Papers

New on arXiv

Wed 12 Aug 2026

423 papers

Reducing Hallucinations in Large Language Models Through Integrated Self-Verification and Retrieval-Augmented GenerationRetrieval-augmented generation → RAG pipeline design and evaluation
Same Chart, Different Story: Bias in Vision-Language Chart InterpretationMultimodal LLMs / vision-language models → Social bias and fairness in VLMs
Unread or Unenforced? Separating Representation from Enforcement Failure in Content GuardsLLM safety and security → Safety alignment, refusal and guardrails
Refusing Everything Looks Safe: Restoring the Benign Arm to Encoded-Prompt EvaluationLLM safety and security → Safety alignment, refusal and guardrails
The Uncontrolled Variable: Vision-Language Model Refusal Responds to Image Presence in Ways Risk Cannot ExplainMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
AdaMem: Adaptive Memory Token Allocation for Soft Compression in Retrieval-Augmented GenerationRetrieval-augmented generation → Generation and context integration
TreeSpark: Calibrated, Load-Adaptive Draft Trees for Semi-Autoregressive Speculative DecodingEfficient LLM inference and serving → Speculative decoding
Diagnosing Faults in Reinforcement Learning Simulators and World Models with Canonical Polynomial InvariantsWorld models → Latent dynamics models and model-based RL +1
ARC-Bench: Closed-Loop Replanning Masks Broken Action Ranking in Frozen JEPA World ModelsWorld models → Latent predictive world models
From Extraction to Governed Memory: Multi-Agent Knowledge Graph Construction with Domain-Expert ReviewLLM agents and tool use → Web, search, and research agents
Mitigating Bias in Large Vision-Language Models via Counterfactual Ensemble DecodingMultimodal LLMs / vision-language models → Social bias and fairness in VLMs
RiskWorld: Object-Centric Latent World Modeling for Autonomous Driving Risk IdentificationWorld models → World models for autonomous driving +1
Tail-Aware Top-k On-Policy DistillationKnowledge distillation → On-policy and self-distillation for LLMs
AeroGround: A Comprehensive Benchmark for Aerial-Ground Collaborative ReasoningLLM reasoning → Spatial and embodied reasoning +1
VideoGAIA: A Benchmark for General AI Assistants on Agentic Video UnderstandingMultimodal LLMs / vision-language models → Video understanding with MLLMs +1
VoiceDesigner: Text-to-Voice Generation and Editing via Unified Diffusion Modeling and Data AugmentationDiffusion models → Audio, speech and music
SteerBench-Work: A Benchmark for Agent Steering at Action BoundariesLLM agents and tool use → General LLM agent frameworks and benchmarks
Excess Separability: Nuisance-Controlled Residual-Stream Probing for Benchmark Contamination DetectionLLM safety and security → Backdoor and data poisoning attacks
Jagged Judges: Epistemic Stability Under Perturbation, Pressure, and PersistenceLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
EgoCITE: Context-Augmented Indexing and Time-Aware Retrieval for Long-Horizon Egocentric MemoryLLM agents and tool use → Multimodal and embodied agents
LLMs Are Not Good Strategists, Yet Memory-Enhanced Agency Boosts ReasoningLLM reasoning → Reasoning for agents, planning and code
When Explanations Betray Backdoors: Black-Box Auditing for Language Model ClassifiersLLM safety and security → Backdoor and data poisoning attacks
PseudoMapLabeler: Confidence-Aware Pseudo-Label Generation for Semi-Supervised Online MappingKnowledge distillation → Distillation for medical imaging
Reasoning Jury: Multi-Model Consensus for Evaluating Reasoning TracesLLM reasoning → Test-time compute and efficient thinking
Scaling Automatic Research Agents via World ModelsWorld models → World models for language agents +1
Do LLMs Beat Nash? Testing Decentralized Coordination in Self-Play Multi-Agent GamesLLM agents and tool use → Social simulation and multi-agent debate
MASCOT: Model-Aware Submodular Coverage for Composite-Attribute Text-to-Image RetrievalMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Can Vision-Language Models Assess Proxemic Risk from Egocentric Robot Images?Multimodal LLMs / vision-language models → Embodied and 3D vision-language models
DIVE: Unlocking Self-Improvement in Frozen Language Models Through Diversity-Driven Skill EvolutionSynthetic data and self-improvement for LLMs → Self-rewarding alignment and self-training
Governed Persistent Memory: Source-Bound State Semantics and Fail-Closed Release for Long-Horizon AgentsLLM agents and tool use → Agent memory systems
SSPO: Structure-Aware Similarity-Weighted Preference Optimization for Neural Combinatorial OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Specification-first convergence with an AI coding agent: a case study of dismantling a core architectural invariant across 189 files in a 717k-line codebase with no test oracle and no human code reviewLLM agents and tool use → Software engineering coding agents
SynWeaver: Website-Prior Task and Trajectory Co-Synthesis for Web AgentsLLM agents and tool use → Web, search, and research agents
LoKiFormer: Locality-aware Attention with Decoupled Knowledge Memory for Efficient Large Language Model PretrainingMixture-of-experts → MoE large language models
AVA-Encoder: Towards Agent-Native Video Representation LearningLLM agents and tool use → Multimodal and embodied agents
The Role Specialization Model (RSM): Coordinating LLM-Based Tools in Agentic Software Development - An Exploratory Case StudyLLM agents and tool use → Software engineering coding agents
AI4AI at Test-Time: Strong-to-Weak Capability Transfer via HarnessesKnowledge distillation → On-policy and self-distillation for LLMs
Redistribution-based Cost Inference Improves Sparse Safe Offline RLOffline RL and imitation learning from data → Safe offline RL
Beyond Trial-and-Error: Agentic Optimization for Image-to-Video AdherenceLLM agents and tool use → Multimodal and embodied agents
VAKRA: Evaluating Multi-Hop Reasoning Across APIs and Retrieval Under Tool-Use PoliciesLLM reasoning → Knowledge- and tool-augmented reasoning
XYZFlow:Scaling Multi dimensional Shortcut Flows for Efficient Generative ModelingDiffusion models → Flow matching, time series and science
Convergent Detour Hijacking: Task-Preserving Resource Amplification in Skill-Based LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
Diagram-MMU: A Multi-Modal Benchmark for Scientific DiagramsMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
SelectLight: Learning to Select Signal Plans Generated by Distributed Model Predictive Control for Urban Traffic NetworksDeep reinforcement learning methods → RL for autonomous driving and traffic
One Frozen Simulator Is Not Enough: Simulator Collapse in Multi-Agent RLLLM agents and tool use → Reinforcement learning for LLM agents
Regime-Gated Residual Mixture-of-Experts for Cross-Sectional Volatility ForecastingMixture-of-experts → MoE for time series and graphs
An Agentic Workflow for Legacy HPC Modernization: Converting the Two-Electron-Integral Core of GAMESSLLM agents and tool use → Medical and scientific domain agents
SCOUT: Unlocking Enhanced Spatial Reasoning via Structured Chain-of-Thought and Multi-Objective Process RewardLLM reasoning → Spatial and embodied reasoning +1
"Pharos Night: Crown Pursuit": An AI-Native Deck-Building and Tactical Arena Game Design Based on Multi-Agent SystemsLLM agents and tool use → Reinforcement learning for LLM agents
Generation as Auxiliary Supervision: Enhancing Visual Understanding at Zero Inference Overhead via Decoupled Embedding PredictionMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
GeoFlow: Efficient Driving Video Generation via Geometry-Aligned PriorsDiffusion models → Diffusion and flow policies for robotics
Learning-Based Behavior Planning for Automated Driving: Real-World Integration and DeploymentEnd-to-end autonomous driving → Imitation-learning end-to-end planners
HSTGFormer: Hyper Spatial-Temporal Graph Transformer for 3D Human Pose EstimationLLM reasoning → Spatial and embodied reasoning
Rethinking Agent Security as a Networking ProblemLLM agents and tool use → Agent threats, guardrails and governance
Context Blindness in DPO: Mitigating Object Hallucination in MLLMs via Context-Calibrated Preference OptimizationMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs +1
Understanding Why Foundation Models Work for Diffusion-Generated Image DetectionDiffusion models → Image generation, text-to-image and editing
Who Thinks Best Depends on How Long You Let Them: Budget-Dependent Rankings in LLM EvaluationLLM reasoning → Test-time compute and efficient thinking
MVFM-3DAD: Multi-view Flow Matching for 3D Anomaly Detection via Density Proxy EstimationDiffusion models → Flow matching, time series and science
RoutePack: Expert Placement and Attention-Aware Data Packing for MoE Reinforcement LearningMixture-of-experts → MoE in reinforcement learning and robotics +1
A corpus-specific clinical RAG system matches or outperforms newer frontier LLMs on HealthBenchRetrieval-augmented generation → Medical and biomedical RAG
SAG: SQL-Retrieval Augmented Generation with Query-Time Dynamic HyperedgesLLM reasoning → Knowledge- and tool-augmented reasoning +1
Do LLMs Take Care of Their Own? Similarity Signals Can Induce CooperationLLM agents and tool use → Social simulation and multi-agent debate
Ready Cohorts: Bounding GPU Opportunity and Avoiding Host Round Trips in LLM-Agent ControlLLM agents and tool use → Software engineering coding agents
QV-PIC: Query-Aware Visual Position-Independent Caching for Efficient RAG ServingRetrieval-augmented generation → Generation and context integration
No One to Blame: A Framework of Constitutive AI UnaccountabilityLLM agents and tool use → Agent threats, guardrails and governance
RT-SEMamba: Real-Time Speech Enhancement Mamba via Progressive Knowledge DistillationState-space sequence models (S4, Mamba) → Speech and audio SSMs +1
Look What the Probes Dragged In! Real-World Chest X-ray Shortcuts in MedCLIPMultimodal LLMs / vision-language models → Medical vision-language models
Faithful, Sufficient and Understandable: Rethinking Graph Counterfactual Explanations via Discrete Diffusion InversionDiffusion models → Flow matching, time series and science
Rethinking Language Model-Based Generative Speech Enhancement in the Latent Space of a Neural Audio CodecDiffusion models → Audio, speech and music
Better Slots, Better Worlds: Representation Quality & Robustness in Object-Centric World ModelsWorld models → Latent predictive world models
Learning Loco-Manipulation From SMPC Demonstrations With Sparse Offline-to-Online RLOffline RL and imitation learning from data → Real-world robot offline RL
LoSA: Near-Lossless Sparse Attention for Training-Free Video Diffusion AccelerationDiffusion models → Video generation
Claim-Level Reliability Assessment for Efficient Test-Time ReasoningLLM reasoning → Test-time compute and efficient thinking
Learning from Unreachable Rewards: Hint-Conditioned Reinforcement Learning for Generative RecommendationDeep reinforcement learning methods → RL training of language and agent models
Retry, Switch, or Abstain? Learning Strategy-Aware Tool-Use Policies via Controlled Error InjectionLLM agents and tool use → Tool learning and function calling
LoongReflect: Boosting Long-Horizon Reflection in Search Agents via Global Perspective DistillationLLM agents and tool use → Agent memory systems
Developing LLM-based Multi-Agent Systems in Software Engineering: A Mixed-Method Experience ReportLLM agents and tool use → Software engineering coding agents
ExRole: From Team Trajectories to Executable Roles in Multi-Agent Language ModelsLLM agents and tool use → General LLM agent frameworks and benchmarks
Evaluating and Calibrating Diffusion Model-derived Uncertainty for Quantitative MRI MappingDiffusion models → Medical imaging and segmentation
LODESTAR: Robust Entropy-Based Answer Selection in Retrieval-Augmented Generation for Question Answering -- Directing Frozen-LLM Entropy with a Reinforcement-Learned Prompt Polarizer under Misleading PassagesRetrieval-augmented generation → RAG robustness and adaptive retrieval
HarmoniDPO: Video-guided Audio Generation via Preference-Optimized DiffusionLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
A Model-Internal Protocol for Assessing Multimodal Models as Integrated SystemsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Policy-as-logic for robust reasoning over rulesLLM reasoning → Reasoning benchmarks and evaluation
On-Policy Self-Distillation for Multi-Dialect ASR: Mastering Dialects, Retaining MandarinKnowledge distillation → Speech and audio distillation
Agent Skills Can Be Harmful: An Empirical Study of Skill-Induced Failures in LLM AgentsLLM agents and tool use → Agent skill learning and evolution
Total Recall at What Cost? Benchmarking the Serving Cost of Agentic Memory SystemsLLM agents and tool use → Agent memory systems
ToolHazard: Scaling Adversarial Environments for Security Evaluation and Alignment of LLM-based AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
LookBack: Where and How to Score LVLM Responses via Visual Reference UsageMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
GeoBridge: Decoupled Semantic Conditioning for Generative Image GeolocalizationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Towards Understanding On-Policy Distillation through the Lens of Test-Time ScalingLLM reasoning → Prompting and chain-of-thought reasoning +1
How China-Origin Vision-Language Models Move from Refusal to Reframing in State AlignmentMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
CoDiR: Confidence-Guided Diffusion Refinement for Semi-Supervised Histopathology SegmentationDiffusion models → Medical imaging and segmentation
MiDashengLM-Gen: Unified Audio Scene Generation via LLM-Driven Autoregressive Flow MatchingDiffusion models → Audio, speech and music
Anti-Shortcut Distillation via Temporal Negative Knowledge TransferKnowledge distillation → Knowledge distillation for LLMs
GRPO for Financial Advice Generation: Outperforming Commercial LLMs under CATE EvaluationLLM reasoning → Reasoning for agents, planning and code
TradingMoE: Routing the Right Experts in Evolving MarketsMixture-of-experts → MoE for time series and graphs
VOLA: Improving Open-World Driving by VLM-Based Semantic Attribute PredictionEnd-to-end autonomous driving → VLMs for autonomous driving
The Sleeping Agent: What Gist-Based Context Compression Loses and WhyLLM agents and tool use → Agent memory systems +1
Policy-Induced Hand Priors in Humanoid Dual-Arm Manipulation: Diagnosing and Mitigating Initial-Pose DependenceVision-language-action models → Human-video and latent-action pretraining
AWARe: Mitigating Catastrophic Forgetting via Activation-Weighted Adaptive REtentionLLM reasoning → Multimodal reasoning +1

All 423