Papers

New on arXiv

Sat 5 Sep 2026

213 papers

Do Chess Explanations Reflect Model Decisions? Behavioral and Token-Level Tests of LLM Reasoning FaithfulnessLLM reasoning → Reasoning for agents, planning and code
MLLMs Hallucinate when Information Distribution Drifts in Synergy HeadsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
OpenDiscoveryTrace: Process Traces for Evaluating AI Scientist WorkflowsLLM reasoning → Reasoning for agents, planning and code
CST-WM: A Causally Structured World Model for Embodied Visual TrackingWorld models → Embodied and robot world models
SignDino: Self-Supervised Sign Language Representation Learning via Temporal-Axis Self-DistillationKnowledge distillation → Speech and audio distillation
Steering Geometry: Validating Human Value Geometry in LLM Steering SpaceLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment +1
Object-Aware Background-Controlled Editing via Weighted Velocity GuidanceDiffusion models → Image generation, text-to-image and editing
GloVLA: Let Geometry Move and Local VLA Interact for Robust Object-Centric Manipulation in Unstructured EnvironmentsVision-language-action models → 3D and spatial-aware VLA
MobileVLA-R1 2.0: RL-Enhanced Reasoning for Mobile Robot ControlVision-language-action models → Reasoning and chain-of-thought in VLA
It is Not Yet Another Tool: Creating and Deploying an Agentic AI Companion in a Security Operations CenterLLM agents and tool use → Agent threats, guardrails and governance
VDiff-Bench: A Challenging Benchmark for Fine-Grained Image Difference IdentificationMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Scratchy: Visual-Scratchpad Multimodal Reasoning for Cryptographic Proof Generation in EasyCryptLLM reasoning → Reasoning for agents, planning and code
RefGuard: Identity-Aware Language-Guided Robot Manipulation via Joint Target-Anchor-Frame GroundingVision-language-action models → 3D and spatial-aware VLA
From Gaze to Meaning: A Training-Free AI Agent for Unified Grounding and ExplanationLLM agents and tool use → Multimodal and embodied agents
SCIRIGOR:Evaluating Open-Ended Scientific Analysis Beyond Final ScoresLLM agents and tool use → Web, search, and research agents
Spectral Prioritized Sweeping in Nonstationary Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Diamond Agent: Agentic Control of Federated HPC Resources as a ServiceLLM agents and tool use → Software engineering coding agents
ExpertLens: Visualizing Embedding Spaces for Post-Hoc Explainability in MoE Enhanced RetrieversMixture-of-experts → MoE large language models
Protocol Compression Changes Which Party Pays: Bilateral Cost in Cross-Organization LLM Agent CommunicationLLM agents and tool use → Agent memory systems
Substrate-Portable Execution for Production LLM WorkflowsLLM agents and tool use → Software engineering coding agents
PAGR: Proof-Carrying Algebraic-Geometric Retrieval: A Quiver-, Provenance-, and Sheaf-Theoretic Framework for Grounded LLM RetrievalRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
SAP: State-Guided Data Synthesis with Argument Provenance for Multi-Turn Tool UseLLM agents and tool use → Tool learning and function calling +1
Where Success Breaks: Failure-Boundary Learning for Robust Vision-Language-Action ModelsVision-language-action models → Failure recovery and robustness
VERPO: Verified Evidence Regularized Policy OptimizationLLM reasoning → Reinforcement learning for reasoning
Automatic Red Teaming for Implicit Vulnerabilities of Text-to-Image ModelsLLM safety and security → Vision-language and text-to-image safety
LayerRoute: Action-Conditioned Mixture-of-Layers Routing for Vision-Language-Action PoliciesMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Beyond Retraining-Free MoE Compression: A Cost-Normalized Study of Post-Compression AdjustmentMixture-of-experts → Efficient MoE training and inference systems
ACE: Adapter Consolidation across Experts for Parameter-Efficient Fine-Tuning of MoE LLMsMixture-of-experts → Mixture of LoRA experts
Don't Lose Entities from Retrieval to Generation: Dual Entity Recovery RAG for multi-hop QARetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
Explaining AI Agents Through Execution TracesLLM agents and tool use → Software engineering coding agents
SCoCaT: Success Conditioned Constrained Reinforcement Learning for Spacecraft DockingDeep reinforcement learning methods → RL for robot navigation and UAV control
DAREBench: Deployment-Aware and Reliable Evaluation of Models as AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
GradeTrap: Authority Cues in Images Shift VLM Judgments Despite Explicit Instructions to Ignore ThemMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
DriveZero: End-to-End Driving Beyond Human DemonstrationsEnd-to-end autonomous driving → Imitation-learning end-to-end planners
Generating Adversarial Texts for Machine Translation via GRPOLLM reasoning → Reinforcement learning for reasoning
Evaluating Deep-Search Agents under Hierarchical Web Evidence PoisoningLLM agents and tool use → Agent threats, guardrails and governance
How to Learn from What a Human Would Avoid? Intervention-Aware World Models with Real-World RL for Dexterous ManipulationWorld models → Latent dynamics models and model-based RL +1
Memory in Deep Time-Series ModelsState-space sequence models (S4, Mamba) → Core SSM architectures for language
Geometry-Aware Test-Time Learning for Quantitative Spatial ReasoningLLM reasoning → Spatial and embodied reasoning +1
Agentic Pressure: The Endogenous Entropy of Reliable AutonomyLLM agents and tool use → Agent threats, guardrails and governance
MOAE: Multi-Objective Agent Evolution with Pareto-Preserving SearchLLM agents and tool use → General LLM agent frameworks and benchmarks
DART: Distributional Adversarial Recurrent Training for Algorithm LearningLLM reasoning → Test-time compute and efficient thinking
Accelerating Diffusion Transformers with Gaussian Process Rectified Feature CacheDiffusion models → Diffusion foundations and sampling
VibeCheck: Assessing the Quality of LLM-Generated Unit Tests: A Multi-agent Empirical Study across Heterogeneous RepositoriesLLM agents and tool use → Software engineering coding agents
Beyond Cross-Lingual Transfer: Benchmarking Propagation Boundaries in Multilingual LLM UnlearningLLM safety and security → Machine unlearning
Intent Drift at SME Scale: Deployment Practice, Not Model Capability, Determines Agentic ComplianceLLM agents and tool use → Agent threats, guardrails and governance
Test-Time Weak-to-Strong Alignment: Transferring Implicit Rewards from Weak to Strong Flow ModelsLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
On-the-go Forgetting without Explicit Unlearning via ERASELLM safety and security → Machine unlearning
SurveyAgent-HKA: A multi-agent framework for scientific survey generation with LLMs and human knowledge augmentationLLM agents and tool use → Web, search, and research agents
Rethinking the Evaluation of Efficiency Methods for Multi-Agent SystemsLLM reasoning → Reasoning for agents, planning and code +1
Solving versus Verifying: Catching Contradictions in Tax Reasoning SystemsLLM reasoning → Reasoning for agents, planning and code
GIF: Agentic Generation of Interactive and Functional Object Compositions for Robot LearningLLM agents and tool use → Multimodal and embodied agents
STAR-Pro: Stage-Wise Token Adaptive Reduction with Progressive Refinement for Efficient Large Vision-Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Structurally Close, Temporally Distant: Measuring Security Exposure in Long-Horizon LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
UniRRM: Unified Reasoning Reward Models Across Languages and Evaluation ParadigmsLLM reasoning → Reinforcement learning for reasoning +1
From Narrative to Auditable Forecasts: A Structured Scaffold for Agentic ForecastingLLM agents and tool use → Financial trading and forecasting agents
EvoSafeHarness: Evolving Model- and Domain-Specific Harnesses for Securing AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
GenPuzzle: Benchmarking Visual Reasoning in Image Generation ModelsLLM reasoning → Multimodal reasoning
From Review to Authorization: Key-Isolated Threshold Signing for LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
AlignDiff: Exploiting Model-Intrinsic Information for Better Preference Data SelectionLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Multimodal Resource-Exhaustion Attacks on Vision-Language Models via Joint Pixel-Prompt OptimizationMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Grounded and Faithful P&ID Reasoning: Constraining Vision-Language Models with Recovered Evidence GraphsLLM reasoning → Multimodal reasoning +1
SAFEGuard: Detect Optimization-Based Jailbreak Attacks Through Harmful Semantic Analysis and Fluency MeasurementLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
AgentBrew: Offline Tool-Use Agent Learning from Raw Real-World TrajectoriesLLM agents and tool use → Tool learning and function calling
Learning Counterfactual World Models for Embodied Reasoning under Partial ObservabilityWorld models → Latent predictive world models
CR-VLA-Force: Learning Control-aware Compliance VLA Model for Robust Contact-rich Robotic ManipulationVision-language-action models → Tactile, force-aware and humanoid VLA +1
Beyond Top-k Skill Retrieval: Diversity-Aware Skill Routing for LLM AgentsLLM agents and tool use → Agent skill learning and evolution
Generalizing HVAC Control With Domain Randomized Reinforcement LearningDeep reinforcement learning methods → RL for energy, networks and operations
CONDUIT: A Unified Residual-Stream Restoration Framework for KV Cache Reuse in Vision-Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Agentic BAIM-LLM Evaluation (ABLE): Benchmarking LLM Use of Protein Design ToolsLLM agents and tool use → Medical and scientific domain agents
Physical policy gradient theorem for in situ stochastic-adjoint trainingDeep reinforcement learning methods → RL theory and sample complexity
Spillover-Aware Multi-Value Steering for Pluralistic LLM AlignmentLLM alignment / RLHF / preference optimisation → LLM preference optimization methods +1
Safety Monitors Mostly Catch What the Model Already RefusesLLM safety and security → Safety alignment, refusal and guardrails
Bait-and-Recover: Poisoning Internal Refusal Signals to Defend LLMs against White-Box Editing JailbreaksLLM safety and security → Safety alignment, refusal and guardrails +1
Closed-Loop Evaluation of Bird's-Eye-View Maps from Cross-View Transformers as Inputs to Behavior-Cloning PoliciesOffline RL and imitation learning from data → Imitation and offline RL for driving
Distilling Vision-Language Models for On-Device Fire UnderstandingMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Travel Time Prediction in Supply Chain Management Using Machine Learning
A Tutorial on Prompt Engineering: From Messy Thoughts to AI Workflows
Checkpoints Are Not Enough: Trust Calibration in CoSLR, a Human-AI System for Systematic Literature Reviews
CHART: A Harness-Rotation Curriculum for Harness-Robust Search Agents
The Corroboration Illusion: When More News Makes LLM Forecasts Less True
Universal Observatory Graphs for Distributed Sky Coverage and Artificial Intelligence Based Interplanetary Routing
Replay-Gated Neural Execution: Decoupling Persistent Behavioral Specifications from Neural Realizations in Frozen Language Models
An Evolutionary Computation Framework for Multi-Agent Q-Learning with Mean-Field Environmental Feedback
Preserving Subject-Clarity in Image Outpainting with Multiscale Wavelet Supervision
A machine-checked proof of the Dong-Yang classification of optimal (n,4) binary codes for BSCs
Hierarchical Fourier Approximation for Variational Quantum Distribution Learning
Representation Learning for Sample-Efficient CATE Estimation by Leveraging Multiple Outcomes
Robust conditional dimension reduction for dissimilarity data
IM-ENGINE: Image Editing for Embodied Data Generation
A Comparative Study of GAN-Based Deep Learning Models for Pneumonia Detection in Chest X-Rays
SIDE: Sensor Impersonation Detection at the Edge via Sequence Prediction
Correction as Annotation: Bootstrapping a Dependency Parser for Documentary Medieval Latin
Beyond the Flag: Clinical Framing Closes the Moderation Gap in Suicide Risk Measurement
Adapting Vision Foundation Models to Acoustics for Pose-Free 3D Sonar Reconstruction
SeaCausal-FL: Federated Fuzzy Causal Learning for Maritime IoT Fault Diagnosis and Counterfactual Reasoning
One Model, Two Worlds: Bidirectional Sonar-Optical Translation
Task-Oriented Semantic Communication for Hazard Warning and Remote Operation in Connected Vehicle Platoons
Spatial Attention Supervision for Defect Localization: Exploiting Ground-Truth Masks as Training Signal in Diffusion-Augmented Defect Detection
SWE-Test: Benchmarking LLM Vulnerability Discovery via Input Prediction

All 213