Papers

New on arXiv

Sat 30 May 2026

281 papers

CHaystack: Benchmarking Chinese Document Retrieval and VQARetrieval-augmented generation → Multimodal RAG
PersonaTrail: Benchmarking Personalized Web Agents through Browsing TrailsLLM agents and tool use → Agent memory systems
Harmonic: Hierarchical State Space Models for Efficient Long-Context Language ModelingState-space sequence models (S4, Mamba) → Core SSM architectures for language
Post-Training Recipe, More Than Model Family, Shapes Multi-Agent LLM Conversational BehaviorLLM agents and tool use → Social simulation and multi-agent debate
Large Language Models Should Learn Personalized Rather Than Aggregated Human PreferencesLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Query Lens: Interpreting Sparse Key-Value Features with Indirect EffectsMechanistic interpretability → Sparse autoencoder training and evaluation
MultiTurnPSB: Evaluating Multi-Turn Jailbreak Attacks an dClassifier-Based Defenses for Medical AI SafetyLLM safety and security → Jailbreaking
Hallucination Is Linearly Decodable from Mid-Layer Hidden States in Quantized LLMsMechanistic interpretability → Hallucination detection from model internals
Generative Multi-Robot Motion Planning via Diffusion Modeling with Multi-Agent Reinforcement Learning GuidanceDiffusion models → Diffusion and flow policies for robotics
CV-Arena: An Open Benchmark for Instructional Computer Vision Problem Solving with Human-AI Collaborative PreferencesLLM agents and tool use → Multimodal and embodied agents
A Circuit, Not The Circuit: Non-Unique Causal Localisation of the Mamba-2 State SinkState-space sequence models (S4, Mamba) → Core SSM architectures for language +1
Single-Channel Tissue Segmentation via Cross-Modal Distillation from Foundation ModelsKnowledge distillation → Distillation for medical imaging
Benchmarking Security Risk Detection and Verification in Open Agentic Skill EcosystemsLLM agents and tool use → Agent threats, guardrails and governance
A Machine-to-Machine Knowledge-Guided LLM Agent for Generalizable Radiotherapy Treatment PlanningLLM agents and tool use → Medical and scientific domain agents
Adversarial Feeds Steer LLM Agent Decisions Against Their DefaultsLLM agents and tool use → Social simulation and multi-agent debate
Reason, Retrieve, Re-rank: A Zero-Shot Reasoning-Aware Framework for Composed Video RetrievalLLM reasoning → Multimodal reasoning
MLLM-Microscope: Unlocking Hidden Structure Within Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Ryze: Evidence-Enriched Data Synthesis from Biomedical PapersSynthetic data and self-improvement for LLMs → Medical synthetic data and agents
Chunking Methods on Retrieval-Augmented Generation - Effectiveness Evaluation Against Computational Cost and LimitationsRetrieval-augmented generation → Multimodal RAG
Task diversity produces systematic transfer but inhibits continual reinforcement learningDeep reinforcement learning methods → Deep RL algorithms and representations
Benchmarks for Vision-Language Models in Urban Perception Should Be Reliability-Aware and NegotiatedMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Enhancing LLM Metacognition via Cognitive Pairwise TrainingLLM reasoning → Reinforcement learning for reasoning
MoEIoU: Rethinking Bounding-Box Regression as a Mixture of ExpertsMixture-of-experts → Vision mixture-of-experts
Certificate-Guided Evaluation of Reinforcement Learning GeneralizationDeep reinforcement learning methods → Deep RL algorithms and representations
Decoupled Behavioral Cloning for Scalable Inductive Generalization in RL from SpecificationsOffline RL and imitation learning from data → Imitation learning methods +1
Coarse-to-Fine Compositional Diffusion for Long-Horizon PlanningDiffusion models → Diffusion and flow policies for robotics
Momento: Evaluating Persistent Memory and Reasoning with Multi-Session Agentic ConversationsLLM agents and tool use → Agent memory systems
The Right Inference Strategy Is All You Need: Nearly Training-Free Domain-Wise Inference for EgoCross ChallengeMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
RoboStressBench: Benchmarking VLM Robustness to Physical Visual Stress in Embodied ScenesMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
SkillPager: Query-Adaptive Intra-Skill Navigation via Semantic Node RetrievalLLM agents and tool use → Agent skill learning and evolution
Not All Flips Are Conformity: Decomposing Stance Convergence in Multi-Agent LLM DebateLLM reasoning → Reasoning for agents, planning and code +1
Cross-Generational Transfer of Adversarial Attacks Reveals Non-Monotonic Safety Alignment in LLMsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Generative Diffusion Priors for 3D Mapping of the Dark UniverseDiffusion models → Inverse problems and image restoration
Quality-Diversity Evolution for Discovering Diverse Vulnerabilities in LLM SafetyLLM safety and security → Prompt injection and agent security
DASH: Dual-Branch Score Distillation for Guidance-Calibrated Compact Diffusion ModelsDiffusion models → Diffusion foundations and sampling
MBench: A Comprehensive Benchmark on Memory Capability for Video World ModelsWorld models → Video-generation world models and simulators
DINO-GFSA: Geo-Localization via Semantic Gated Fusion and Mamba-based Sequential AggregationState-space sequence models (S4, Mamba) → Point clouds, detection and remote sensing
FlowOVD: Learning Generative Latent Flows for Zero-shot Open-vocabulary DetectionMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection
Behavior-Invariant Task Representation Learning with Transformer-based World Models for Offline Meta-Reinforcement LearningWorld models → Latent dynamics models and model-based RL +1
Latent Diffusion Pretraining for Crystal Property PredictionDiffusion models → Flow matching, time series and science
GIRL-DETR: Gradient-Isolated Reinforcement Learning for Video Moment RetrievalDeep reinforcement learning methods → RL training of language and agent models
SafeVLA-Bench: A Benchmark for the Success-Safety Gap in Vision-Language-Action ModelsVision-language-action models → Safety and adversarial attacks
Logit Distillation on Manifolds: Mapping by LearningKnowledge distillation → General knowledge distillation methods
FALAT: Tracing Failures in LLM Agent Trajectories via Dependency-Guided SearchLLM reasoning → Reasoning for agents, planning and code +1
Confidence-Adaptive SwiGLU for Mixture-of-ExpertsMixture-of-experts → Expert routing strategies
CoMIC: Collaborative Memory and Insights Circulation for Long-Horizon LLM Agents in Cloud-Edge SystemsLLM agents and tool use → Agent memory systems
Internalize the Temperature: On-Policy Self-Distillation as Policy Reheater for Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning +2
Score × Decoder: A Unified View of Unsupervised Inference-Time Scaling for Hallucination MitigationLLM reasoning → Hallucination detection and mitigation
ViBE: Co-Optimizing Workload Skew and Hardware Variability for MoE ServingMixture-of-experts → Efficient MoE training and inference systems
EMA: Approximate Nearest Neighbor Search with General Attribute Filtering and Dynamic UpdatesRetrieval-augmented generation → Vector search and ANN indexes
Infeasible optimization problems and the hierarchical augmented Lagrangian method in imitation learningOffline RL and imitation learning from data → Imitation learning methods
Latent Reward Steering: An Adaptive Inference-Time Framework that Implicitly Promotes Cognitive Behaviors in Reasoning LLMsLLM reasoning → Test-time compute and efficient thinking +1
WaveFilter: Enhancing the Long-Context Capability of Diffusion LLMs via Wavelet-Guided KV Cache FilteringEfficient LLM inference and serving → Diffusion LLM and MoE inference
EPIC: Efficient and Parallel Inference under CFG Constraints for Diffusion Language ModelsDiffusion models → Discrete and text diffusion
MOSAIC: Modular Orchestration for Structured Agentic Intelligence and CompositionLLM agents and tool use → Software engineering coding agents
Wavelet-Fusion Diffusion Model for Multimodal Brain MRI Synthesis with Modality and Metadata ConditioningDiffusion models → Medical imaging and segmentation
Dialectics of Alignment: Harnessing Unsafe Knowledge for Dynamic Safety RoutingMixture-of-experts → Expert routing strategies +1
OCC-RAG: Optimal Cognitive Core for Faithful Question AnsweringLLM reasoning → Knowledge- and tool-augmented reasoning
Regularized Offline Policy Optimization with Posterior Hybrid Bayesian BeliefOffline RL and imitation learning from data → Offline RL algorithms +1
The Paradox of Outcome Optimization: A Causal Information-Theoretic Bound on Reasoning Shortcuts in LLMsLLM reasoning → Reinforcement learning for reasoning
T-CLIP: Enabling Thermal Perception for Contrastive Language-Image PretrainingMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Moxia: A Trust-First Neuro-Symbolic Execution Architecture for Self-Explaining Mathematical ReasoningLLM reasoning → Reasoning benchmarks and evaluation
SKIP: Sparse Keyframe Interpolation Paradigm for Efficient Embodied World ModelsWorld models → Embodied and robot world models
FineVerify: Scaling Test-Time Compute with Fine-Grained Self-Verification for Agentic SearchLLM reasoning → Reasoning for agents, planning and code +1
Collaborative Few-Step Distillation and Low-Bit Quantization for Wan2.2 Dual-Expert Video Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Scaling Behavior of Single LLM-Driven Multi-Agent SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
The Invitation Trap: Proactive Availability Backdoor in LLMs via Conversational InductionLLM safety and security → Backdoor and data poisoning attacks
MESA: Improving MoE Safety Alignment via Decentralized ExpertiseLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +2
ForeSci: Evaluating LLM Agents for Forward-Looking AI Research JudgmentLLM agents and tool use → Web, search, and research agents
Hidden Thoughts Are Not Secret: Reasoning Trace Exposure in LLMsLLM reasoning → Prompting and chain-of-thought reasoning
Global-Local Attention Decomposition for Terrain Encoding in Humanoid Perceptive LocomotionDeep reinforcement learning methods → RL for robot locomotion and manipulation
Robust Reasoning via Dynamic Token Selection for Distribution-Aligned Self-DistillationKnowledge distillation → Knowledge distillation for LLMs
MM-Snowball: Evaluating and Mitigating Hallucination Snowballing in Multimodal Multi-Turn DialogueMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
FlowNar: Scalable Streaming Narration for Long-Form VideosMultimodal LLMs / vision-language models → Video understanding with MLLMs
MemPro: Agentic Memory Systems as Evolvable ProgramsLLM agents and tool use → Agent memory systems
Pause and Think: A Dataset and Benchmark for Video-Grounded Assistive Action SuggestionLLM reasoning → Multimodal reasoning +1
Linguistics-Aware Non-Distortionary LLM WatermarkingLLM safety and security → Watermarking generated content and models
TRACE: Trajectory Risk-Aware Compression for Long-Horizon Agent SafetyLLM agents and tool use → Agent threats, guardrails and governance
MemGraphRAG: Memory-based Multi-Agent System for Graph Retrieval-Augmented GenerationRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG +1
CARE-RL: Capability-Aware Reinforcement Learning for Mitigating Cross-Domain ConflictsLLM reasoning → Reinforcement learning for reasoning +1
Toward Agentic Governance: What Shapes LLM-Agent Intervention in Public Forums?LLM agents and tool use → Social simulation and multi-agent debate
ASAP: Advancing Medical Volumetric Representation Learning with Anatomy-aware Semantically-adaptive Pre-trainingMultimodal LLMs / vision-language models → Medical vision-language models
Critic-R: Improving Agentic Search using Instruction-tuned Retrievers with Natural Language Introspective FeedbackLLM agents and tool use → Web, search, and research agents
Improving Visual Representation Alignment Generation with GRPODiffusion models → Diffusion foundations and sampling
Sandboxed Coding Agents are Competitive Omni-modal Task SolversMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
LASER: Loss-Aware Singular-value Decomposition and Rank Allocation for Efficient Low-Precision Vision-Language ModelsLLM reasoning → Multimodal reasoning +1
Same Payload, Different Channel: Measuring Trust Asymmetry in Tool-Using Language ModelsLLM agents and tool use → Agent threats, guardrails and governance +1
Decomposed On-Policy Distillation for Vision-Language Reasoning: Steering Gradients for Visual GroundingLLM reasoning → Multimodal reasoning +1
DeepLatent: Think with Images via Parallel Latent Visual ReasoningLLM reasoning → Multimodal reasoning +1
Interpretable Policy Distillation for Power Grid Topology ControlDeep reinforcement learning methods → RL for energy, networks and operations
Probe Before You Edit: Probing-Guided Molecular Optimization for LLM Agents in Structure-Based Drug DesignLLM agents and tool use → Medical and scientific domain agents
ETC: Extreme Token Compression via Task-aware Visual Information Distillation in VLMsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
PACE: Phase-Aware Chunk Execution for Robot Policies with Action ChunkingVision-language-action models → Efficient VLA inference and adaptation
DREAM-S: Speculative Decoding with Searchable Drafting and Target-Aware Refinement for Multimodal GenerationMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
KACE: Knowledge-Adaptive Context Engineering for Mathematical ReasoningLLM reasoning → Mathematical reasoning
DriveAnchor: Progressive Anchor-based Flow Learning for Autonomous Driving PlanningDiffusion models → Diffusion and flow policies for robotics +1
Acting with AI: An Interaction-Based Framework for Agentic Tort LiabilityLLM agents and tool use → Agent threats, guardrails and governance
PaCo-VLA: Passivity-Shielded Compliance Prior for Contact-Rich Vision-Language-Action ManipulationVision-language-action models → Tactile, force-aware and humanoid VLA
Skill or Skip? Learning Selective Skill Invocation in Agentic Tasks via Dual-Granularity Preference LearningLLM agents and tool use → Agent skill learning and evolution
LaSR: Context-Aware Speech Recognition via Latent ReasoningLLM reasoning → Audio-language model reasoning

All 281