Searcharxiv⌕ Search

arXiv subjects

Wenxuan Wang

Publications and source records attributed to Wenxuan Wang.

At least 19 recordsLinked to original sources

Dual-Granularity Semantic Prompting for Language Guidance Infrared Small Target Detection

Infrared small target detection remains challenging due to limited feature representation and severe background interference, resulting in sub-optimal performance. While recent CLIP-inspired methods attempt to leverage textual guidance for detection, they are hindered by inaccurate text descriptions and reliance on manual annotations. To overcome these limitations, we propose DGSPNet, an end-to-end language prompt-driven framework. Our approach integrates dual-granularity semantic prompts: coarse-grained textual priors (e.g., 'infrared image', 'small target') and fine-grained personalized semantic descriptions derived through visual-to-textual mapping within the image space. This design not only facilitates learning fine-grained semantic information but also can inherently leverage language prompts during inference without relying on any annotation requirements. By fully leveraging the precision and conciseness of text descriptions, we further introduce a text-guide channel attention (TGCA) mechanism and text-guide spatial attention (TGSA) mechanism that enhances the model's sensitivity to potential targets across both low- and high-level feature spaces. Extensive experiments demonstrate that our method significantly improves detection accuracy and achieves state-of-the-art performance on three benchmark datasets.

cs.CV↗

Interactive 3D Cutaway Visualization for Subsurface Target Localization in Robot-Assisted Surgery

Purpose: Blended augmented reality (AR) overlays can provide ambiguous depth cues in which subsurface anatomy floats above the tissue surface. We investigate whether an interactive 3D cutaway visualization can improve robot-assisted target localization. Methods: We propose Virtual Box, an interactive 3D cutaway interface that allows users to independently manipulate the viewing perspective to inspect subsurface targets. In a controlled, counterbalanced within-subject tissue phantom study, 15 engineering students used Virtual Box and a depth-enhanced blended AR overlay to position a da Vinci instrument at virtual targets in a water tank. We used registered 3D transrectal ultrasound to measure 3D localization error and depth error. We also collected completion time and subjective ratings. Results: Virtual Box reduced mean 3D error from 12.92 to 4.96 mm (61.6% reduction) and depth error from 9.48 to 2.07 mm (78.2% reduction; both p < 0.001). Spatial understanding, depth perception, and confidence ratings were higher (all unadjusted p < 0.001), but completion time increased from 21.64 to 32.58 s (p = 0.008). Overall workload did not differ significantly. Conclusion: Virtual Box improved localization accuracy and perceived spatial understanding in this controlled task, with additional completion time, supporting further evaluation for surgical navigation.

cs.RO↗

Agent Planning Benchmark: A Diagnostic Framework for Planning Capabilities in LLM Agents

Planning is central to LLM agents: before acting, an agent must decompose goals, select tools, reason over constraints, and decide when a task is infeasible. Yet existing agent evaluations often report only end-to-end success, making it difficult to determine whether failures stem from planning or execution. We introduce Agent Planning Benchmark (APB), a planning-specific diagnostic benchmark with 4,209 multimodal cases across 22 domains and five settings, covering holistic planning, feedback-conditioned step-wise planning, and robustness under extraneous tools, broken tools, and unsolvable tasks. Across 12 MLLMs, APB reveals systematic weaknesses in long-horizon planning, tool-noise robustness, calibrated refusal, and inference-time refinement. We further validate APB on 200 ToolSandbox tasks and 200 $τ^2$-bench tasks, where APB-guided refinement consistently improves plan correctness, plan grade, and downstream execution metrics across three representative models. APB thus serves as an upstream diagnostic complement to execution benchmarks. The APB benchmark and code are available in \href{https://github.com/Mikivishy/AgentPlanningBenchmark}{this URL}.

cs.CL↗

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. Current Multimodal Large Language Models (MLLMs) are trained directly on complex downstream tasks. Does it guarantee their performance on these foundational visual capabilities? To systematically investigate this gap, we introduce VisFacor, 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 VisFacor, 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 55.9%. 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↗

Soloist: A Generative Model for Blind Voice Separation in the Wild

Real-world voice separation must recover individual voices from overlapping speech, environmental noise, and music, yet isolated source recordings are rarely available for supervision. We introduce Soloist, a generative model for blind voice separation that autoregressively generates individual voice streams as audio-codec sequences conditioned on the mixture, without requiring a target-speaker query or a predefined voice count. We propose a two-stage training paradigm to learn from synthetic mixtures and adapt to real recordings. The first stage combines supervised separation on synthetic mixtures with isolated source targets and an auxiliary task that predicts speaker timelines from real recordings, learning who speaks when from natural acoustics. The second stage applies reinforcement learning to real recordings, with rewards supplied by a model judge that compares candidate separations based on voice purity and content preservation without isolated source targets. Across diverse scenarios, Soloist recovers clean speech from multi-speaker overlap and noisy recordings, and extracts singing vocals from music accompaniment while maintaining relatively high pitch accuracy and intelligibility. On real-world recordings with unknown voice counts, Soloist also separates voices well. Finally, we integrate Soloist into a full-duplex conversation processing pipeline that assembles separated outputs into continuous speaker tracks, demonstrating its practical utility in real conversations.

cs.SD↗

A locking-free virtual element method for the three-dimensional quasi-static electroporoelasticity equations

We develop and analyze a locking-free virtual element method for the three-dimensional quasi-static electroporoelasticity equations. The pressure, electric, magnetic, and displacement fields are discretized by the lowest-order nodal, edge, face, and Stokes-like virtual element spaces, respectively, so that the scheme works on general polyhedral meshes, while the temporal discretization is carried out by a backward Euler scheme. The locking phenomenon is overcome by defining the Stokes-like space on a refined mesh $\widetilde{\mathcal{T}}_h$, on which we can construct an interpolation operator $\mathrm{I}_s$ satisfying $÷(\mathrm{I}_s\bm v)|_K=Π^K_0÷\bm v$, which renders the error estimates independent of the Lamé constant $λ$. Moreover, since the exact sequence $\curl\bm V^e_h=\bm V^f_{h,0}$ holds, the magnetic field is updated explicitly, reducing the four-field system to a three-field one and lowering the computational cost. Numerical experiments are provided to demonstrate the performance of the proposed method.

math.NA↗

SubtleMemory: A Benchmark for Fine-Grained Relational Memory Discrimination in Long-Horizon AI Agents

Persistent AI assistants, such as OpenClaw, accumulate large collections of related memories over long-term interactions. As these memories grow, they may reinforce one another, diverge across contexts, or directly conflict, making correct assistance depend on memory relations rather than isolated recall. Existing long-term memory benchmarks do not systematically probe how agents preserve and utilize such relations during downstream tasks. To address this gap, we introduce SubtleMemory, a benchmark for fine-grained relational memory discrimination in long-running AI agents. SubtleMemory constructs relation-controlled latent semantic artifacts whose variants instantiate complementary, nuanced, or contradictory relations, and embeds them into realistic user-agent histories, requiring agents to recover distributed relational structures during later queries and instructions. The benchmark contains 1,522 evaluation instances over 10 long histories, grounded in 1,090 relation-controlled memory-variant sets and spanning user-related and non-user-related queries. Evaluating six standalone memory systems, two Claw-style agents with native memory modules, and three Claw-style agents with plugin memory modules, we find that current systems remain weak on fine-grained relational memory discrimination. We further introduce diagnostic protocols that reveal distinct capability profiles across memory preservation, retrieval, and downstream reasoning stages.

cs.AI↗

Benchmarking Jailbreak Guardrails for Embodied Agents

Embodied agents powered by large language models and vision-language models are increasingly deployed in physical environments, but jailbreak attacks can induce these agents to perform physically harmful actions. A growing number of guardrail methods have been proposed to intercept dangerous behavior before it is executed, yet existing safety benchmarks evaluate the embodied models themselves, leaving it unclear how well these guardrails actually defend an embodied agent in practice. We present the first systematic evaluation of jailbreak guardrails for embodied agents. To compare guardrails under identical conditions, we build a pluggable evaluation framework that treats the embodied agent as a fixed backend and each guardrail as a module that can intervene at the perception, planning, or control stage. We subject six representative guardrails to template-based and automated jailbreak attacks as well as safe instructions, and assess them at the system level along three dimensions: defense effectiveness, measured by the bypass rate and the hazard success rate in the simulator; usability, measured by the false-positive rate and the task completion rate on safe instructions; and efficiency, measured by the latency overhead added at runtime. Experiments on guardrails that span different intervention stages, decision mechanisms, and input modalities reveal a clear trade-off among the three dimensions, and show that no single guardrail dominates in all settings. We further analyze how intervention stage, decision mechanism, and input modality shape safety outcomes, and we offer practical guidance for selecting and designing guardrails for embodied agents.

cs.AI↗

Internalizing Agent Experience into Diffusion Model Weights via On-Policy Context Distillation

Wrapping an image generation model in an agentic harness can effectively boost Text-to-Image task performance: the harness can leverage memory, skills, workflow orchestration, result verification, and iterative refinement to continually construct and revise prompts, thereby eliciting better images. These gains, however, remain external to the diffusion model and are realized only while the full harness runs. We propose Diffusion On-Policy Context Distillation (D-OPCD), which treats the agent-improved prompt as privileged context and distills the knowledge encoded in the agent harness into the weights of the diffusion model, so that the model retains part of the harness's benefit when conditioned on the original query alone. Using a Text-to-Image agent equipped with our proposed Auto Skill Evolver (ASE), we show that D-OPCD can internalize harness capabilities into the generator's weights, raising the average direct-generation score from 60.52 to 65.09 across four benchmarks. With this knowledge absorbed into the weights, the harness can shed its saturated skills and resume evolving: a second ASE round on the updated generator improves on a skill-free harness by additional 1.83 points, pointing toward text-to-image systems in which harness and model keep improving each other through continual co-evolution.

cs.AI↗

3D Software Synthesis Driven by Constraint-Expressive Intermediate Representation

Graphical user interface (UI) software has undergone a fundamental transformation from traditional two-dimensional (2D) desktop/web/mobile interfaces to spatial three-dimensional (3D) environments. While existing work has made remarkable success in automated 2D software generation, such as HTML/CSS and mobile app interface code synthesis, the generation of 3D software still remains under-explored. Current methods for 3D software generation usually generate the 3D environments as a whole and cannot modify or control specific elements in the software. Furthermore, these methods struggle to handle the complex spatial and semantic constraints inherent in the real world. To address the challenges, we present Scenethesis, a novel requirement-sensitive 3D software synthesis approach that maintains formal traceability between user specifications and generated 3D software. Scenethesis is built upon ScenethesisLang, a domain-specific language that serves as a granular constraint-aware intermediate representation (IR) to bridge natural language requirements and executable 3D software. It serves both as a comprehensive scene description language enabling fine-grained modification of 3D software elements and as a formal constraint-expressive specification language capable of expressing complex spatial constraints. By decomposing 3D software synthesis into stages operating on ScenethesisLang, Scenethesis enables independent verification, targeted modification, and systematic constraint satisfaction. Our evaluation demonstrates that Scenethesis accurately captures over 80% of user requirements and satisfies more than 90% of hard constraints while handling over 100 constraints simultaneously. Furthermore, Scenethesis achieves a 42.8% improvement in BLIP-2 visual evaluation scores compared to the state-of-the-art method.

cs.CV↗

Characterization of Sobolev regularity of plurisubharmonic functions

Let $f$ be a nonzero holomorphic germ at $0 \in \mathbb C^n$ with $f(0)=0$, and let $χ$ be a $C^2$ non-decreasing convex function on the left half-line. We establish sharp necessary and sufficient conditions for the local Sobolev regularity of the plurisubharmonic function $v=χ(\log|f|).$ The criteria for the $L^p$-integrability of the classical Laplacian and for $W^{1,p}$-regularity are given by a weighted integral involving $χ''$ and $χ'$, respectively, and depend on $f$ only through the smallest multiplicity of $\operatorname{Div}(f)$. We also obtain the $W^{2,p}_{\mathrm{loc}}$ criterion for $1<p<\infty$. At the endpoint $p=1$, we prove that \[ v\in W^{2,1}_{\mathrm{loc}} \quad\Longleftrightarrow\quad χ'\in L^1((-\infty,A)), \] equivalently, $v$ is locally bounded. As applications, we obtain counterexamples to Calderón-Zygmund theory and characterize a class of functions in the local Monge--Ampère domain. % and exhibit a natural family in $W^{2,1}_{\mathrm{loc}}$ whose limit fails to belong to $W^{2,1}_{\mathrm{loc}}$.

math.CV↗

Spin-polarized Superconductivity and High-Chern Insulators in Twisted Rhombohedral Graphene Family

Rhombohedral multilayer graphene has emerged as a remarkably versatile platform for exploring strong correlation driven quantum states arising from low-energy topological flat bands. When reconstructed by the moire superlattice, these bands host a wide range of emergent novel states, including integer and fractional Chern insulators and unconventional superconductivity. Here, we firstly report the simultaneous emergence of widespread spin polarized SC and high Chern insulators in twisted bilayer multilayer RMG system 2+n where n=4,5,6. The SC states in 2+n system exhibit different responses to the in plane magnetic field, with SC being suppressed, enhanced and induced by in plane magnetic field . The latter two are consistent with spin-triplet pairing. Along with SC, angle and layer dependent HCIs with tunable Chern numbers emerge. Moreover, the fractional high Chern insulator in the system survives under high in plane magnetic field which can induce SC in the same device. Our work not only establishs twisted bilayer multilayer rhombohedral graphene as a unified platform for studying SC and high Chern insulators, but also opens a pathway towards multiple copropagating chiral Majorana channels by coupling spin-polarized SC to high Chern insulators.

cond-mat.mes-hall↗

The PIMMUR Principles: Ensuring Validity in Collective Behavior of LLM Societies

Large language models (LLMs) are increasingly used to simulate human collective behavior, yet claims that such simulations are human-like remain largely untested. We conducted a systematic audit (pre-registered on OSF) of LLM-based social simulations across four databases (Scopus, IEEE Xplore, ACM Digital Library, and arXiv). Across 576 studies reported in 350 recent papers, we applied six methodological evaluations: agent Profile, Interaction, Memory, Minimal-Control, Unawareness, and Realism (PIMMUR). Coding every study against pre-specified rules, we revealed that PIM were met more often than MUR. Frontier LLMs correctly identified the underlying social experiment in 65.2% of cases, and 50.6% of prompts imposed constraints that pre-determined the outcome. These compliance rates are upper bounds, because incomplete methodological reporting (for example, unreleased prompts) limits the available evidence. Reproducing five representative experiments (e.g., opinion dynamics), we found that reported collective phenomena often vanish or reverse once PIMMUR principles are enforced, indicating that many "emergent" behaviors are methodological artifacts rather than genuine social dynamics. Current LLM simulations may therefore capture model-specific biases rather than universal features of human social behavior, raising concerns about their use as scientific proxies for human society.

cs.CL↗

Medical Reasoning in the Era of LLMs: A Systematic Review of Enhancement Techniques and Applications

The proliferation of Large Language Models (LLMs) in medicine has enabled impressive capabilities, yet a critical gap remains in their ability to perform systematic, transparent, and verifiable reasoning, a cornerstone of clinical practice. This has catalyzed a shift from single-step answer generation to the development of LLMs explicitly designed for medical reasoning. This paper provides the first systematic review of this emerging field. We propose a taxonomy of reasoning enhancement techniques, categorized into training-time strategies (e.g., supervised fine-tuning, reinforcement learning) and test-time mechanisms (e.g., prompt engineering, multi-agent systems). We analyze how these techniques are applied across different data modalities (text, image, code) and in key clinical applications such as diagnosis, education, and treatment planning. Furthermore, we survey the evolution of evaluation benchmarks from simple accuracy metrics to sophisticated assessments of reasoning quality and visual interpretability. Based on an analysis of 60 seminal studies from 2022-2025, we conclude by identifying critical challenges, including the faithfulness-plausibility gap and the need for native multimodal reasoning, and outlining future directions toward building efficient, robust, and sociotechnically responsible medical AI.

cs.CL↗

ScopeBench: Evaluating Cultural Norm Scope Awareness in LLMs

Safety evaluation in large language models treats harm as culturally invariant, yet a model cannot correctly apply a safety norm without knowing whose norms bind. We introduce ScopeBench, a benchmark that treats cultural scope as a first-class safety dimension. It comprises 3,323 norms spanning 49 countries across nine cultural clusters, each annotated as Universal or Specific. We define three scope error types identically across four evaluation tasks: Scope Inflation, Scope Deflation, and Scope Displacement. Evaluating 21 frontier LLMs across seven model families, we surface three structural failures. First, the dominant error type inverts with prompt format: Scope Deflation reaches 70% in single-answer classification, while Scope Inflation reaches 33% when each candidate culture is judged independently. Second, scope-classification accuracy is empirically decoupled from behavioral safety: the six classifiers achieving at least 97% Task 1 accuracy still produce harmful responses to 40-42% of scope-violating statements. Third, no model is safe in interaction, with harmful-response rates ranging from 29% to 51% across the cohort, excluding one model that systematically refuses the task. We release all data, code, and model outputs to establish cultural scope as a distinct alignment axis.

cs.CL↗

A Survey of Large Models in Sports

Sports have witnessed growing global enthusiasm in recent years, serving as a vital force for physical health, cultural exchange, social connection, and economic growth. The rapid advancement of large models, particularly (multimodal) large language models (M)LLMs, has demonstrated transformative potential to reshape sports understanding, analysis, and interaction across diverse domains. This paper presents a comprehensive survey of large models in sports, including (i) an overview of tasks and applications across different participant groups; (ii) a detailed analysis of sports-related datasets and benchmarks; and (iii) a critical discussion of current challenges and future directions. Our goal is to establish a foundation for advancing research and practical development of large-model-driven sports intelligence. An open-source GitHub repository is maintained at: https://github.com/Road2Redemption/Awesome_Large_Models_In_Sports1.

cs.CL↗

RealGUINoise: An Interactive Cross-Platform Benchmark for GUI Agent Robustness under Real-World Interface Noise

Graphical User Interface (GUI) agents and Computer-Using Agents (CUAs) are rapidly becoming practical tools. However, real-world deployment increasingly exposes performance failures and safety risks, while a major yet underexplored source of these problems lies in the complex and noisy conditions of everyday interfaces. Existing benchmarks largely assume clean environments or focus narrowly on security-specific settings, and lack a unified framework for consistent, automated end-to-end evaluation across diverse agents and platforms. Hence, we introduce RealGUINoise, an interactive cross-platform, extensible benchmark for systematically evaluating GUI Agents under common realistic interface noise in fully interactive environments. Specifically, RealGUINoise comprises 42 noise types spanning web, desktop, and mobile tasks and integrates 7 representative agent frameworks. It evaluates these agents on real-world daily tasks through real-time interaction, comparing their performance against task-specific golden rubrics and clean-environment trajectories in terms of reliability, safety, and trajectory-level behavior. Our experiments show that these noises not only degrade task performance but also substantially redirect agents' action trajectories and increase their propensity for unsafe behavior. These findings expose a critical gap between capability in clean environments and dependable operation in real-world settings, establishing RealGUINoise as a testbed for developing more robust and trustworthy GUI agents.

cs.HC↗

From Sports to Safety: Benchmarking Proactive Risk Inference in MLLMs

Timely anticipation of physical hazards is essential for real-world safety, yet existing MLLM evaluations focus on harmful content or general risks, leaving proactive physical hazard prediction underexplored. Sports provide a well-suited testbed: accident causes span diverse injury dimensions and pre-accident spatiotemporal cues draw on reasoning capabilities shared with broader safety domains such as autonomous driving and fall detection. We introduce SPRINT (Sports Proactive Risk INference Testbed), a benchmark of 2,888 real-world sports videos (2,440 accident, 448 safe controls) spanning 14 sports and 3 environmental settings. Accident videos feature fine-grained annotations of early hazard cues, accident timing, and hierarchical causes; safe videos are manually verified as accident-free and serve to diagnose prompt-induced false alarms. Evaluating state-of-the-art MLLMs under diverse prompts and temporal windows reveals a sharp gap between hazard sensitivity and understanding: the best model exceeds 95% in signaling hazards yet falls below 50% in identifying their causes. Diagnostic experiments further show that explicit danger queries trigger severe false alarms even on hazard-free videos. These findings indicate that current MLLMs exhibit only superficial proactive safety, lacking stable, cause-grounded early warning, and underscore the need for reliable proactive safety in dynamic physical environments. Data and code will be open-sourced upon acceptance.

cs.CV↗