Papers

New on arXiv

Wed 4 Jun 2025

471 papers

Simulating Human Behavior with the Psychological-mechanism Agent: Integrating Feeling, Thought, and ActionLLM agents and tool use → Social simulation and multi-agent debate
Training a Scientific Reasoning Model for ChemistryLLM reasoning → Clinical and biomedical applications
Mechanistic Interpretability of Diffusion Models: Circuit-Level Analysis and Causal ValidationDiffusion models → Diffusion foundations and sampling +1
Self-Composing Policies for Scalable Continual Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
T2-RAGBench: Text-and-Table Benchmark for Evaluating Retrieval-Augmented GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
RedDebate: Safer Responses Through Multi-Agent Red Teaming DebatesLLM agents and tool use → Social simulation and multi-agent debate +1
MANBench: Is Your Multimodal Model Smarter than Human?Multimodal LLMs / vision-language models → MLLM training and visual instruction tuning
RoE-FND: A Case-Based Reasoning Approach with Dual Verification for Fake News Detection via LLMsLLM reasoning → Multimodal reasoning
CyclicReflex: Improving Reasoning Models via Cyclical Reflection Token SchedulingLLM reasoning → Test-time compute and efficient thinking
Beyond the Norm: A Survey of Synthetic Data Generation for Rare EventsSynthetic data and self-improvement for LLMs → Model collapse from synthetic data
SiLIF: Structured State Space Model Dynamics and Parametrization for Spiking Neural NetworksState-space sequence models (S4, Mamba) → Core SSM architectures for language
AI Agent Behavioral ScienceLLM agents and tool use → Social simulation and multi-agent debate
SAVVY: Spatial Awareness via Audio-Visual LLMs through Seeing and HearingLLM reasoning → Audio-language model reasoning
Vision-Language Models Mistake Head Orientation for Gaze Direction: Nonverbal Conversation CuesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Homogeneous Keys, Heterogeneous Values: Exploiting Local KV Cache Asymmetry for Long-Context LLMsEfficient LLM inference and serving → KV cache optimization
Object-level Self-Distillation for Vision PretrainingKnowledge distillation → Detection and dense prediction distillation
AD-EE: Early Exiting for Fast and Reliable Vision-Language Models in Autonomous DrivingMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection +2
Robust Anti-Backdoor Instruction Tuning in LVLMsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
SGN-CIRL: Scene Graph-based Navigation with Curriculum, Imitation, and Reinforcement LearningDeep reinforcement learning methods → RL for robot navigation and UAV control
Matching Markets Meet LLMs: Algorithmic Reasoning with Ranked PreferencesLLM reasoning → Test-time compute and efficient thinking
Leveraging Reward Models for Guiding Code Review Comment GenerationLLM alignment / RLHF / preference optimisation → Process reward models for reasoning
Aligning Large Language Models with Implicit Preferences from User-Generated ContentLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Watermarking Degrades Alignment in Language Models: Analysis and MitigationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods +1
RETRO SYNFLOW: Discrete Flow Matching for Accurate and Diverse Single-Step RetrosynthesisDiffusion models → Flow matching, time series and science
HMAR: Efficient Hierarchical Masked Auto-Regressive Image GenerationDiffusion models → Image generation, text-to-image and editing
MedAgentGym: A Scalable Agentic Training Environment for Code-Centric Reasoning in Biomedical Data ScienceLLM reasoning → Clinical and biomedical applications +1
Unsupervised Meta-Testing with Conditional Neural Processes for Hybrid Meta-Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Bridging the Performance Gap Between Target-Free and Target-Based Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Is Perturbation-Based Image Protection Disruptive to Image Editing?Diffusion models → Robustness, privacy and adversarial aspects
Through the Stealth Lens: Attention-Aware Defenses Against Poisoning in RAGRetrieval-augmented generation → Privacy and security of RAG
A Statistical Physics of Language Model ReasoningLLM reasoning → Test-time compute and efficient thinking
WorldPrediction: A Benchmark for High-level World Modeling and Long-horizon Procedural PlanningWorld models → Embodied and robot world models
A Risk-Aware Reinforcement Learning Reward for Financial TradingDeep reinforcement learning methods → Adversarial robustness and RL in finance
HuGeDiff: 3D Human Generation via Diffusion with Gaussian SplattingDiffusion models → 3D, point cloud and scene generation
RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for RoboticsLLM reasoning → Spatial and embodied reasoning +1
RedRFT: A Light-Weight Benchmark for Reinforcement Fine-Tuning-Based Red TeamingLLM safety and security → Prompt injection and agent security
The Cost of Dynamic Reasoning: Demystifying AI Agents and Test-Time Scaling from an AI Infrastructure PerspectiveLLM reasoning → Reasoning for agents, planning and code +1
A Lyapunov Drift-Plus-Penalty Method Tailored for Reinforcement Learning with Queue StabilityDeep reinforcement learning methods → RL theory and sample complexity
Relational reasoning and inductive bias in transformers and large language modelsLLM reasoning → Reasoning benchmarks and evaluation
Backbone Augmented Training for AdaptationsDiffusion models → Image generation, text-to-image and editing
Automated Skill Discovery for Language Agents through Exploration and Iterative FeedbackLLM agents and tool use → Agent skill learning and evolution
DrSR: LLM based Scientific Equation Discovery with Dual Reasoning from Data and ExperienceLLM reasoning → Knowledge- and tool-augmented reasoning
Evaluating MLLMs with Multimodal Multi-image Reasoning BenchmarkLLM reasoning → Multimodal reasoning +2
RSVP: Reasoning Segmentation via Visual Prompting and Multi-modal Chain-of-ThoughtLLM reasoning → Multimodal reasoning +1
Voyager: Long-Range and World-Consistent Video Diffusion for Explorable 3D Scene GenerationDiffusion models → 3D, point cloud and scene generation
Struct2D: A Perception-Guided Framework for Spatial Reasoning in MLLMsLLM reasoning → Spatial and embodied reasoning +1
Pseudo-Simulation for Autonomous DrivingEnd-to-end autonomous driving → Adversarial scenarios and safety testing
OWMM-Agent: Open World Mobile Manipulation With Multi-modal Agentic Data SynthesisLLM agents and tool use → Multimodal and embodied agents +1
FullDiT2: Efficient In-Context Conditioning for Video Diffusion TransformersDiffusion models → Video generation
Diffusion Domain Teacher: Diffusion Guided Domain Adaptive Object DetectorDiffusion models → Image generation, text-to-image and editing
Does Thinking More always Help? Mirage of Test-Time Scaling in Reasoning ModelsLLM reasoning → Test-time compute and efficient thinking
Advancing Multimodal Reasoning: From Optimized Cold Start to Staged Reinforcement LearningLLM reasoning → Multimodal reasoning +1
EPiC: Towards Lossless Speedup for Reasoning Training through Edge-Preserving CoT CondensationLLM reasoning → Prompting and chain-of-thought reasoning
Cascadia: An Efficient Cascade Serving System for Large Language ModelsEfficient LLM inference and serving → Serving systems and request scheduling
R-Search: Empowering LLM Reasoning with Search via Multi-Reward Reinforcement LearningLLM reasoning → Knowledge- and tool-augmented reasoning
Long or short CoT? Investigating Instance-level Switch of Large Reasoning ModelsLLM reasoning → Prompting and chain-of-thought reasoning
SuperWriter: Reflection-Driven Long-Form Generation with Large Language ModelsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
OpenThoughts: Data Recipes for Reasoning ModelsLLM reasoning → Reasoning benchmarks and evaluation
Physics-Constrained Flow Matching: Sampling Generative Models with Hard ConstraintsDiffusion models → Flow matching, time series and science
Horizon Reduction Makes RL ScalableOffline RL and imitation learning from data → Offline RL algorithms
Image Editing As Programs with Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
SLAC: Safe and Efficient Real-Robot Reinforcement Learning via Unsupervised Simulation Pre-TrainingDeep reinforcement learning methods → RL for robot locomotion and manipulation
MMR-V: What's Left Unsaid? A Benchmark for Multimodal Deep Reasoning in VideosLLM reasoning → Multimodal reasoning +1
macOSWorld: A Multilingual Interactive Benchmark for GUI AgentsLLM agents and tool use → GUI and mobile agents
TRiSM for Agentic AI: A Review of Trust, Risk, and Security Management in LLM-based Agentic Multi-Agent SystemsLLM agents and tool use → Agent threats, guardrails and governance
Guided Speculative Inference for Efficient Test-Time Alignment of LLMsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods +1
Rectified Sparse AttentionEfficient LLM inference and serving → KV cache optimization
TextAtari: 100K Frames Game Playing with Language AgentsLLM agents and tool use → Reinforcement learning for LLM agents
Multimodal Tabular Reasoning with Privileged Structured InformationLLM reasoning → Knowledge- and tool-augmented reasoning
LaF-GRPO: In-Situ Navigation Instruction Generation for the Visually Impaired via GRPO with LLM-as-Follower RewardLLM reasoning → Spatial and embodied reasoning
Progressive Mastery: Customized Curriculum Learning with Guided Prompting for Mathematical ReasoningLLM reasoning → Reinforcement learning for reasoning
Crowd-SFT: Crowdsourcing for LLM AlignmentLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Lacuna Inc. at SemEval-2025 Task 4: LoRA-Enhanced Influence-Based Unlearning for LLMsLLM safety and security → Machine unlearning
Think Like a Person Before Responding: A Multi-Faceted Evaluation of Persona-Guided LLMs for Countering HateLLM reasoning → Reasoning benchmarks and evaluation
Autonomous Vehicle Lateral Control Using Deep Reinforcement Learning with MPC-PID DemonstrationEnd-to-end autonomous driving → Reinforcement learning for driving +1
Mitigating Hallucinations in Large Vision-Language Models via Entity-Centric Multimodal Preference OptimizationMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs +1
Privacy and Security Threat for OpenAI GPTsLLM safety and security → Prompt injection and agent security
Rex-Thinker: Grounded Object Referring via Chain-of-Thought ReasoningLLM reasoning → Multimodal reasoning
AI Agents for Conversational Patient Triage: Preliminary Simulation-Based Evaluation with Real-World EHR DataLLM agents and tool use → Medical and scientific domain agents
Interpretability by Design for Efficient Multi-Objective Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
AgentMisalignment: Measuring the Propensity for Misaligned Behaviour in LLM-Based AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Vocabulary-free few-shot learning for Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Stronger Baselines for Retrieval-Augmented Generation with Long-Context Language ModelsRetrieval-augmented generation → RAG pipeline design and evaluation
Solving Inverse Problems via Diffusion-Based Priors: An Approximation-Free Ensemble Sampling ApproachDiffusion models → Inverse problems and image restoration
Structured Pruning for Diverse Best-of-N Reasoning OptimizationLLM reasoning → Test-time compute and efficient thinking
From Real to Synthetic: Synthesizing Millions of Diversified and Complicated User Instructions with Attributed GroundingLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Graph Counselor: Adaptive Graph Exploration via Multi-Agent Synergy to Enhance LLM ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning +2
Diffusion-based Cumulative Adversarial Purification for Vision Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Vision Remember: Recovering Visual Information in Efficient LVLM with Vision Feature ResamplingMultimodal LLMs / vision-language models → Visual token pruning and efficiency
Multiple Stochastic Prompt Tuning for Few-shot Adaptation under Extreme Domain ShiftMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
More or Less Wrong: A Benchmark for Directional Bias in LLM Comparative ReasoningLLM reasoning → Mathematical reasoning
HSSBench: Benchmarking Humanities and Social Sciences Ability for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Boosting Open-Source LLMs for Program Repair via Reasoning Transfer and LLM-Guided Reinforcement LearningLLM reasoning → Reasoning for agents, planning and code
Magic Mushroom: A Customizable Benchmark for Fine-grained Analysis of Retrieval Noise Erosion in RAG SystemsRetrieval-augmented generation → RAG robustness and adaptive retrieval
Prompt Candidates, then Distill: A Teacher-Student Framework for LLM-driven Data AnnotationKnowledge distillation → Knowledge distillation for LLMs
Vulnerability-Aware Alignment: Mitigating Uneven Forgetting in Harmful Fine-TuningLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Algorithm- and Data-Dependent Generalization Bounds for Diffusion ModelsDiffusion models → Diffusion foundations and sampling
AssetOpsBench: Benchmarking AI Agents for Task Automation in Industrial Asset Operations and MaintenanceLLM agents and tool use → Software engineering coding agents
Multi-objective Aligned Bidword Generation Model for E-commerce Search AdvertisingLLM alignment / RLHF / preference optimisation → Preference alignment for recommenders and RAG
Personalized MR-Informed Diffusion Models for 3D PET Image ReconstructionDiffusion models → Medical imaging and segmentation

All 471