inadvertently exacerbates sycophancy in instruction-following tasks. The framework, code and benchmark are available at https://github.com/LanceZPF/EvalHall. △ Less

Submitted 15 July, 2026; originally announced August 2026.

Comments: 47 pages, 17 figures

Teacher Retains Full Tokens, Student Merges Efficiently: TM20K for E-Commerce Sequence Modeling in Ad Recommendation

Authors: Xinchun Li, Duoru Zheng, Wenlin Zhao, Haoran Ding, Ziyi Zhou, Jingxuan Tan, Huizhi Yang, Yuchen Jiang, Zhe Chen, Yuchao Zheng, Linlan Chen, Dongjian Wang, Dongyue Wang, Xiaosong Li, Hongyue Mao, Yaocheng Tan

Abstract: Benefiting from ultra-long behavior sequence modeling, existing recommender systems bring users a better experience via simultaneously considering their long-term and short-term interests. Nevertheless, extended sequence lengths introduce substantial burdens on training efficiency and serving throughput. Prior approaches typically utilize search-based or cluster-based compression on ultra-long seq… ▽ More Benefiting from ultra-long behavior sequence modeling, existing recommender systems bring users a better experience via simultaneously considering their long-term and short-term interests. Nevertheless, extended sequence lengths introduce substantial burdens on training efficiency and serving throughput. Prior approaches typically utilize search-based or cluster-based compression on ultra-long sequences at the cost of fine-grained information, or rely on various lightweight target attention structures incapable of sufficient sequential feature extraction. In this paper, we balance the effectiveness and efficiency for ultra-long sequence modeling via full transformer modeling accompanied with a two-stage knowledge distillation framework. First, both teacher and student models take the full attention mechanism rather than pure target-sequence attention for effective sequence scaling. For student models, we propose several simple yet well-motivated token merge approaches, significantly compressing the sequence length while maintaining an acceptable performance. Then, a one-time teacher is heavily trained with full sequence tokens, further boosting the performance of student models via knowledge distillation. The proposed paradigm named TM20K has been successfully deployed in ByteDance's e-commerce advertising recommender system that extends the e-commerce sequence length to 20K, delivering substantial improvements in key business metrics (e.g., ADSS +1.036\%) while keeping the training and serving cost nearly the same as the online state-of-the-art model (e.g., serving latency only +5.6\%). △ Less

Submitted 13 August, 2026; v1 submitted 7 August, 2026; originally announced August 2026.

Comments: ByteDance 20K Ultra-long Sequence Modeling for Ad E-Commerce Recommendation

StateFlow: Sequence Pipeline Parallelism for Long-Context Modeling with Linear Recurrence

Authors: Wenxuan Zhao, Yingfa Chen, Xu Han, Wenjing Han, Tianbo Huang, Zhiyu Li, Ao Sun, Jingheng Xu, Lin Gan, Guangwen Yang

Abstract: Long-context training is increasingly important for large language models, and linear attention and state space models have become popular for improving long-context efficiency. However, efficiently parallelizing long-sequence training for recurrent and hybrid models remains challenging. We present StateFlow, a sequence pipeline parallelism system for models with linear recurrence. StateFlow par… ▽ More Long-context training is increasingly important for large language models, and linear attention and state space models have become popular for improving long-context efficiency. However, efficiently parallelizing long-sequence training for recurrent and hybrid models remains challenging. We present StateFlow, a sequence pipeline parallelism system for models with linear recurrence. StateFlow partitions each sequence into chunks and schedules their execution while propagating boundary states and gradients across chunks, thereby reducing activation lifetimes and improving training throughput. StateFlow further uses profile-guided nonuniform chunking to balance recurrence and softmax attention computation in hybrid models, and overlaps state transitions that expose limited parallelism with surrounding computation. Applying StateFlow to models with up to 32B parameters and 256K context length, we achieve up to \(2.22\times\) throughput improvements and \(2.45\times\) memory reduction compared to conventional pipeline parallelism, enabling otherwise infeasible configurations. △ Less

Submitted 7 August, 2026; originally announced August 2026.

CalibForge: Adversarial Solver Calibration for Scaling Learnable Terminal Tasks

Authors: Fanzhe Meng, Guoxin Chen, Jiale Zhao, Shuang Sun, Zhiyu Lin, Wayne Xin Zhao, Ruihua Song, Ji-Rong Wen, Kai Jia

Abstract: Training terminal agents requires executable and verifiable tasks that are not merely solvable, but appropriately challenging for learning. Executable validation establishes feasibility, yet does not reveal how a task behaves relative to a given solver setting. In this paper, we present CalibForge, an autonomous terminal-task synthesis system that uses verified solver behavior to revise candidate… ▽ More Training terminal agents requires executable and verifiable tasks that are not merely solvable, but appropriately challenging for learning. Executable validation establishes feasibility, yet does not reveal how a task behaves relative to a given solver setting. In this paper, we present CalibForge, an autonomous terminal-task synthesis system that uses verified solver behavior to revise candidate tasks through adversarial solver calibration. Multi-solver calibration targets disagreement within a heterogeneous solver pool, whereas contrastive solver calibration targets a designated strong-pass/weak-fail relation; both operationalize a solver-relative learnable zone anchored in demonstrated solvability. Using CalibForge, we construct 5,431 calibrated terminal tasks. Our ablations show that both strategies yield more effective supervision than authoring and validation alone or ordinary single-solver feedback. Models trained on the full collection achieve 32.58% and 47.57% on Terminal-Bench 2.0. The largest improvements over the corresponding base model reach 24.71 percentage points on Terminal-Bench 2.0, 27.68 points on SWE-bench Pro, and 30.04 points on Doc2Repo. Together, these results support solver-relative learnability as a practical target for constructing effective and transferable agent training data. △ Less

Submitted 6 August, 2026; originally announced August 2026.

Comments: Dataset: https://huggingface.co/datasets/AweAI-Team/CalibForge. Repository: https://github.com/AweAI-Team/CalibForge

ePIC Early Science Report

Authors: D. Abbott, N. Abdelrahman, S. Abhijit, I. Abualrob, R. B. Achari, J. Adam, L. Adamczyk, K. Adkins, A. Affolder, K. Agarwal, J. Agarwala, N. Agrawal, C. A. Aidala, W. Akers, A. Al-bataineh, S. N. Alam, M. Alekseev, P. R. Altieri, J. -S. Alvarado Gallenao, S. B. L. Amar, R. Ammendola, I. Amos Cali, G. An, D. Anderson, E. Anderssen , et al. (774 additional authors not shown)

Abstract: This Early Science Report from the ePIC Collaboration outlines the compelling physics program achievable during the first years of operation of the Electron-Ion Collider (EIC), prior to the establishment of the full design luminosity and energy range. The analyses are based on realistic early-running beam configurations and detailed Geant4 ePIC detector simulations, hit digitization and data recon… ▽ More This Early Science Report from the ePIC Collaboration outlines the compelling physics program achievable during the first years of operation of the Electron-Ion Collider (EIC), prior to the establishment of the full design luminosity and energy range. The analyses are based on realistic early-running beam configurations and detailed Geant4 ePIC detector simulations, hit digitization and data reconstruction. The projected studies from the physics working groups of ePIC span inclusive, semi-inclusive, exclusive, diffractive and tagging, as well as jet and heavy flavor measurements in both electron-proton and electron-ion collisions. Even before the collider reaches its full design performance, these measurements will constrain parton distribution functions in nucleons and nuclei, access transverse-momentum-dependent and spin-dependent observables, probe gluon dynamics in nuclei, and initiate a program of imaging of quarks and gluons. Each measurement is directly connected to the core science pillars of the EIC, identified in the 2018 report by the National Academy of Sciences: understanding the origin of the nucleon mass, unraveling the spin structure of the nucleon, and exploring the emergent properties of dense gluonic matter. The results presented here provide examples that demonstrate that the early years of EIC running with ePIC will deliver novel world-leading insights into Quantum Chromodynamics. In addition, the early science program will establish measurement and analysis methodologies that will pave the way to the subsequent full EIC physics program. △ Less

Submitted 5 August, 2026; originally announced August 2026.

Report number: epic-AN-AC-2026-004

DreamWAM: Beyond RGB Future Prediction for World Action Models

Authors: Shanglin Yuan, Weiheng Zhao, Xin Shi, Haoyi Jiang, Xianda Guo, Liu Liu, Wenyu Liu, Wei Sui, Xinggang Wang

Abstract: World Action Models (WAMs) learn action-relevant representations by predicting how the observed world will evolve. Most existing WAMs define this future in RGB space, where task-relevant state transitions are entangled with nuisance variations in texture, illumination, background, and viewpoint. We argue that WAMs should explicitly predict action-relevant future state rather than relying on RGB pr… ▽ More World Action Models (WAMs) learn action-relevant representations by predicting how the observed world will evolve. Most existing WAMs define this future in RGB space, where task-relevant state transitions are entangled with nuisance variations in texture, illumination, background, and viewpoint. We argue that WAMs should explicitly predict action-relevant future state rather than relying on RGB prediction alone. We introduce DreamWAM, which reformulates future prediction as structured world modeling beyond RGB, representing future states through complementary views of appearance, motion, geometry, and semantics. During training, DreamWAM combines joint latent denoising of RGB and motion with lightweight gated residual branches for geometry and semantics. Shared attention between VideoDiT and ActionDiT allows the action branch to learn from these future-state predictions, while all beyond-RGB supervision branches are disabled at inference and deployment remains RGB-only. Across both no-rollout and joint video-action inference, DreamWAM consistently improves the matched RGB-only baselines on LIBERO, from 97.30\% to 98.40\% and from 98.00\% to 98.90\%, respectively. The gains become larger under unseen LIBERO-Plus perturbations, from 51.36\% to 63.44\% and from 69.16\% to 75.47\%. The same robustness extends to real-world manipulation, where DreamWAM attains an average success rate of 74.4\% across unseen changes in lighting, background, and object layout, compared with 55.6\% for Fast-WAM-Joint. These results show that robust world-action learning depends not only on predicting the future, but on representing it in a form that matters for action. The code and models are publicly released at https://github.com/hustvl/DreamWAM. △ Less

Submitted 5 August, 2026; originally announced August 2026.

A-SR: Self-Evolving Agentic LLMs for Symbolic Regression via Hierarchical Coordination

Authors: Wenxiao Zhao, Dong Liu, Kaiyi Xu, Feng Liu, Zhen Zhao, Fei Ben, Shu Wang, Wenhao Li, Ying Nian Wu, Fenghua Ling, Haobo Li, Lei Bai

Abstract: Symbolic regression aims to discover closed-form equations from data, but existing LLM-guided methods often rely on a unified proposal loop that compresses heterogeneous search failures into a scalar score and a single prompt. We propose A-SR, a self-evolving agentic framework that shifts the control unit from expression edits to role-conditioned evidence views. A-SR coordinates formula discovery… ▽ More Symbolic regression aims to discover closed-form equations from data, but existing LLM-guided methods often rely on a unified proposal loop that compresses heterogeneous search failures into a scalar score and a single prompt. We propose A-SR, a self-evolving agentic framework that shifts the control unit from expression edits to role-conditioned evidence views. A-SR coordinates formula discovery through routing among coordination protocols, an online evaluator-reward role policy, and state-routed process memory. During search, evaluator feedback characterizes reliability and productivity, updates role-level utilities, and routes elite motifs, failure traces, and validity diagnostics to different agents. The framework self-evolves at two timescales: within a run, it adapts the search process without updating LLM parameters; across runs, recorded trajectories can be distilled into open-source LLMs as role-conditioned proposal priors. Averaged over the four LSR-Synth scientific domains in LLM-SRBench, A-SR improves Acc@0.01 over baselines from 25.79% to 48.30% with Llama3.1-8B, while A-SR-LoRA improves the corresponding Qwen3-4B result from 24.58% to 38.29%. On four real-world scientific discovery tasks, A-SR obtains the best in-distribution or out-of-distribution normalized mean squared error on 7 of 8 reported metrics. △ Less

Submitted 6 August, 2026; v1 submitted 5 August, 2026; originally announced August 2026.

Comments: 18 pages, 8 figures, including appendix

Enabling Urgency-aware Robot Swarm Intralogistics using Smart IoT Tags

Authors: Youssef Alboraei, Murray Groves, Shane Wen, Wenda Zhao, Senhui Qiu, Mohammud J. Bocus, Robert Piechocki, Sabine Hauert, Kerstin Eder

Abstract: Warehouse items differ in how urgently they must be moved: perishable goods, pharmaceutical shipments, and just-in-time production materials must be delivered sooner than the rest of the stock. Decentralised robot swarms suit warehouses that cannot justify fixed automation infrastructure, but current swarm controllers treat all items alike or rely on an external scheduler to set priorities, so urg… ▽ More Warehouse items differ in how urgently they must be moved: perishable goods, pharmaceutical shipments, and just-in-time production materials must be delivered sooner than the rest of the stock. Decentralised robot swarms suit warehouses that cannot justify fixed automation infrastructure, but current swarm controllers treat all items alike or rely on an external scheduler to set priorities, so urgent items wait as long as ordinary ones. This paper presents a swarm logistics system in which each warehouse carrier holds an ultra-low-power Internet-of-Things (IoT) tag that broadcasts the urgency of its item over Bluetooth Low Energy (BLE). Robots read these broadcasts directly and weigh urgency against travel distance when choosing which carrier to serve, so prioritisation happens at the item level without central scheduling. The system is evaluated in simulation and validated on real robots and IoT-tagged carriers against a proximity-only baseline. In the physical trials, priority alignment (i.e. proportion of urgent items served first), improved from 0.41 to 0.64, with a nonsignificant trend toward lower 95th-percentile (P95) delivery latency and throughput within 1.2% of the baseline. In simulation, the benefit grew with system size: across three larger configurations, P95 latency fell by 5.2% to 11.8% and priority alignment improved by 41.7% to 51.6%. Attaching urgency to the items themselves therefore allows a decentralised swarm to serve time-critical stock sooner while keeping the low infrastructure requirements that make swarm systems attractive for warehouse automation. △ Less

Submitted 5 August, 2026; originally announced August 2026.

Comments: 16 pages, 8 figures, 3 Tables

Faster-WAM: Efficient Inference-Time Future Conditioning for Robust World Action Models

Authors: Weiheng Zhao, Haoyi Jiang, Xin Shi, Liu Liu, Fan Huang, Zhizhong Su, Wei Sui, Xinggang Wang

Abstract: World Action Models (WAMs) improve robot manipulation by learning how the environment evolves beyond the current observation. However, existing approaches face a fundamental dilemma: Joint-WAMs preserve future-aware representations during inference but incur prohibitive computation costs, while efficient alternatives remove future modeling at inference time and may lose the robustness benefits of… ▽ More World Action Models (WAMs) improve robot manipulation by learning how the environment evolves beyond the current observation. However, existing approaches face a fundamental dilemma: Joint-WAMs preserve future-aware representations during inference but incur prohibitive computation costs, while efficient alternatives remove future modeling at inference time and may lose the robustness benefits of temporal reasoning. In this work, we revisit the role of future representations in WAMs and show that inference-time future conditioning is critical for generalization under distribution shifts. This observation motivates Faster-WAM, an efficient future-conditioning WAM that preserves future representations while avoiding expensive video-action interaction. Faster-WAM introduces a sparse future-conditioning framework that computes future representations once and selectively reuses them throughout action denoising. Specifically, we propose SparseMoT to replace ubiquitous layer-wise fusion with selective video-action interaction at a compact subset of network stages, and Interval KV-Fusion to aggregate multi-depth future representations without increasing attention complexity. Experiments demonstrate that Faster-WAM achieves a substantially better performance-efficiency trade-off than existing WAMs. On the out-of-distribution LIBERO-Plus benchmark, Faster-WAM improves success rate from 49.14% to 73.57% compared with Fast-WAM, while running 2.21$\times$ faster than Joint-WAM. It further achieves state-of-the-art performance on LIBERO and RoboTwin 2.0, while demonstrating strong robustness in real-world manipulation. △ Less

Submitted 4 August, 2026; originally announced August 2026.

MissClick: Exploiting Digit-Serialized Coordinates to Attack GUI Grounding Models

Authors: Yu Ran, Wentao Zhao, Xin Zhang, Yi Pan

Abstract: Recent GUI visual grounding models generate screen coordinates as sequences of digit tokens that are parsed into numerical values and mapped to executable clicks. The security implications of this coordinate generation process have been largely overlooked. We observe that each coordinate digit is predicted as a categorical token, yet after parsing, changing a hundreds-place digit by one changes th… ▽ More Recent GUI visual grounding models generate screen coordinates as sequences of digit tokens that are parsed into numerical values and mapped to executable clicks. The security implications of this coordinate generation process have been largely overlooked. We observe that each coordinate digit is predicted as a categorical token, yet after parsing, changing a hundreds-place digit by one changes the corresponding numerical coordinate component by 100 units, which can induce a large displacement of the executed click. This observation motivates attack objectives that account for the numerical and place-value structure of coordinate outputs rather than treating them as ordinary text. Moreover, untargeted and targeted attacks impose different success conditions--displacing the click outside the correct region versus into an attacker-specified region--and therefore benefit from different objectives. We propose MissClick, a simple and effective white-box adversarial attack with two goal-specific objectives: MissClick-U maximizes soft-coordinate displacement for untargeted disruption, while MissClick-T minimizes a place-weighted target-digit loss for targeted hijacking. Compared with existing attacks against GUI grounding models on OS-Atlas and UGround across desktop, web, and mobile platforms, MissClick-U achieves untargeted success rates of 75.07\% and 72.93\% (+16.62 and +30.72 pp), and MissClick-T achieves targeted success rates of 44.86\% and 62.67\% (+31.73 and +47.06 pp). Attack objective comparison further shows that soft-coordinate displacement yields the highest untargeted attack success rate, whereas place-weighted target-digit optimization yields the highest targeted attack success rate, revealing distinct objective preferences for the two attack goals. △ Less

Submitted 4 August, 2026; originally announced August 2026.

LLaDA MoE v2: Scaling Mixture-of-Experts Diffusion Language Models

Authors: Fengqi Zhu, Shaoxuan Xu, Jingyang Ou, Zebin You, Yipeng Xing, Huabin Liu, Xiaolu Zhang, Jun Zhou, Zhenzhong Lan, Yankai Lin, Wayne Xin Zhao, Jianguo Li, Chongxuan Li, Ji-Rong Wen

Abstract: Diffusion language models (dLLMs) offer an alternative to autoregressive (AR) language modeling, yet the scaling behavior of Mixture-of-Experts (MoE) dLLMs remains poorly understood. We systematically characterize how optimization hyperparameters, compute allocation, and architecture scale for MoE dLLMs, identifying quantitative differences from scaling trends previously reported for AR models. Sp… ▽ More Diffusion language models (dLLMs) offer an alternative to autoregressive (AR) language modeling, yet the scaling behavior of Mixture-of-Experts (MoE) dLLMs remains poorly understood. We systematically characterize how optimization hyperparameters, compute allocation, and architecture scale for MoE dLLMs, identifying quantitative differences from scaling trends previously reported for AR models. Specifically, for optimization, the optimal nominal batch size grows faster, while the optimal learning rate decays more rapidly with compute. For model--data allocation, IsoFLOP analysis reveals a slight data-side tilt: the optimal token budget grows faster than activated model-side computation. For MoE architecture, larger scales increasingly favor larger expert pools at fixed activated capacity, while moderate expert granularity remains consistently effective and the preferred fraction of activated capacity assigned to shared experts remains stable across scales. Guided by these findings, we train LLaDA MoE v2, a 30B-A3B dLLM, from scratch on 23.5T tokens. With approximately 65\% as many pretraining tokens as Qwen3, LLaDA MoE v2 approaches Qwen3 on several knowledge, reasoning, and coding benchmarks. After supervised fine-tuning alone, it outperforms SDAR Chat on seven of eight reasoning and coding benchmarks and remains close to Qwen3 on several tasks. These results establish practical scaling laws and design principles for MoE dLLMs. △ Less

Submitted 4 August, 2026; originally announced August 2026.

Dynamic Resolution Routing for Efficient Egocentric Grounding

Authors: Huixin Sun, Wangbo Zhao, Fanyue Wei, Qiuxia Lin, Pengzhan Sun, Angela Yao

Abstract: Egocentric visual grounding requires high-resolution inputs to localize small objects. However, scaling Multimodal Large Language Models to this domain is constrained by the excessive cost of visual token processing. We identify that current efficient strategies based on token reduction are unreliable for selecting object-centric spatial evidence. To overcome this, we propose SmartRes, a framework… ▽ More Egocentric visual grounding requires high-resolution inputs to localize small objects. However, scaling Multimodal Large Language Models to this domain is constrained by the excessive cost of visual token processing. We identify that current efficient strategies based on token reduction are unreliable for selecting object-centric spatial evidence. To overcome this, we propose SmartRes, a framework that performs efficiency optimization in the pixel space via dynamic resolution routing. SmartRes first encodes a low-resolution view for global context and uses a lightweight router to activate high-resolution patches in object-centric regions and constructs an order-preserving visual sequence. To further enable robust routing under severe foreground-background imbalance, we introduce a margin-regularized routing objective that increases foreground-background logit separation and improves foreground recall. Experiments on Ego4D and EgoIntention show that SmartRes reduces visual tokens by up to 67% while retaining 86.4% of full-resolution performance, and achieves up to 1.66X faster inference than state-of-the-art token reduction methods with higher accuracy. Furthermore, strong performance on small object grounding indicates the effectiveness of SmartRes towards egocentric applications. Code will be publicly available. △ Less

Submitted 2 August, 2026; originally announced August 2026.

Collaborative Memory Augmentation for Generative Recommendation

Authors: Enze Liu, Zhen Tian, Wayne Xin Zhao

Abstract: Generative Recommendation (GR) has exhibited great potential by modeling item transitions as a sequence-to-sequence task. Despite the success of GR, existing frameworks primarily focus on modeling individual user sequences within a constrained internal parametric space, failing to explicitly leverage cross-user collaborative signals. To address this issue, we propose \textbf{OMEGA}, a cOllaborativ… ▽ More Generative Recommendation (GR) has exhibited great potential by modeling item transitions as a sequence-to-sequence task. Despite the success of GR, existing frameworks primarily focus on modeling individual user sequences within a constrained internal parametric space, failing to explicitly leverage cross-user collaborative signals. To address this issue, we propose \textbf{OMEGA}, a cOllaborative MEmory augmentation framework for Generative recommendAtion. OMEGA bridges the gap between implicit parametric knowledge and explicit collaborative signals. We first introduce a latent context compression method that utilizes learnable query tokens to distill sequential user behavior into compact representations, significantly reducing storage overhead. These compressed representations are aggregated into a collaborative memory bank, serving as an explicit repository of global behavioral patterns. To ensure precise knowledge acquisition, we design a lightweight and target-aware retrieval mechanism that identifies pertinent memories by considering both sequence-level and target-level similarities. Furthermore, a context-aware integration module, equipped with a gated cross-attention mechanism, is employed to adaptively fuse the retrieved collaborative memories with the local user context while mitigating the interference of noisy patterns. Empirical evaluations on multiple real-world datasets demonstrate that OMEGA significantly outperforms existing advanced GR models, validating the potential of external memory as a complement to the generative paradigm. △ Less

Submitted 2 August, 2026; originally announced August 2026.

Comments: Accepted by KDD 2026 Research Track

Smart membrane: high content in situ monitoring barrier on chip with artificial neural network

Authors: Bo Tang, Victor Krajka, Mengxi Liu, Wei Zhao, Gazal Goekkus, Paul Lukowicz, Lili Zhu, Pu Chen, Stephan Reichl, Andreas Dietzel

Abstract: Conventional transepithelial electrical resistance (TEER) technique provides only a low-content analysis of cell-layer conditions, necessitating repeated microscopic assessments of morphology and cell-cell contacts outside the incubator for barrier-on-chip systems. This work presents a novel high-content TEER device in the form of a novel nanoporous membrane that facilitates continuous electrical… ▽ More Conventional transepithelial electrical resistance (TEER) technique provides only a low-content analysis of cell-layer conditions, necessitating repeated microscopic assessments of morphology and cell-cell contacts outside the incubator for barrier-on-chip systems. This work presents a novel high-content TEER device in the form of a novel nanoporous membrane that facilitates continuous electrical measurement of cell-substrate impedance sensing (ECIS). The ultrathin (700 nm) membrane, composed of ultra-low-stress SixNy, is monolithically integrated into wafer-level fabricated chips sealed with glass lids. Coplanar ECIS electrodes were connected to custom electronics to record impedance under sinusoidal excitation. Human umbilical vein endothelial cells (HUVECs) were seeded and continuously recorded impedance spectra were compared with bright-field and fluorescence microscopy, revealing distinct phases of monolayer formation. With one-dimensional convolutional neural network (Conv1d) and Kolmogorov-Arnold Network (KAN) trained with a small amount of Nyquist-diagrams, phases of (I) adherence, (II) outspreading, (III) confluence and (IV) barrier maturity with tight junction formation could be recognized with 95% confidence. As further proof of concept, reversible and irreversible barrier weakening using modulators PN159 and BAC was identified in this way. Our studies have demonstrated that an immediate and automatable non-invasive detection of in-vitro barrier dynamics within barrier-on-chip systems, eliminating the need for microscopy and endpoint staining. We expect this ECIS technique will find broad applications in organ-on-chip systems for in situ monitoring physiological or pathological states of tissue barrier. △ Less

Submitted 4 August, 2026; v1 submitted 2 August, 2026; originally announced August 2026.

Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning

Authors: Xujun Che, Yuchen Yuan, Weida Zhao, Chenyang Yu

Abstract: Error-penalized scoring rules ($+1$ for a correct answer, $-λ$ for a wrong one, $0$ for abstaining) are increasingly prescribed against hallucination: a rational agent facing such a rule answers exactly when its correctness probability exceeds Chow's threshold $t^\ast=λ/(1+λ)$. We prove that a KL-anchored gradient learner can do the opposite. When abstention is a discrete action, the reward gradie… ▽ More Error-penalized scoring rules ($+1$ for a correct answer, $-λ$ for a wrong one, $0$ for abstaining) are increasingly prescribed against hallucination: a rational agent facing such a rule answers exactly when its correctness probability exceeds Chow's threshold $t^\ast=λ/(1+λ)$. We prove that a KL-anchored gradient learner can do the opposite. When abstention is a discrete action, the reward gradient and the anchor's restoring force are throttled by the same gate-saturation factor and die together: under explicit conditions (among them, blanket answering loses score in expectation and prompts share a bounded readout) the model drifts toward refusing everything, its mean training reward rising to zero like $1/t$ in training time $t$, so the curve reads as improvement while coverage collapses. The advantage estimator compounds the failure: in its sparse-answer regime, group normalization silently replaces every designed penalty with an effective penalty of one, moving the learned threshold from $t^\ast$ to $1/2$. The repair is structural: train a mandatory confidence report with a strictly proper score plus a correctness reward, and abstain only at deployment by thresholding the report. The always-emitted report has no gate to saturate, so no shared factor can kill its reward gradient and its anchor together, and its calibrated optimum is attracting. Simulations confirm every prediction, and experiments on language models at two scales confirm the mechanism live: the rule silences questions the models demonstrably still solve within ten optimizer steps, an ablation isolates the cause, and report-level training raises coverage, accuracy, and calibration together. △ Less

Submitted 31 July, 2026; originally announced August 2026.

Is It Time for the Renaissance of Salient Object Detection in the Era of MLLMs?

Authors: Wenzhuo Zhao, Xiuzhi Li, Zhongkuan Mao, Ronghao Xian, Yao Jiang, Zhao Gao, Keren Fu, Qijun Zhao, Jian Cheng

Abstract: The zero-shot capabilities of multimodal large language models (MLLMs) are pushing salient object detection (SOD) beyond task-specific supervision. To disentangle MLLMs beyond conventional mask-based evaluation, we decompose SOD into localization and segmentation, and re-engineer datasets with phrases, boxes, and attributes, establishing a diagnostic benchmark for MLLM saliency perception (SaliLLM… ▽ More The zero-shot capabilities of multimodal large language models (MLLMs) are pushing salient object detection (SOD) beyond task-specific supervision. To disentangle MLLMs beyond conventional mask-based evaluation, we decompose SOD into localization and segmentation, and re-engineer datasets with phrases, boxes, and attributes, establishing a diagnostic benchmark for MLLM saliency perception (SaliLLM). SaliLLM uncovers a striking capability mismatch: MLLMs outperform state-of-the-art (SOTA) methods in localization, yet remain substantially weaker in segmentation. Further analyses attribute this gap primarily to mismatches between MLLMs and annotations over foreground cardinality, granularity, and extent. Motivated by this diagnosis, we recast zero-shot SOD as protocol-aligned Foreground Organization and introduce the first training-free framework that leverages Gestalt-inspired Collaborative attention for Unified SOD (FOCUS). FOCUS couples top-down Bayesian-surprise calibration of protocol-conditioned foreground granularity with bottom-up propagation of MLLMs evidence over entity-centric perceptual manifolds induced by self-supervised features, yielding coherent object extents as prompts for a general segmenter. Across 13 RGB, RGB-D, and RGB-T SOD benchmarks, FOCUS generally surpasses SOTA methods without training, reducing mean absolute error by 11\%, 34\%, and 48\% compared with fully, weakly, and self-supervised methods, respectively. Our findings signal the renaissance of SOD: from task-specific supervision to zero-shot foreground organization. Code is available in the supplementary material. △ Less

Submitted 31 July, 2026; originally announced July 2026.

Comments: 10 pages, 4 figures, conference

Self-Supervised Skill Optimization

Authors: Siran Peng, Cuiyu Yang, Tianyu Fu, Tianshuo Zhang, Haoyuan Zhang, Weisong Zhao, Anyang Su, Minghui Wu, Huiying Li, Xiangyu Zhu, Chenxu Zhao, Zhen Lei

Abstract: Agent skills provide frozen large language model (LLM) agents with reusable procedural guidance, and recent work shows that such skills can be optimized with ground-truth (GT) feedback. Many applications, however, lack GT labels, task scores, rewards, or reliable task-specific evaluators. We therefore introduce Self-Supervised Skill Optimization (SSO), a comparative framework that learns a reusabl… ▽ More Agent skills provide frozen large language model (LLM) agents with reusable procedural guidance, and recent work shows that such skills can be optimized with ground-truth (GT) feedback. Many applications, however, lack GT labels, task scores, rewards, or reliable task-specific evaluators. We therefore introduce Self-Supervised Skill Optimization (SSO), a comparative framework that learns a reusable skill from unlabeled task instances alone. At each step, SSO runs the current skill on an unlabeled batch, uses a subset of the resulting executions to generate complete skill probes, and runs the probes on the same batch. An LLM judge compares the resulting answers, trajectories, artifacts, or terminal states. A separate behavior extractor identifies behavioral differences without seeing the judge's decisions. SSO uses these decisions to aggregate evidence for and against the observed behaviors across instances. It then ranks the behaviors by the resulting evidence and renders a new complete skill from the highest-ranked behaviors. The update is accepted only if the new skill outperforms the current one on an unlabeled validation set. SSO outperforms existing GT-free prompt optimizers on both closed-ended and open-ended tasks. On closed-ended benchmarks, it approaches and sometimes exceeds the strongest GT-based skill optimizer without using any GT feedback. △ Less

Submitted 30 July, 2026; originally announced July 2026.

ViP-Rig: Visual-Prompted Controllable Rigging

Authors: Zihan Qin, Mingze Sun, Yifan Mao, Jialei Xu, Jingfeng Guo, Changrong Hu, Wenbo Zhao, Junjun Jiang, Xianming Liu

Abstract: Rigging is inherently task-dependent because the same mesh may require different skeletons and deformation behaviors across animation tasks. In practice, artists often inspect an initial rig and repeatedly edit its skeletal structure and deformation behavior to meet specific animation requirements. Existing automatic methods primarily generate a plausible rig from geometry, offering limited explic… ▽ More Rigging is inherently task-dependent because the same mesh may require different skeletons and deformation behaviors across animation tasks. In practice, artists often inspect an initial rig and repeatedly edit its skeletal structure and deformation behavior to meet specific animation requirements. Existing automatic methods primarily generate a plausible rig from geometry, offering limited explicit control over the resulting skeleton and deformation behavior. In this work, we present ViP-Rig, a visual-prompted framework that supports both prompt-first rigging and result-guided editing by injecting features extracted from user-drawn or edited 2D skeletal and rigidity prompts into frozen pretrained backbones. Specifically, ViP-Rig consists of two stages, Skeleton Generation and Skinning Prediction. In the first stage, the skeletal sketch is processed by the Dense-to-Compact Visual Prompt Encoding to produce compact, fixed-length conditioning tokens. The resulting tokens are injected into a frozen pretrained autoregressive generator through gated adapters to control joint placement and branching structure while preserving the generator's geometric prior. In the second stage, the rigidity map is processed using the same visual encoding design, while the pretrained skinning backbone remains frozen. The resulting tokens are symmetrically injected into the point and joint streams to modulate point-joint compatibility and the resulting skinning weights. Experiments on Articulation-XL2.0 and zero-shot evaluation on ModelsResource show that ViP-Rig more accurately recovers target skeletons and skinning weights than geometry-conditioned baselines under prompt-guided evaluation. Qualitative results further demonstrate explicit and localized control in both prompt-first rigging and result-guided editing. △ Less

Submitted 30 July, 2026; originally announced July 2026.

Comments: 8 pages, 4 figures. Zihan Qin and Mingze Sun contributed equally. Xianming Liu is the corresponding author

AutoSupervision: Closing the Feedback Loop in Scientific Workflows with Grounded Revision Verification

Authors: Haobo Li, Eunseo Jung, Wenxiao Zhao, Feng Liu, Jiong Wang, Kaiyi Xu, Zijie Guo, Zixin Chen, Ben Fei, Fenghua Ling, Lei Bai

Abstract: Recent advances in large language models (LLMs) have enabled AI systems to assist scientific research and peer review. However, an essential capability for reliable AI-assisted scientific workflows remains underexplored: verifying whether reviewer feedback leads to meaningful and evidence-supported manuscript improvements. We introduce AutoSupervision, which evaluates whether scientific manuscript… ▽ More Recent advances in large language models (LLMs) have enabled AI systems to assist scientific research and peer review. However, an essential capability for reliable AI-assisted scientific workflows remains underexplored: verifying whether reviewer feedback leads to meaningful and evidence-supported manuscript improvements. We introduce AutoSupervision, which evaluates whether scientific manuscript revisions genuinely address reviewer concerns through grounded evidence. AutoSupervision leverages transparent peer-review records as a natural source of supervision, where reviewer comments specify scientific concerns, author responses describe claimed resolutions, and revised manuscripts provide evidence of changes. Given reviewer comments, author responses, and revised manuscripts, models must characterize reviewer concerns, determine whether concerns have been addressed, and identify supporting manuscript evidence. We construct AutoSupervision from 56,000 Nature Communications articles and corresponding review records. Then we conducted experiments on LLMs, the ablation study, and the case study. Our results show that while LLMs perform well in characterizing reviewer concerns, with GPT-5.5 achieving a score of 0.754, evidence-based verification remains the primary bottleneck, with the best-performing model reaching only 0.501. △ Less

Submitted 30 July, 2026; originally announced July 2026.

Thinking Once Is Enough: Intermediate-Layer Evidence Routing for High-Resolution VQA

Authors: Zhongkuan Mao, Xianjie Liu, Tianyu Meng, Yidong Wang, Wenzhuo Zhao, Ronghao Xian, Yao Jiang, Fei Shen, Junfeng Fang, Yong Dai, Yi Zhang, Keren Fu

Abstract: High-resolution visual question answering (HR-VQA) is often treated as a problem of insufficient evidence acquisition, where failing multimodal large language models must inspect images again through cropping, re-encoding, or multi-round search. We show that this view is incomplete: in many cases, fine-grained evidence has already survived visual encoding and become identifiable and influential wi… ▽ More High-resolution visual question answering (HR-VQA) is often treated as a problem of insufficient evidence acquisition, where failing multimodal large language models must inspect images again through cropping, re-encoding, or multi-round search. We show that this view is incomplete: in many cases, fine-grained evidence has already survived visual encoding and become identifiable and influential within an intermediate-layer routing window, but is later diluted before answer generation. We propose Thinking-Once, a \textbf{training-free, single-visual-pass} evidence-routing method that reconstructs question-conditioned attention at this window, preserves core entity tokens and compact background context, and routes this evidence to later layers without extra visual encoding. Across five base models, Thinking-Once consistently improves or matches the corresponding base setting, increasing the average scores on V$^*$Bench, HRBench-4K, and HRBench-8K by \textit{+3.1}, \textit{+3.0}, and \textit{+2.7} points while reducing the average peak memory by about 4,GB. On Qwen2.5-VL-7B, it improves the three benchmarks by \textit{+9.9}, \textit{+4.6}, and \textit{+5.5} points, raising the cross-benchmark mean from 72.5 to 79.1. With the ZwZ-8B base model, Thinking-Once reaches a mean score of 82.7. Against 11 open-source HR-VQA baselines, it obtains the best or tied-best score on all three benchmark averages and the best overall mean; for example, compared with DeepScan, it reduces V$^*$Bench inference time by \textbf{97.2\%} while improving the cross-benchmark mean from 77.8 to 79.1. These results show that HR-VQA can be improved by routing already encoded evidence rather than repeatedly acquiring new visual inputs. Code is available in the appendix. △ Less

Submitted 30 July, 2026; originally announced July 2026.

Private Face Recognition Training Dataset Publication via Identity-Decoupled and Geometry-Preserving Face Distillation

Authors: Shuhuan Chen, Xiangyu Zhu, Weisong Zhao, Siran Peng, Tianshuo Zhang, Haoyuan Zhang, Haichao Shi, Xiao-Yu Zhang, Zhen Lei

Abstract: Publishing private face recognition~(FR) training datasets is privacy-sensitive because faces expose identity information. Private FR training dataset publication mitigates this risk by releasing protected proxies as substitutes for private training faces. However, training FR models with such data introduces an identity paradox: \emph{the identity cues that make released faces useful for recognit… ▽ More Publishing private face recognition~(FR) training datasets is privacy-sensitive because faces expose identity information. Private FR training dataset publication mitigates this risk by releasing protected proxies as substitutes for private training faces. However, training FR models with such data introduces an identity paradox: \emph{the identity cues that make released faces useful for recognition supervision are also the cues that make them linkable to real individuals.} A protected face should be decoupled from the original identity, yet still behave as a reliable identity sample for training. Removing these cues too aggressively may destroy the class structure needed for recognition learning, whereas preserving them too faithfully may increase source-identity linkability. We argue that this paradox stems from conflating source-aligned identity semantics with recognition-useful proxy identity geometry. The former should be suppressed to reduce linkage to private individuals, while the latter should be preserved for FR learning. Based on this insight, we propose \textbf{Private Face Distillation}, an identity-decoupling and geometry-preserving framework. It uses Orthogonal Geometry Preservation to construct decoupled proxy identities from private identity representations while maintaining hyperspherical geometry, and Relational Topology Alignment to preserve identity relations for recognition learning. Experiments across multiple domain-shifted FR scenarios show that Private Face Distillation achieves stronger utility than the evaluated publication baselines. On IJB-C surveillance, it improves $\mathrm{TAR}@\mathrm{FAR}{=}1\text{e-}{3}$ by 3.94\% over the baseline while reducing source-identity linkability. These results suggest that private FR training dataset publication should decouple source-identity correspondence while preserving proxy identity geometry. △ Less

Submitted 30 July, 2026; originally announced July 2026.

Unbiased Data-Driven Determination of the Nuclear Dipole Amplitude in the Color Glass Condensate

Authors: Si-Wei Dai, Haowu Duan, Long-Gang Pang, Guang-You Qin, Shu-Yi Wei, Han-Zhong Zhang, Wenbin Zhao

Abstract: Gluon saturation limits the growth of parton densities at small Bjorken-$x$ and is expected to be most pronounced in heavy nuclei. Yet quantitative extractions of the nuclear gluon dipole amplitude have long relied on parametrized initial conditions, introducing uncontrolled model dependence that obscures genuine nuclear effects. We introduce a physics-informed neural-network framework that embeds… ▽ More Gluon saturation limits the growth of parton densities at small Bjorken-$x$ and is expected to be most pronounced in heavy nuclei. Yet quantitative extractions of the nuclear gluon dipole amplitude have long relied on parametrized initial conditions, introducing uncontrolled model dependence that obscures genuine nuclear effects. We introduce a physics-informed neural-network framework that embeds the collinearly improved Balitsky-Kovchegov evolution equation directly into the training objective, allowing the impact-parameter-averaged dipole amplitude to be determined from data without assuming a functional form for its initial condition. Applying this framework to forward-hadron nuclear-modification-factor and coherent $J/ψ$ photoproduction data, we extract the $^{208}$Pb dipole amplitude at $x_0=0.01$ with QCD evolution and momentum-space positivity enforced throughout training. The evolved amplitude reproduces the measured cross sections across the available kinematic range and yields a saturation-scale ratio $Q_{s0,\mathrm{Pb}}^2/Q_{s0,p}^2 = 3.17^{+0.17}_{-0.10}$, consistent with simple geometric scaling. The extracted Pb initial condition is well described by a McLerran-Venugopalan-type form, in contrast to the proton, reflecting the higher color-charge density of a large nucleus. Using the same amplitude, we predict the rapidity dependence of the transverse-momentum ratio in $pp$, $p$Pb, and Pb$p$ collisions, finding agreement with recent LHCb measurements at low multiplicity without any system-dependent parameters. This work provides the first unbiased, data-driven determination of nuclear structure in the saturation regime and establishes a general strategy for embedding nonlinear evolution equations into machine-learning extractions of dynamically constrained observables. △ Less

Submitted 29 July, 2026; originally announced July 2026.

Collaborative Weighting with Pessimistic Critic for Mitigating Overestimation in Off-Policy Reinforcement Learning

Authors: Gong Gao, Xiao Lai, Ziqi Xie, Guojie Chen, Xianhui Liu, Weidong Zhao

Abstract: Deep off-policy reinforcement learning algorithms for continuous control typically rely on neural value function approximation to guide policy improvement. However, temporal-difference (TD) learning introduces noisy targets, resulting in non-stationary optimization, while greedy policy updates amplify early-stage estimation errors. The recursive propagation of such errors leads to persistent overe… ▽ More Deep off-policy reinforcement learning algorithms for continuous control typically rely on neural value function approximation to guide policy improvement. However, temporal-difference (TD) learning introduces noisy targets, resulting in non-stationary optimization, while greedy policy updates amplify early-stage estimation errors. The recursive propagation of such errors leads to persistent overestimation bias and degraded training stability in actor-critic methods. Existing approaches attempt to alleviate this issue via prioritized sampling or modified value learning objectives, but often overemphasize high-uncertainty transitions caused by limited data coverage or bootstrapping errors, thereby further amplifying bias.In this paper, we propose Collaborative Weighting Actor-Critic (CWAC), a unified framework that explicitly accounts for predictive uncertainty in value estimation. CWAC employs distributional critic to model return uncertainty and introduces a collaborative weighting mechanism that jointly reweights TD-errors and uncertainty, enabling robust learning from reliable samples while suppressing noisy updates. In addition, we incorporate a stochastic pessimistic value estimation scheme via sampling from the return distribution, which effectively mitigates error propagation during policy improvement. CWAC can be seamlessly integrated into existing off-policy algorithm frameworks such as SAC, TD3, and DDPG with minimal overhead. Empirical results demonstrate that our proposed method significantly enhances performance across a diverse range of simulated tasks. Our code is publicly available at https://anonymous.4open.science/r/CWAC-348E. △ Less

Submitted 29 July, 2026; originally announced July 2026.

Exploring Structures in Physics Problems: Can AI Agents Discover Statistical Mechanical Mappings?

Authors: Wanyu Zhao, Wanbing Zhao

Abstract: An important skill in theoretical physics is to recognize when a new problem can be transformed into a known model. We study this skill as an AI-agent task: can LLM-based agents discover statistical mechanical mappings from a raw partition function to a tractable representation? To probe this question, we introduce StatMechBench-v0, a benchmark of six Ising-type problems covering transfer-matrix m… ▽ More An important skill in theoretical physics is to recognize when a new problem can be transformed into a known model. We study this skill as an AI-agent task: can LLM-based agents discover statistical mechanical mappings from a raw partition function to a tractable representation? To probe this question, we introduce StatMechBench-v0, a benchmark of six Ising-type problems covering transfer-matrix methods, gauge-removable disorder, and planar/Pfaffian structure. We evaluate a simple propose-verify-revise agent across multiple LLMs and problem phrasings. The results show that numerical feedback often helps agents repair code and recover correct partition functions. However, agents can also pass the numerical checks while misidentifying the underlying tractable class or understating computational complexity. This both reveals limitations in current LLM reasoning and calls for a verification stack that goes beyond numerical agreement, incorporating, for example, symbolic checks and structural invariants. Our study provides an early evaluation and design directions for AI agents aimed at structural discovery in theoretical physics. △ Less

Submitted 28 July, 2026; originally announced July 2026.

Comments: Accepted to the AID-Wild workshop at CAIS 2026

Nanoscale Storage of Incompatible Elements at Olivine Grain Boundaries in Natural Basalts

Authors: Wenhao Zhao, Reid Cooper, Stephen Parman, Austin Akey, Greg Hirth

Abstract: Grain boundaries are pervasive in polycrystalline olivine, yet their structure and trace-element chemistry remain poorly constrained. We characterize boundaries in undeformed olivine aggregates from Piton de la Fournaise (La Reunion) and Mauna Loa (Hawaii) by correlating electron backscatter diffraction, transmission electron microscopy, and atom probe tomography. The boundaries are crystalline, w… ▽ More Grain boundaries are pervasive in polycrystalline olivine, yet their structure and trace-element chemistry remain poorly constrained. We characterize boundaries in undeformed olivine aggregates from Piton de la Fournaise (La Reunion) and Mauna Loa (Hawaii) by correlating electron backscatter diffraction, transmission electron microscopy, and atom probe tomography. The boundaries are crystalline, with a structural width of about 1 nm and no continuous glassy film resolved. Ca, Al, P, Na, and Ti are selectively enriched, whereas Mg, Fe, Ni, and Mn remain homogeneous. Several enriched mass windows coincide with nominal rare-earth-element ion positions, but unresolved isobars and the lack of a diagnostic isotope envelope preclude secure assignments. Divalent-cation segregation follows an ionic-size-misfit trend, while aliovalent departures require unconstrained contributions. The Ca interfacial excess is 2.94-2.97 atoms per square nanometer, equivalent to 0.327-0.330 monolayers, and lies near the upper published 1523 K equilibrium segregation isotherm of Hiraga et al. (2004), within the range of basaltic olivine crystallization temperatures. This supports boundary formation during crystal growth, although synneusis followed by high-temperature residence could produce a similar signature. Ca chemical widths are about 6-9 nm, and those of the most strongly enriched elements reach about 10-15 nm, demonstrating decoupling between the narrow structural core and broader chemical segregation. Compared with published atom-probe data on deformed olivine, these boundaries preserve broader halos, although analytical differences make the contrast tentative. Olivine grain boundaries therefore constitute a distinct nanoscale reservoir that should be included alongside crystal interiors and residual melt in trace-element mass balances of olivine-rich aggregates. △ Less

Submitted 28 July, 2026; originally announced July 2026.

Comments: 46 pages; 10 main figures and 10 supplementary figures; supplementary text and tables included

TopoGR: Revealing and Preserving Latent Structure of Semantic ID in Generative Recommendation

Authors: Ziyu Zheng, Zhengshun Du, Yaming Yang, Bin Tong, Guan Wang, Meng Yan, Ziyu Guan, Wei Zhao

Abstract: Semantic ID-based generative recommendation tokenizes each item into a sequence of discrete semantic IDs and predicts the next item by generating semantic IDs. However, existing methods typically regard SIDs as independent discrete symbols, while often overlooking the topology of the learned semantic ID space. We identify a structural mismatch between tokenization and generation: the tokenizer lea… ▽ More Semantic ID-based generative recommendation tokenizes each item into a sequence of discrete semantic IDs and predicts the next item by generating semantic IDs. However, existing methods typically regard SIDs as independent discrete symbols, while often overlooking the topology of the learned semantic ID space. We identify a structural mismatch between tokenization and generation: the tokenizer learns a structured code space with semantic neighborhood relations, whereas the generator consumes semantic ID tokens as independent categorical symbols. Consequently, item relatedness is reduced to exact semantic ID overlap, making it difficult to identify semantically similar items whose semantic IDs do not overlap. To address this issue, we propose TopoGR, a topology-preserving generative recommendation framework based on Bit-decomposable Semantic ID(Binary SID). Each Binary SID is learned in a bit-decomposable form and can be deterministically converted to a standard integer SID, while exposing an explicit Hamming geometry. TopoGR exploits this topology at three stages: binary SID features preserve Hamming proximity at the input layer; Hamming soft targets inject topology-aware supervision; and Hamming-consistent reranking aligns candidate items with the predicted binary prototype during inference. We further verify that the Hamming topology can capture item relatedness beyond exact SID matching. Experiments on four benchmark datasets show that TopoGR consistently outperforms existing state-of-the-art baselines in recommendation performance. △ Less

Submitted 27 July, 2026; originally announced July 2026.

Comments: under review

GAUGE: Grading Agent-Built Financial Models Without a Golden Answer

Authors: Jiacheng Lu, Sinuo Wang, Wentao Zhao, Rui Sun, Cheng Hua, Tao Song, Hui Cai, Beidi Luan, Zhengze Wu, Lingjing Teng, Yijia He, Jing Li, Daxin Jiang, Zuo Bai, Haibing Guan

Abstract: Financial models combine public disclosures with analyst assumptions to produce forecasts and valuations. While some components can be checked mechanically, forecasts, discount rates, and target prices often admit multiple reasonable answers. Existing benchmarks nevertheless tend to grade such outputs against a single expert reference. Using independently built analyst models for the same companie… ▽ More Financial models combine public disclosures with analyst assumptions to produce forecasts and valuations. While some components can be checked mechanically, forecasts, discount rates, and target prices often admit multiple reasonable answers. Existing benchmarks nevertheless tend to grade such outputs against a single expert reference. Using independently built analyst models for the same companies, we find that across 108 directed pairs covering 65 companies, the median single-reference score is 0.33, 92.6% score below 0.70, and no same-vintage pair agrees on implied price within 10%. Point-tolerance grading can therefore penalize disagreement already present among professionals. We introduce GAUGE, a benchmark for evaluating agent-built valuation models against observed analyst practice rather than a single point answer. GAUGE uses 1,001 vendor-classified analyst workbooks and a 196-task evaluation set, with a three-layer observed-practice envelope, 56 auditable facets, eight validity gates, and deterministic structural checks. We validate the benchmark with a 55-participant known-groups study, company-grouped cross-fitting, and judge-stability audits. On the failure-aware score $φ_0$, senior analysts average 88.3, juniors 66.0, and finance students 43.2. Across 24 agents and 1,011 scored generations, the best agent scores 53.4, above the student mean but below every senior and most juniors. It passes 93% of mechanical facets and 78% of judgment facets, with a fleet-median gap of 26 points. Current agents are substantially stronger at model construction than valuation judgment. We release the methodology, a gated de-identified data tier, a controlled training split, a versioned 48-task evaluation core, and a withheld refresh pool. △ Less

Submitted 27 July, 2026; originally announced July 2026.

Comments: 35 pages, including appendices. Code, benchmark materials, and evaluation artifacts will be publicly released

Kimi K3: Open Frontier Intelligence

Authors: Kimi Team, Tongtong Bai, Yifan Bai, Yiping Bao, M. C., Jianfeng Cai, Xinyuan Cai, Peizhou Cao, Yuxuan Cao, Ziwei Chai, Y. Charles, H. S. Che, Guanduo Chen, Guangyu Chen, Guanzheng Chen, Huarong Chen, Jia Chen, Jianlong Chen, Jun Chen, Kexin Chen, Peng Chen, Ruijue Chen, Wentao Chen, Xin Chen, Yang Chen , et al. (377 additional authors not shown)

Abstract: We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model depth. Together with Stable LatentMoE, which effectively activates 16 of 896 routed experts per token… ▽ More We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model depth. Together with Stable LatentMoE, which effectively activates 16 of 896 routed experts per token, and refined training and data recipes, these advances yield an approximately 2.5x improvement in overall scaling efficiency over Kimi K2. Post-training highlights reinforcement learning across general, agentic, and coding domains and multiple reasoning-effort levels, enabling compositional generalization and robust long-horizon execution. At 2.8T scale, Kimi K3 is supported by infrastructure advances in multiple areas: algorithm-system co-design for KDA, perfectly balanced expert-parallel training with efficient memory management, million-token agentic RL with persistent rollout and sandbox states, and deployment innovations. Extensive evaluations show that Kimi K3 achieves frontier-level performance across long-horizon coding, agentic, knowledge, reasoning, and vision tasks. While its overall performance still trails the most powerful proprietary models, namely Claude Fable 5 and GPT-5.6 Sol, Kimi K3 consistently outperforms other open and proprietary models evaluated in our suite. We release the full Kimi K3 model weights to facilitate future research and accelerate the broader deployment and adoption of frontier intelligence. △ Less

Submitted 7 August, 2026; v1 submitted 27 July, 2026; originally announced July 2026.

Comments: K3 tech report

Gravitational Lensing of Gravitational Waves: Towards a Higher-order Geometric-optics Approach

Authors: Zhao Li, Shaoqi Hou, Wen Zhao

Abstract: In this work, we study the gravitational lensing of gravitational waves (GWs) by extending the geometric-optics approximation to higher order. With the help of the Newman-Penrose formalism, we reexpress the GW propagation equations as a series of scalar equations and present explicit expressions for the Weyl scalars that describe the GW polarizations. By combining the approaches of solving geodesi… ▽ More In this work, we study the gravitational lensing of gravitational waves (GWs) by extending the geometric-optics approximation to higher order. With the help of the Newman-Penrose formalism, we reexpress the GW propagation equations as a series of scalar equations and present explicit expressions for the Weyl scalars that describe the GW polarizations. By combining the approaches of solving geodesic deviation and transport equations, we construct a solvable system of equations that describes the evolution of GW polarization along null geodesics. This framework fills the gap left by the leading-order geometric optics and the Kirchhoff diffraction integral, neither of which captures the polarization characteristics of GWs during the lensing process. This work applies the above framework to a Schwarzschild lensing configuration. Through a rigorous theoretical formulation and detailed numerical analysis, our results reveal the emergence of apparent vector and scalar modes in lensed GW signals, which originate from the smearing of the polarization plane and distortion of the wavefront and do not represent genuine dynamical degrees of freedom but rather arise as the propagation effects imposed by gravitational lensing. △ Less

Submitted 27 July, 2026; originally announced July 2026.

Comments: 26 pages, 5 figures

RODR: Riemannian Orthogonally Decoupled Regularization for Disentangled Manifold Representation

Authors: Jiayu Zhu, Wenlai Zhao

Abstract: Point cloud denoising is essentially a geometric recovery task that aims to reconstruct the intrinsic structure of a smooth 2D Riemannian manifold embedded in R^3 from noisy, discrete ambient-space samples. Despite the remarkable progress of modern manifold-aware encoders and generative transport models in geometric representation learning, a fundamental objective-geometry mismatch remains underex… ▽ More Point cloud denoising is essentially a geometric recovery task that aims to reconstruct the intrinsic structure of a smooth 2D Riemannian manifold embedded in R^3 from noisy, discrete ambient-space samples. Despite the remarkable progress of modern manifold-aware encoders and generative transport models in geometric representation learning, a fundamental objective-geometry mismatch remains underexplored. Theoretically, we identified that this mismatched coupling leads to geometric gradient interference, where conflicting optimization objectives result in structural degradation and point clustering. We introduce Riemannian Orthogonally Decoupled Regularization (RODR) to reformulate the optimization trajectory by disentangling the normal (fitting) and tangential (distribution) components. Guided by a vector-attention and entropy-aware adaptive strategy, RODR effectively preserves high-fidelity geometric details while maintaining sampling uniformity. Experiments demonstrate that RODR reaches performance comparable to state-of-the-art baselines and suggests improved distribution regularity and reduced local aggregation effectively. Our work establishes a generic and interpretable framework for disentangled geometric optimization in point cloud processing. △ Less

Submitted 26 July, 2026; originally announced July 2026.

Comments: 10 pages, 4 figures

MSC Class: 68T05; 53B20 ACM Class: I.2.6; G.1.6; I.3.5

From RLVR to RLSVR: Task Transformation Induces Self-Verifiable Rewards for Open-Ended LLM Self-Improvement

Authors: Qinsi Wang, Jing Shi, Huazheng Wang, Kun Wan, Yiran Wu, Bo Liu, Qingyun Wu, Hai Helen Li, Yiran Chen, Handong Zhao, Wentian Zhao

Abstract: Reinforcement Learning with Verifiable Rewards (RLVR) has driven recent progress in reasoning-oriented large language models (LLMs) by enabling large-scale optimization. However, its applicability remains largely limited to domains such as mathematics and coding, where correctness can be deterministically verifiable. Open-ended tasks instead often rely on human preferences, reward models, or LLM-b… ▽ More Reinforcement Learning with Verifiable Rewards (RLVR) has driven recent progress in reasoning-oriented large language models (LLMs) by enabling large-scale optimization. However, its applicability remains largely limited to domains such as mathematics and coding, where correctness can be deterministically verifiable. Open-ended tasks instead often rely on human preferences, reward models, or LLM-based judges, introducing evaluation bias, judge capability bottlenecks, and additional inference costs. Drawing on the principle of self-supervised learning, which constructs pretext tasks to derive supervision from the data itself, we propose Reinforcement Learning with Self-Verifiable Rewards (RLSVR), a task-transformation-based training paradigm for extending RLVR to open-ended tasks. RLSVR transforms open-ended tasks into verifiable proxy environments whose internal rules and interaction outcomes automatically generate reward signals. We instantiate RLSVR with SpyRL, a Self-PlaY Reinforcement Learning method inspired by social deduction game Who Is the Spy?. Agents receive asymmetric information, complete the same target task, and vote to identify a designated spy. Because the spy identity is predetermined, voting outcomes provide fully verifiable rewards, while successful identification remains closely related to output quality. Experiments on text summarization, creative writing, and mathematical reasoning show that SpyRL outperforms existing self-improvement methods on non-verifiable tasks and yields consistent gains on verifiable reasoning tasks. These results demonstrate that task transformation can extend scalable RLVR-based self-improvement beyond inherently verifiable domains. Models and code have been released at https://github.com/wangqinsi1/RLSVR/tree/SpyRL. △ Less

Submitted 30 July, 2026; v1 submitted 26 July, 2026; originally announced July 2026.

Comments: COLM 2026