Papers

New on arXiv

Sun 2 Aug 2026

299 papers

RADAR: Readiness for AI Discovery and Agentic ReachLLM agents and tool use → Agent threats, guardrails and governance
M3-Former: Multimodal Transformer with Mixture-of-Experts for Long-Term Vessel Trajectory PredictionMixture-of-experts → MoE in reinforcement learning and robotics
Spyre-Accelerated Retrieval-Augmented Generation on IBM LinuxONE: A Cloud-Native Architecture for Secure, High-Throughput Enterprise AI InferenceRetrieval-augmented generation → RAG pipeline design and evaluation
CARA: Cognitive Adaptive Recommendation AgentLLM agents and tool use → Agentic recommender systems
BRA-Audit: Budgeted Runtime Auditing for LLM Multi-Agent Systems via Cumulative-Exposure Audit-Point PlacementLLM agents and tool use → Agent threats, guardrails and governance
Position: AI Agents in Scientific Teams Should Be Studied as Human-Agent SystemsLLM agents and tool use → Web, search, and research agents
Unified Multi-Dimensional Benchmark for Complex Graph Reasoning in Large Language ModelsLLM reasoning → Knowledge- and tool-augmented reasoning
Towards Sustainable Learning in Online Education: A Reinforcement Learning ApproachDeep reinforcement learning methods → RL training of language and agent models
Mechanistic Interpretability-Guided Selective Fine-Tuning of Vision-Language Models for Centimeter-Level Flood Depth EstimationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
AeroDPO: Unleashing Lightweight UAV Navigation with High-Fidelity Perception and Automated Preference OptimizationLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
You Don't Need To Stay in The Loop: An Agentic Robotics Loop for Robot-Policy ImprovementLLM agents and tool use → Multimodal and embodied agents
EDATracer: An Agentic Framework for Large-Scale EDA Artifact AnalysisLLM agents and tool use → Software engineering coding agents
DenialRAG: Single-Document RAG Poisoning via Embedded Parametric DenialRetrieval-augmented generation → Privacy and security of RAG
Moving the Safety Barrier: Dynamic Routing Adaptive Alignment Against White-Box AttacksLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Permission Denied: Policy-Graded Evaluation of Coding Agents in Hardened EnvironmentsLLM agents and tool use → Software engineering coding agents
V-Mem: Modality-Routed Retrieval for Long-Term Multimodal Agentic MemoryLLM agents and tool use → Agent memory systems
STAR-VLM: Spatiotemporal Grounding Vision-Language Models for Motion and Velocity Estimation via Automotive Radar SupervisionLLM reasoning → Chain-of-thought reasoning for driving +2
Recursive Vision Language Models for General Symbolic ReasoningLLM reasoning → Test-time compute and efficient thinking +1
Question Begets Question: Self-Evolving Curriculum for Reinforcement Fine-Tuning on Competition MathematicsLLM reasoning → Reinforcement learning for reasoning +1
Deep Agentic Search for Repository-Level Code Question Answering: An Empirical StudyLLM agents and tool use → Software engineering coding agents
Towards Compact Unified Multimodal Tracking: Synergizing Knowledge Distillation with Structural PruningKnowledge distillation → Detection and dense prediction distillation
Slot2Text: Object-Centric Visual Tokenization for Efficient and Spatially Traceable Surgical MLLMsMultimodal LLMs / vision-language models → Medical vision-language models
Retrieval Augmented Biomedical Question Answering with Weak Question Recovery and Neural Reranking for BioASQ Task 14bRetrieval-augmented generation → Medical and biomedical RAG
Computing with Agentic OraclesLLM agents and tool use → Agent memory systems
Where Reasoning Diverges: Localized Multi-Agent Debate for Multi-Hop Question AnsweringLLM agents and tool use → Social simulation and multi-agent debate
PALMs: Using Multi Construct-Grounded Rationales for Modeling Population Preferences in LLMsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Real-Time Hybrid Retrieval in Hyperbolic Space for Retrieval-Augmented Generation on Edge DevicesRetrieval-augmented generation → Generation and context integration
Beyond Routing Saturation: A Long-Horizon Class-Incremental Perspective on Expert Routing in Multimodal Continual Instruction TuningMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP +1
When Replanning Becomes the Bottleneck: Budgeted Replanning for Embodied AgentsLLM agents and tool use → Tool learning and function calling
Training Small LLMs as Spatial Multi-Agent PoliciesLLM agents and tool use → Reinforcement learning for LLM agents
QR-Erase: Efficient Subspace-Based Machine Unlearning with Layer LocalizationLLM safety and security → Machine unlearning
No Single Neuron of Failure: Distributed Safety Alignment Against White-Box AttacksLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Demystifying When and Why VLAs Fail in Contact-Rich Tasks and How to Fix ThemVision-language-action models → Tactile, force-aware and humanoid VLA
SG-WAM: Self-Guided World Modeling in Geometry-Aware Policy SpaceWorld models → Embodied and robot world models
Why Formal Monitors Fail: Attack Distribution Entropy as a Coverage Bound for LTL-Based LLM Agent SafetyLLM agents and tool use → Agent threats, guardrails and governance
DreamTrajectory: Trajectory-Guided Action Generation with World Model Alignment for Mobile ManipulationWorld models → Embodied and robot world models
The Boy Who Cried Wolf: Adversarial Misclassification of Safe Inputs as Unsafe in Multimodal GuardrailsLLM safety and security → Vision-language and text-to-image safety
CRAFTS: Collaborative Role-Adaptive Fine-Tuning of LLM Agents for Chemical Process SimulationLLM agents and tool use → Medical and scientific domain agents
Asking Questions the Right Way: A Multi-Agent Conversational System for Prompt Formulation in Complex Task ResolutionLLM agents and tool use → General LLM agent frameworks and benchmarks
EviSD: Evidence-Conditioned Self-Distillation for Search-Augmented AgentsKnowledge distillation → On-policy and self-distillation for LLMs
HopRefusalBench: Diagnosing Refusal Failures in Search-Augmented Agents for Multi-Hop ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning +1
PixVL: Self-Supervised Training of Pixel-Level MLLMs via a Unified Mask--Text Consistency CycleMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Prompt-Induced Waste in Coding Agents: Reasoning, Effort, Harness Design, and End-to-End CostLLM agents and tool use → Software engineering coding agents
Hybrid AI for Explainable and Accurate Conversational Agents in eGovernmentLLM agents and tool use → General LLM agent frameworks and benchmarks
LongChart VQA: A Comprehensive Benchmark for MLLMs with Complex Multi-Chart ReasoningLLM reasoning → Multimodal reasoning +1
Cognitive Demand Steering for Adaptive Meta-Reasoning in Large Language ModelsLLM reasoning → Test-time compute and efficient thinking
Remember-R1: Mitigating Long-Context Visual Forgetting through Reinforcement LearningLLM reasoning → Multimodal reasoning +2
AlphaG-OPD: Reliability-Gated Sibling Counterfactuals for On-Policy Distillation in Symbolic Alpha Factor DiscoveryKnowledge distillation → On-policy and self-distillation for LLMs
UDT: Reconciling U-Nets and Diffusion Transformers with Data-Adaptive Token ReductionDiffusion models → Diffusion foundations and sampling +1
CrossLex: A Source-Grounded Benchmark for Cross-Jurisdictional Legal Reasoning in Large Language ModelsLLM reasoning → Reasoning for agents, planning and code
Stop When Memory Suffices: Evidence-Conditioned Progressive Execution for LLM AgentsLLM agents and tool use → Agent memory systems
Training nGPTMixture-of-experts → Expert routing strategies
Rethinking Video Token Compression with a Global Codebook: Learning Once, Compressing EverywhereLLM reasoning → Multimodal reasoning +1
ACE-GraphRAG: Agentic Context Engineering for Hierarchical GraphRAGRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG +1
Hermite Curves as Trajectory Priors for Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation
Distill What the Student Can See: Fisher-Projected On-Policy Distillation for Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
A Benchmark Dataset for MLLM-Generated Image Detection: GPT Image2 & Nano Banana2Multimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
RestoreKV: Recovering Full-Cache Behavior Under Aggressive Query-Agnostic KV Cache EvictionEfficient LLM inference and serving → KV cache optimization
Evaluating VLMs on Multimodal Aristotelian Persuasion TasksMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Learning What to Remember and What to Internalize in LLM Self-Evolution via Adaptive Memory-Parameter CoordinationLLM agents and tool use → General LLM agent frameworks and benchmarks
Perspectives on Tsallis Statistics for Artificial IntelligenceDeep reinforcement learning methods → RL theory and sample complexity
EndoWAM: A Grounded World-Action Model for Generalizable Endoscopic NavigationWorld models → Medical and surgical world models
VaRS-Doc: Interpretation-Aware Variant Representations via Latent Self-Probing for Visual Document RetrievalRetrieval-augmented generation → Multimodal RAG
Climate-Dyna Deep Hedging for XVAs: Model-Based Reinforcement Learning, Residual Climate HVA, and Hedge-Instrument DiscoveryDeep reinforcement learning methods → Adversarial robustness and RL in finance
It's the Decoding Format, Not the Perturbation: Auditing Consistency-Based Selection for Vision-Language Test-Time ScalingLLM reasoning → Multimodal reasoning +1
ReBRAC-v2: The Return of the KingOffline RL and imitation learning from data → Offline RL algorithms +1
PRISM: Privileged Probabilistic Latent Supervision for End-to-End Autonomous Driving Motion PlanningEnd-to-end autonomous driving → Reasoning VLA and VLM driving policies
A Unified Benchmark for Privacy-preserving Vector SearchRetrieval-augmented generation → Privacy and security of RAG
MADE: Belief-Driven Dual-Agent Coordination for Autonomous Model DeploymentLLM agents and tool use → General LLM agent frameworks and benchmarks
3DZip: Spatial-Aware Feature Diversity-Guided Token Compression for 3D Question AnsweringLLM reasoning → Spatial and embodied reasoning +2
The Graph Language: How Knowledge Graphs Speak to Large Language ModelsLLM reasoning → Knowledge- and tool-augmented reasoning
Think in Sets for Streaming Video Token CompressionLLM reasoning → Multimodal reasoning +1
PATH-Bench: Path-Dependent Evaluation of Lifelong AgentsLLM agents and tool use → Agent memory systems
Policy Optimality Measurement for Multi-Vehicle Decision-Making: From Extrinsic Indicators to Intrinsic QualityDeep reinforcement learning methods → RL for autonomous driving and traffic
MiniWorld: Democratizing the Training of Video World Models from ScratchWorld models → Video-generation world models and simulators +1
SoK: Intent-Oriented Systematization of Multi-Turn LLM JailbreaksLLM safety and security → Jailbreaking
Lethe: How Hard Is It to Forget? A Benchmark for Federated Unlearning in Medical ImagingLLM safety and security → Machine unlearning
Separate-and-Detect: Unified Drum Transcription and Stem Generation via Latent DiffusionDiffusion models → Audio, speech and music
When Collaboration Becomes a Trigger: Collective Evidence-Threshold Backdoors in Multi-Agent SystemsLLM agents and tool use → Agent threats, guardrails and governance +1
Attend to Your Own Thoughts: Breaking the Barrier for Post-Training Quantization of Reasoning LLMs through the Lens of 1.58-Bit QuantizationLLM reasoning → Test-time compute and efficient thinking
One Query, Many Scales: Sparse Mixture-of-Experts for Efficient Hierarchical Cross-View Geo-LocalizationMixture-of-experts → Vision mixture-of-experts
Control Under Compression: Reliability Frontiers for Tool-Using AgentsLLM agents and tool use → Tool learning and function calling
Credit the Right Box: Marginal Contribution Assignment for Structured Visual PerceptionLLM reasoning → Multimodal reasoning +1
Don't Offer What Can't Be Done: Deterministic Executability Gating for LLM Skill Selection at ScaleLLM agents and tool use → Agent skill learning and evolution
FactorJEPA: Factorizing Monolithic Futures into Layout-Agent-Interaction Channels for Crowded and Chaotic Global South Urban WorldsWorld models → Video-generation world models and simulators
Decoy Images Amplify Caption-Mediated Defenses Against Encoded JailbreaksLLM safety and security → Vision-language and text-to-image safety
What Could the Agent See at 19:05? Generating Temporal Enterprise Scenarios from Real Research and Replaying Them to Evaluate AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
WAM-Diff2: Hierarchical AR-to-Diffusion Distillation for Highly Efficient Autonomous Driving VLAVision-language-action models → VLA for autonomous driving +2
Why LLMs Give In: Conversational Factors and Reasoning Behind Medical SycophancyLLM reasoning → Clinical and biomedical applications
Cloud-ScPO: Hidden-State Geometry for Semi-Supervised Preference Optimization in LLM ReasoningLLM reasoning → Mathematical reasoning +1
RL Bootstrapping of OpenVLA-OFT for a Novel Robot EmbodimentVision-language-action models → Reinforcement learning and post-training
Toward Fine-Grained Forgetting:Attribute Unlearning for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Who Belongs in the Eval Set? A Capability-Taxonomy-Driven Pipeline for Curating Regression Eval Sets in Agent-Extensibility PlatformsLLM agents and tool use → General LLM agent frameworks and benchmarks
From AI Technical Debt to Agentic Technical Debt: A Systematic Mapping of Root Causes and Manifestations in Agentic AI SystemsLLM agents and tool use → Software engineering coding agents
Auditing Discovery Claims: A Two-Sided Criterion for Agentic Science, with the Negative Side DecidableLLM agents and tool use → Web, search, and research agents
One-Sided Quantile Coupling for Flow MatchingDiffusion models → Flow matching, time series and science
MonitorVLM-v2: A Deployed Vision-Language Framework for Real-Time Safety Violation DetectionLLM reasoning → Reasoning for agents, planning and code
Search-GRT: Guided Retrieval Training of Search Agents to Optimize for Complex Question AnsweringLLM agents and tool use → Web, search, and research agents
Mind the Gap: Zero-Query Jailbreaks via Filter-Generator Discrepancy in Text-to-Image SystemsLLM safety and security → Vision-language and text-to-image safety
PROGRESS: Coverage-guided RL to Train Search-augmented LLM AgentLLM reasoning → Knowledge- and tool-augmented reasoning +1

All 299