arXiv subjects

Wenxuan Wang

Publications and source records attributed to Wenxuan Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Rethinking Depth Pruning for Vision Transformers: A Heterogeneity-Aware Perspective

While prior studies have successfully compressed vision Transformers (ViTs) through various pruning techniques, most have concentrated on width pruning to achieve significant reductions in model size. Depth pruning, which removes entire layers from a ViT, is notoriously difficult for accuracy recovery despite its potential to deliver higher speedups, limiting the acceleration achieved by existing joint width-and-depth pruning methods. In this work, we reveal that the failure of existing depth pruning methods lies in their neglect of heterogeneity between different layers, and we introduce HetDPT, a heterogeneity-aware depth pruning method that avoids dimension mismatch. Comprehensive experiments on ImageNet-1K, CIFAR-100, COCO, and ADE20K validate our method: HetDPT achieves a 1.58$\times$ speedup for DeiT-B while maintaining accuracy and a 1.39$\times$ speedup for DeiT-S with nearly no accuracy degradation. Furthermore, when combined with width pruning, HetDPT+ sets a new state-of-the-art record in extreme ViT pruning, enhancing the acceleration ratio from 4.24$\times$ to 5.19$\times$ for the Isomorphic-Pruning-2.6G configuration while maintaining near-lossless accuracy; our code is available at https://github.com/Efficient-AI-for-All/HetDPT.

cs.CV↗

arXiv2026

Beyond SFT-to-RL: Pre-alignment via Black-Box On-Policy Distillation for Multimodal RL

The standard post-training recipe for large multimodal models (LMMs) applies supervised fine-tuning (SFT) on curated demonstrations followed by reinforcement learning with verifiable rewards (RLVR). However, SFT introduces distributional drift that neither preserves the model's original capabilities nor faithfully matches the supervision distribution. This problem is further amplified in multimodal reasoning, where perception errors and reasoning failures follow distinct drift patterns that compound during subsequent RL. We introduce PRISM, a three-stage pipeline that mitigates this drift by inserting an explicit distribution-alignment stage between SFT and RLVR. Building on the principle of on-policy distillation (OPD), PRISM casts alignment as a black-box, response-level adversarial game between the policy and a Mixture-of-Experts (MoE) discriminator with dedicated perception and reasoning experts, providing disentangled corrective signals that steer the policy toward the supervision distribution without requiring access to teacher logits. While 1.26M public demonstrations suffice for broad SFT initialization, distribution alignment demands higher-fidelity supervision; we therefore curate 113K additional demonstrations from Gemini 3 Flash, featuring dense visual grounding and step-by-step reasoning on the hardest unsolved problems. Experiments on Qwen3-VL show that PRISM consistently improves downstream RLVR performance across multiple RL algorithms (GRPO, DAPO, GSPO) and diverse multimodal benchmarks, improving average accuracy by +4.4 and +6.0 points over the SFT-to-RLVR baseline on 4B and 8B, respectively. Our code, data, and model checkpoints are publicly available at https://github.com/XIAO4579/PRISM.

cs.CV↗

arXiv2026

A Survey of Deep Learning for Geometry Problem Solving

Geometry problem solving, a crucial aspect of mathematical reasoning, is vital across various domains, including education, the assessment of AI's mathematical abilities, and multimodal capability evaluation. The recent surge in deep learning technologies, particularly the emergence of multimodal large language models, has significantly accelerated research in this area. This paper presents a survey of the applications of deep learning in geometry problem solving, including (i) a comprehensive summary of the relevant tasks in geometry problem solving; (ii) a thorough review of related deep learning methods; (iii) a detailed analysis of evaluation metrics and methods; and (iv) a critical discussion of state-of-the-art performance, existing challenges, and promising future directions. Our objective is to offer a comprehensive and practical reference of deep learning for geometry problem solving, thereby fostering further advancements in this field. We maintain a list of relevant papers: https://github.com/majianz/dl4gps.

cs.CL↗

arXiv2026

Probing Multimodal Large Language Models on Cognitive Biases in Chinese Short-Video Misinformation

Short-video platforms have become major channels for misinformation, where deceptive claims frequently leverage visual experiments and social cues. While Multimodal Large Language Models (MLLMs) have demonstrated impressive reasoning capabilities, their robustness against misinformation entangled with cognitive biases remains under-explored. In this paper, we introduce a comprehensive evaluation framework using a high-quality, manually annotated dataset of 200 short videos spanning four health domains. This dataset provides fine-grained annotations for three deceptive patterns-experimental errors, logical fallacies, and fabricated claims-each verified by evidence such as national standards and academic literature. We evaluate eight frontier MLLMs across five modality settings. Experimental results demonstrate that Gemini-2.5-Pro achieves the highest performance in the multimodal setting with a belief score of 71.5/100, while o3 performs the worst at 35.2. Furthermore, we investigate social cues that induce false beliefs in videos and find that models are susceptible to biases like authoritative channel IDs.

cs.CL↗

arXiv2026

MOSS-Audio Technical Report

MOSS-Audio is a unified audio-language model for speech, environmental sound, and music understanding, supporting audio captioning, time-aware question answering, timestamped transcription, and audio-grounded reasoning. MOSS-Audio couples a dedicated audio encoder with a modality adapter and a large language model: the encoder produces 12.5 Hz temporal representations, the adapter projects them into the decoder space, and the decoder generates autoregressive text outputs. Two design choices are central to the system: DeepStack cross-layer feature injection, which exposes the decoder to acoustic information from multiple encoder depths, and time markers, which provide explicit temporal cues by inserting timestamp markers into the audio-token stream. At the data level, we design an event-preserving audio annotation pipeline that segments raw audio at coherent event boundaries, applies branch-specific annotation to speech, music, and general audio, and merges the results into unified captions for pretraining. The intermediate branch-specific captions are further retained to support the construction of task-oriented SFT data. The model is pretrained on large-scale audio-language data, with time-aware objectives incorporated to support temporal grounding, and then undergoes multi-stage post-training to enhance instruction following and audio-grounded reasoning. We release 4B and 8B variants in both Instruct and Thinking configurations. MOSS-Audio achieves strong performance across general audio understanding, speech captioning, ASR, and timestamped ASR, positioning it as a promising understanding foundation for future voice agents.

cs.SD↗

arXiv2026

Fractional High-Chern Insulator in Twisted Rhombohedral Graphene

The realization of fractional Chern insulators opens up the possibility of exploring fractionally charged excitations and anyonic statistics in the absence of a magnetic field. A central question is whether lattice-based systems can give rise to radically new states, distinct from those observed in traditional fractional quantum Hall systems. In this work, we investigate a new type of moiré flat band system composed of Bernal bilayer graphene and rhombohedral tetralayer graphene. We discover an unprecedented richness of quantum anomalous Hall insulators with Chern numbers from C = 1 to C = 7 at v = 1 and around v = 3. Remarkably, we observe an exotic fractional Chern insulator with C = 7/3 around v = 2/3 which is beyond all known fractional Chern insulators described by either the Jain sequence or current high Chern theory. Our work expands the understanding of fractionally charged excitations beyond the Landau level basis and offers a new moire platform for exploring anyons.

cond-mat.mes-hall↗

arXiv2026

Image-Conditioned Instance Prompt Network for Referring Remote Sensing Image Segmentation

Referring Remote Sensing Image Segmentation (RRSIS) is a situated, task-driven cross-modal task related to the embodied perception paradigm, requiring models to align visual-spatial features with linguistic intentions for precise target perception. Recent research has focused on refining the granularity of textual features and optimizing image-text feature fusion to better guide target feature representations. However, insufficient descriptive granularity and sensitivity to semantic shifts can cause bottlenecks in cross-modal feature fusion. To address these issues, we propose the Image-Conditioned Instance Prompt Network (ICIPNet) with Bilateral Information Fusion, which is designed to alleviate bottlenecks in cross-modal feature fusion. ICIPNet introduces an Image-Conditioned Instance Prompt (ICIP) module to generate self-adaptive visual and semantic representations without external knowledge. The Bilateral Information Fusion (BIF) module enhances feature fusion along the token and channel dimensions. Experiments demonstrate that the proposed ICIPNet outperforms existing RRSIS models.

cs.CV↗

arXiv2026

AutoMonitor-Bench: Evaluating the Reliability of LLM-Based Misbehavior Monitor

We introduce AutoMonitor-Bench, the first benchmark designed to systematically evaluate the reliability of LLM-based misbehavior monitors across diverse tasks and failure modes. AutoMonitor-Bench consists of 3,010 carefully annotated test samples spanning question answering, code generation, and reasoning, with paired misbehavior and benign instances. We evaluate monitors using two complementary metrics: Miss Rate (MR) and False Alarm Rate (FAR), capturing failures to detect misbehavior and oversensitivity to benign behavior, respectively. Evaluating 12 proprietary and 10 open-source LLMs, we observe substantial variability in monitoring performance and a consistent trade-off between MR and FAR, revealing an inherent safety-utility tension. To further explore the limits of monitor reliability, we construct a large-scale training corpus of 153,581 samples and fine-tune Qwen3-4B-Instruction to investigate whether training on known, relatively easy-to-construct misbehavior datasets improves monitoring performance on unseen and more implicit misbehaviors. Our results highlight the challenges of reliable, scalable misbehavior monitoring and motivate future work on task-aware designing and training strategies for LLM-based monitors.

cs.CL↗

arXiv2026

Human Cognitive Benchmarks Reveal Foundational Visual Gaps in MLLMs

Humans develop perception through a bottom-up hierarchy: from basic primitives and Gestalt principles to high-level semantics. In contrast, current Multimodal Large Language Models (MLLMs) are trained directly on complex downstream tasks, often bypassing these foundational visual capabilities. To systematically investigate this gap, we introduce VisFactor, a benchmark that digitizes 20 vision-centric subtests from FRCT, a well-established cognitive psychology assessment spanning four domains of human visual cognition. Furthermore, we design algorithms to automatically construct and validate unlimited test cases with controllable difficulty. Using VisFactor, we evaluate 39 frontier MLLMs, including both proprietary (e.g., GPT, Gemini) and open-source (e.g., LLaMA, Qwen) models. The best model achieves a score of only 54.0%. Analysis reveals good internal consistency (Cronbach's alpha = 0.94) and construct validity (compared to existing vision benchmarks). Models consistently fail on tasks such as mental rotation, spatial relation inference, and figure-ground discrimination, regardless of model size or prompting strategy. These findings suggest that performance improvements on existing general benchmarks might represent castles in the air instead of a genuine mastery of human-like visual cognition.

cs.CV↗

arXiv2026

Identifying the Achilles' Heel: An Iterative Method for Dynamically Uncovering Factual Errors in Large Language Models

Large Language Models (LLMs) like ChatGPT are foundational in various applications due to their extensive knowledge from pre-training and fine-tuning. Despite this, they are prone to generating factual and commonsense errors, raising concerns in critical areas like healthcare, journalism, and education to mislead users. Current methods for evaluating LLMs' veracity are limited by the need for extensive human labor, test data contamination, or limited scope, hindering efficient and effective exposure of errors. To address these challenges, we propose HalluHunter, a novel, fully automated framework for systematically uncovering factual inaccuracies in LLMs. HalluHunter employs a knowledge-graph-based approach, extracting fact triplets to generate diverse question types for single- and multi-hop reasoning using rule-based Natural Language Processing (NLP) techniques. Its iterative process starts with random triplet selection for question generation, followed by adaptive selection in subsequent iterations, targeting triplets where LLMs frequently err based on their performance analysis. Our extensive tests on nine prominent LLMs reveal that HalluHunter can trigger factual errors in up to 55% of tested questions. Moreover, we demonstrate that HalluHunter's test cases, particularly in adaptive selection, could further expose the weaknesses in benchmarking the factuality in LLMs meanwhile maintaining the coverage of questions. All code, data, and results are available at this link: https://github.com/Mysterchan/HalluHunter.

cs.SE↗

arXiv2026

Learning to Ask: When LLM Agents Meet Unclear Instruction

Equipped with the capability to call functions, modern large language models (LLMs) can leverage external tools for addressing a range of tasks unattainable through language skills alone. However, the effective execution of these tools relies heavily not just on the advanced capabilities of LLMs but also on precise user instructions, which often cannot be ensured in the real world. To evaluate the performance of LLMs tool-use under imperfect instructions, we meticulously examine the real-world instructions queried from users, analyze the error patterns, and build a challenging tool-use benchmark called Noisy ToolBench (NoisyToolBench). We find that due to the next-token prediction training objective, LLMs tend to arbitrarily generate the missed argument, which may lead to hallucinations and risks. To address this issue, we propose a novel framework, Ask-when-Needed (AwN), which prompts LLMs to ask questions to users whenever they encounter obstacles due to unclear instructions. Moreover, to reduce the manual labor involved in user-LLM interaction and assess LLMs performance in tool utilization from both accuracy and efficiency perspectives, we design an automated evaluation tool named ToolEvaluator. Our experiments demonstrate that the AwN significantly outperforms existing frameworks for tool learning in the NoisyToolBench. We will release all related code and datasets to support future research.

cs.CL↗

arXiv2026

A Survey on the Safety and Security Threats of Computer-Using Agents: JARVIS or Ultron?

Recently, AI-driven interactions with computing devices have advanced from basic prototype tools to sophisticated, LLM-based systems that emulate human-like operations in graphical user interfaces. We are now witnessing the emergence of \emph{Computer-Using Agents} (CUAs), capable of autonomously performing tasks such as navigating desktop applications, web pages, and mobile apps. However, as these agents grow in capability, they also introduce novel safety and security risks. Vulnerabilities in LLM-driven reasoning, with the added complexity of integrating multiple software components and multimodal inputs, further complicate the security landscape. In this paper, we present a systematization of knowledge on the safety and security threats of CUAs. We conduct a comprehensive literature review and distill our findings along four research objectives: \textit{\textbf{(i)}} define the CUA that suits safety analysis; \textit{\textbf{(ii)} } categorize current safety threats among CUAs; \textit{\textbf{(iii)}} propose a comprehensive taxonomy of existing defensive strategies; \textit{\textbf{(iv)}} summarize prevailing benchmarks, datasets, and evaluation metrics used to assess the safety and performance of CUAs. Building on these insights, our work provides future researchers with a structured foundation for exploring unexplored vulnerabilities and offers practitioners actionable guidance in designing and deploying secure Computer-Using Agents.

cs.CL↗

arXiv2026

Beyond the Leaderboard: Rethinking Medical Benchmarks for Large Language Models

Large language models (LLMs) show significant potential in healthcare, prompting numerous benchmarks to evaluate their capabilities. However, concerns persist regarding the reliability of these benchmarks, which often lack clinical fidelity, robust data management, and safety-oriented evaluation metrics. To address these shortcomings, we introduce MedCheck, the first lifecycle-oriented assessment framework specifically designed for medical benchmarks. Our framework deconstructs a benchmark's development into five continuous stages, from design to governance, and provides a comprehensive checklist of 46 medically-tailored criteria. Using MedCheck, we conducted an in-depth empirical evaluation of 53 medical LLM benchmarks. Our analysis uncovers widespread, systemic issues, including a profound disconnect from clinical practice, a crisis of data integrity due to unmitigated contamination risks, and a systematic neglect of safety-critical evaluation dimensions like model robustness and uncertainty awareness. Based on these findings, MedCheck serves as both a diagnostic tool for existing benchmarks and an actionable guideline to foster a more standardized, reliable, and transparent approach to evaluating AI in healthcare.

cs.CL↗

arXiv2026

Inference-Time Scaling of Verification: Self-Evolving Deep Research Agents via Test-Time Rubric-Guided Verification

Recent advances in Deep Research Agents (DRAs) are transforming automated knowledge discovery and problem-solving. While the majority of existing efforts focus on enhancing policy capabilities via post-training, we propose an alternative paradigm: self-evolving the agent's ability by iteratively verifying the policy model's outputs, guided by meticulously crafted rubrics. This approach gives rise to the inference-time scaling of verification, wherein an agent self-improves by evaluating its generated answers to produce iterative feedback and refinements. We derive the rubrics based on an automatically constructed DRA Failure Taxonomy, which systematically classifies agent failures into five major categories and thirteen sub-categories. We present DeepVerifier, a rubrics-based outcome reward verifier that leverages the asymmetry of verification and outperforms vanilla agent-as-judge and LLM judge baselines by 12%-48% in meta-evaluation F1 score. To enable practical self-evolution, DeepVerifier integrates as a plug-and-play module during test-time inference. The verifier produces detailed rubric-based feedback, which is fed back to the agent for iterative bootstrapping, refining responses without additional training. This test-time scaling delivers 8%-11% accuracy gains on challenging subsets of GAIA and XBench-DeepSearch when powered by capable closed-source LLMs. Finally, to support open-source advancement, we release DeepVerifier-4K, a curated supervised fine-tuning dataset of 4,646 high-quality agent steps focused on DRA verification. These examples emphasize reflection and self-critique, enabling open models to develop robust verification capabilities.

cs.AI↗

arXiv2026

AgentDoG: A Diagnostic Guardrail Framework for AI Agent Safety and Security

The rise of AI agents introduces complex safety and security challenges arising from autonomous tool use and environmental interactions. Current guardrail models lack agentic risk awareness and transparency in risk diagnosis. To introduce an agentic guardrail that covers complex and numerous risky behaviors, we first propose a unified three-dimensional taxonomy that orthogonally categorizes agentic risks by their source (where), failure mode (how), and consequence (what). Guided by this structured and hierarchical taxonomy, we introduce a new fine-grained agentic safety benchmark (ATBench) and a Diagnostic Guardrail framework for agent safety and security (AgentDoG). AgentDoG provides fine-grained and contextual monitoring across agent trajectories. More Crucially, AgentDoG can diagnose the root causes of unsafe actions and seemingly safe but unreasonable actions, offering provenance and transparency beyond binary labels to facilitate effective agent alignment. AgentDoG variants are available in three sizes (4B, 7B, and 8B parameters) across Qwen and Llama model families. Extensive experimental results demonstrate that AgentDoG achieves state-of-the-art performance in agentic safety moderation in diverse and complex interactive scenarios. All models and datasets are openly released.

cs.AI↗

arXiv2026

EgoEsportsQA: An Egocentric Video Benchmark for Perception and Reasoning in Esports

While video large language models (Video-LLMs) excel in understanding slow-paced, real-world egocentric videos, their capabilities in high-velocity, information-dense virtual environments remain under-explored. Existing benchmarks focus on daily activities, yet lack a rigorous testbed for evaluating fast, rule-bound reasoning in virtual scenarios. To fill this gap, we introduce EgoEsportsQA, a pioneering video question-answering (QA) benchmark for grounding perception and reasoning in expert esports knowledge. We curate 1,745 high-quality QA pairs from professional matches across 3 first-person shooter games via a scalable six-stage pipeline. These questions are structured into a two-dimensional decoupled taxonomy: 11 sub-tasks in the cognitive capability dimension (covering perception and reasoning levels) and 6 sub-tasks in the esports knowledge dimension. Comprehensive evaluations of state-of-the-art Video-LLMs reveal that current models still fail to achieve satisfactory performance, with the best model only 71.58%. The results expose notable gaps across both axes: models exhibit stronger capabilities in basic visual perception than in deep tactical reasoning, and they grasp overall macro-progression better than fine-grained micro-operations. Extensive ablation experiments demonstrate the intrinsic weaknesses of current Video-LLM architectures. Further analysis suggests that our dataset not only reveals the connections between real-world and virtual egocentric domains, but also offers guidance for optimizing downstream esports applications, thereby fostering the future advancement of Video-LLMs in various egocentric environments.

cs.CV↗

arXiv2026

ComPASS: Towards Personalized Agentic Social Support via Tool-Augmented Companionship

Developing compassionate interactive systems requires agents to not only understand user emotions but also provide diverse, substantive support. While recent works explore empathetic dialogue generation, they remain limited in response form and content, struggling to satisfy diverse needs across users and contexts. To address this, we explore empowering agents with external tools to execute diverse actions. Grounded in the psychological concept of "social support", this paradigm delivers substantive, human-like companionship. Specifically, we first design a dozen user-centric tools simulating various multimedia applications, which can cover different types of social support behaviors in human-agent interaction scenarios. We then construct ComPASS-Bench, the first personalized social support benchmark for LLM-based agents, via multi-step automated synthesis and manual refinement. Based on ComPASS-Bench, we further synthesize tool use records to fine-tune the Qwen3-8B model, yielding a task-specific ComPASS-Qwen. Comprehensive evaluations across two settings reveal that while the evaluated LLMs can generate valid tool-calling requests with high success rates, significant gaps remain in final response quality. Moreover, tool-augmented responses achieve better overall performance than directly producing conversational empathy. Notably, our trained ComPASS-Qwen demonstrates substantial improvements over its base model, achieving comparable performance to several large-scale models. Our code and data are available at https://github.com/hzp3517/ComPASS.

cs.CL↗

arXiv2026

Curing Miracle Steps in LLM Mathematical Reasoning with Rubric Rewards

In this paper, we observe that current models are susceptible to reward hacking, leading to a substantial overestimation of a model's reasoning ability. This is evidenced by a high incidence of false positives-solutions that reach the correct answer through an unsound process. Through a systematic analysis with human verification, we establish a taxonomy of these failure modes, identifying patterns like Miracle Steps-abrupt jumps to a correct output without a valid preceding derivation. Probing experiments suggest that these Miracle Steps are linked to answer-recall shortcuts, including memorization from pretraining, where the model accesses the correct answer independently of its reasoning chain. To mitigate this systemic issue, we introduce the Rubric Reward Model (RRM), a process-oriented reward function that evaluates the entire reasoning trajectory against problem-specific rubrics. The RRM explicitly penalizes logical flaws and encourages rigorous deduction. When integrated into an RL pipeline, RRM-based training consistently outperforms outcome-only supervision across four math benchmarks. Notably, it boosts Verified Pass@1024 on AIME2024 from 26.7% to 62.6% and reduces the incidence of Miracle Steps by 71%. Our work demonstrates that rewarding the solution process is crucial for building accurate and reliable models.

cs.CL↗