Papers

New on arXiv

Tue 17 Jun 2025

364 papers

Think Clearly: Improving Reasoning via Redundant Token PruningLLM reasoning → Test-time compute and efficient thinking +1
Adaptive Action Duration with Contextual Bandits for Deep Reinforcement Learning in Dynamic EnvironmentsDeep reinforcement learning methods → Deep RL algorithms and representations
LoRA-Mixer: Coordinate Modular LoRA Experts Through Serial Attention RoutingMixture-of-experts → Mixture of LoRA experts
RedTopic: Toward Topic-Diverse Red Teaming of Large Language ModelsLLM safety and security → Prompt injection and agent security
AIMatDesign: Knowledge-Augmented Reinforcement Learning for Inverse Materials Design under Data ScarcityDeep reinforcement learning methods → RL for energy, networks and operations
Understanding Verbatim Memorization in LLMs Through Circuit DiscoveryMechanistic interpretability → Knowledge and representations inside LLMs
Can Vision Language Models Understand Mimed Actions?Multimodal LLMs / vision-language models → Video understanding with MLLMs
VIDEE: Visual and Interactive Decomposition, Execution, and Evaluation of Text Analytics with Intelligent AgentsLLM agents and tool use → Web, search, and research agents
Utility-Driven Speculative Decoding for Mixture-of-ExpertsMixture-of-experts → Efficient MoE training and inference systems +1
FramePrompt: In-context Controllable Animation with Zero Structural ChangesDiffusion models → Video generation
Toward Principled LLM Safety Testing: Solving the Jailbreak Oracle ProblemLLM safety and security → Jailbreaking
Mercury: Ultra-Fast Language Models Based on DiffusionLLM reasoning → Diffusion reasoning and test-time scaling
SafeRL-Lite: A Lightweight, Explainable, and Constrained Reinforcement Learning LibraryDeep reinforcement learning methods → Safe and constrained RL
Semantic uncertainty in advanced decoding methods for LLM generationLLM reasoning → Test-time compute and efficient thinking +1
FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent SpaceDiffusion models → Image generation, text-to-image and editing
OAgents: An Empirical Study of Building Effective AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
SHADE-Arena: Evaluating Sabotage and Monitoring in LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
Stable CDE Autoencoders with Acuity Regularization for Offline Reinforcement Learning in Sepsis TreatmentOffline RL and imitation learning from data → Safe offline RL
GCN-Driven Reinforcement Learning for Probabilistic Real-Time Guarantees in Industrial URLLCDeep reinforcement learning methods → RL for energy, networks and operations
NeuroMoE: A Transformer-Based Mixture-of-Experts Framework for Multi-Modal Neurological Disorder ClassificationMixture-of-experts → Medical and biomedical MoE
Revisiting Reinforcement Learning for LLM Reasoning from A Cross-Domain PerspectiveLLM reasoning → Reinforcement learning for reasoning
Structured Moral Reasoning in Language Models: A Value-Grounded Evaluation FrameworkLLM reasoning → Reasoning benchmarks and evaluation
MDBench: A Synthetic Multi-Document Reasoning Benchmark Generated with Knowledge GuidanceLLM reasoning → Reasoning benchmarks and evaluation
FORTRESS: Frontier Risk Evaluation for National Security and Public SafetyLLM safety and security → Safety alignment, refusal and guardrails
Frequency-Calibrated Membership Inference Attacks on Medical Image Diffusion ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
Winter Soldier: Backdooring Language Models at Pre-Training with Indirect Data PoisoningLLM safety and security → Backdoor and data poisoning attacks
PeRL: Permutation-Enhanced Reinforcement Learning for Interleaved Vision-Language ReasoningLLM reasoning → Multimodal reasoning +1
DETONATE: A Benchmark for Text-to-Image Alignment and Kernelized Direct Preference OptimizationDiffusion models → Diffusion and flow policies for robotics +1
OS-Harm: A Benchmark for Measuring Safety of Computer Use AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
Agentic Plan Caching: Test-Time Memory for Fast and Cost-Efficient LLM AgentsLLM agents and tool use → Software engineering coding agents
CDP: Towards Robust Autoregressive Visuomotor Policy Learning via Causal DiffusionDiffusion models → Diffusion and flow policies for robotics
ASCD: Attention-Steerable Contrastive Decoding for Reducing Hallucination in MLLMMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
RobotSmith: Generative Robotic Tool Design for Acquisition of Complex Manipulation SkillsLLM agents and tool use → Multimodal and embodied agents
Reasoning with Exploration: An Entropy PerspectiveLLM reasoning → Reinforcement learning for reasoning
Optimizing Length Compression in Large Reasoning ModelsLLM reasoning → Test-time compute and efficient thinking
Cost-Aware Routing for Efficient Text-To-Image GenerationDiffusion models → Image generation, text-to-image and editing
Exploring Speaker Diarization with Mixture of ExpertsMixture-of-experts → Multimodal MoE models
Ring-lite: Scalable Reasoning via C3PO-Stabilized Reinforcement Learning for LLMsLLM reasoning → Reinforcement learning for reasoning +1
AgentDistill: Training-Free Agent Distillation with Generalizable MCP BoxesLLM agents and tool use → General LLM agent frameworks and benchmarks +2
Casper: Inferring Diverse Intents for Assistive Teleoperation with Vision Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
AGENTSAFE: Benchmarking the Safety of Embodied Agents on Hazardous InstructionsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Unified Software Engineering Agent as AI Software EngineerLLM agents and tool use → Software engineering coding agents
AIRTBench: Measuring Autonomous AI Red Teaming Capabilities in Language ModelsLLM safety and security → Prompt injection and agent security
Massive Supervised Fine-tuning Experiments Reveal How Data, Layer, and Training Factors Shape LLM Alignment QualityLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Recognition through Reasoning: Reinforcing Image Geo-localization with Large Vision-Language ModelsWorld models → World models for language agents +2
StreetLens: Enabling Human-Centered AI Agents for Neighborhood Assessment from Street View ImageryLLM agents and tool use → Multimodal and embodied agents
Issue Retrieval and Verification Enhanced Supplementary Code Comment GenerationRetrieval-augmented generation → RAG for code
SENIOR: Efficient Query Selection and Preference-Guided Exploration in Preference-based Reinforcement LearningLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
GuiLoMo: Allocating Expert Number and Rank for LoRA-MoE via Bilevel Optimization with GuidedSelection VectorsMixture-of-experts → Mixture of LoRA experts
Revisiting Chain-of-Thought Prompting: Zero-shot Can Be Stronger than Few-shotLLM reasoning → Prompting and chain-of-thought reasoning
Probabilistic Aggregation and Targeted Embedding Optimization for Collective Moral Reasoning in Large Language ModelsLLM reasoning → Reasoning benchmarks and evaluation
Exploring MLLMs Perception of Network Visualization PrinciplesMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Align Your Flow: Scaling Continuous-Time Flow Map DistillationDiffusion models → Flow matching, time series and science
NetRoller: Interfacing General and Specialized Models for End-to-End Autonomous DrivingEnd-to-end autonomous driving → Large language models for driving
TGDPO: Harnessing Token-Level Reward Guidance for Enhancing Direct Preference OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Risk Estimation of Knee Osteoarthritis Progression via Predictive Multi-task Modelling from Efficient Diffusion Model using X-ray ImagesDiffusion models → Medical imaging and segmentation
Doppelganger Method: Breaking Role Consistency in LLM Agent via Prompt-based Transferable Adversarial AttackLLM agents and tool use → Agent threats, guardrails and governance
Train Once, Forget Precisely: Anchored Optimization for Efficient Post-Hoc UnlearningLLM safety and security → Machine unlearning
SIRI-Bench: Challenging VLMs' Spatial Intelligence through Complex Reasoning TasksLLM reasoning → Spatial and embodied reasoning
Toward Safety-First Human-Like Decision Making for Autonomous Vehicles in Time-Varying Traffic FlowEnd-to-end autonomous driving → Cooperative and V2X driving
LingoLoop Attack: Trapping MLLMs via Linguistic Context and State Entrapment into Endless LoopsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
GUI-Robust: A Comprehensive Dataset for Testing GUI Agent Robustness in Real-World AnomaliesLLM agents and tool use → GUI and mobile agents
LexiMark: Robust Watermarking via Lexical Substitutions to Enhance Membership Verification of an LLM's Textual Training DataLLM safety and security → Watermarking generated content and models
Dense360: Dense Understanding from Omnidirectional PanoramasMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Adapting Lightweight Vision Language Models for Radiological Visual Question AnsweringMultimodal LLMs / vision-language models → Medical vision-language models
Vela: Scalable Embeddings with Voice Large Language Models for Multimodal RetrievalMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Model compression using knowledge distillation with integrated gradientsKnowledge distillation → General knowledge distillation methods
MoORE: SVD-based Model MoE-ization for Conflict- and Oblivion-Resistant Multi-Task AdaptationMixture-of-experts → Mixture of LoRA experts
MoTE: Mixture of Ternary Experts for Memory-efficient Large Multimodal ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
LongLLaDA: Unlocking Long Context Capabilities in Diffusion LLMsDiffusion models → Discrete and text diffusion
Unsupervised Skill Discovery through Skill Regions DifferentiationDeep reinforcement learning methods → Deep RL algorithms and representations
RAGtifier: Evaluating RAG Generation Approaches of State-of-the-Art RAG Systems for the SIGIR LiveRAG CompetitionRetrieval-augmented generation → RAG pipeline design and evaluation
Adaptive Reinforcement Learning for Unobservable Random DelaysDeep reinforcement learning methods → Deep RL algorithms and representations
Causally Steered Diffusion for Automated Video Counterfactual GenerationDiffusion models → Video generation
Factored Classifier-Free GuidanceDiffusion models → Diffusion foundations and sampling
HALO: Hierarchical Reinforcement Learning for Large-Scale Adaptive Traffic Signal ControlDeep reinforcement learning methods → RL for autonomous driving and traffic
Advancing Safe Mechanical Ventilation Using Offline RL With Hybrid Actions and Clinically Aligned RewardsOffline RL and imitation learning from data → Safe offline RL
Excessive Reasoning Attack on Reasoning LLMsLLM reasoning → Reasoning for agents, planning and code
ELLIS Alicante at CQs-Gen 2025: Winning the critical thinking questions shared task: LLM-based question generation and selectionLLM reasoning → Reasoning benchmarks and evaluation
AviationLLM: An LLM-based Knowledge System for Aviation TrainingRetrieval-augmented generation → RAG robustness and adaptive retrieval +1
FRIDU: Functional Map Refinement with Guided Image DiffusionDiffusion models → 3D, point cloud and scene generation
Expectation Confirmation Preference Optimization for Multi-Turn Conversational Recommendation AgentLLM agents and tool use → Agentic recommender systems +1
ADRD: LLM-Driven Autonomous Driving Based on Rule-based Decision SystemsEnd-to-end autonomous driving → Large language models for driving
Steering Robots with Inference-Time InteractionsOffline RL and imitation learning from data → Robot manipulation imitation learning
From What to Respond to When to Respond: Timely Response Generation for Open-domain Dialogue AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Re-Initialization Token Learning for Tool-Augmented Large Language ModelsLLM agents and tool use → Tool learning and function calling
Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMsLLM reasoning → Reinforcement learning for reasoning
Causes in neuron diagrams, and testing causal reasoning in Large Language Models. A glimpse of the future of philosophy?LLM reasoning → Reasoning benchmarks and evaluation
Xolver: Multi-Agent Reasoning with Holistic Experience Learning Just Like an Olympiad TeamLLM reasoning → Reasoning for agents, planning and code +1
Narrate2Nav: Real-Time Visual Navigation with Implicit Language Reasoning in Human-Centric EnvironmentsLLM reasoning → Spatial and embodied reasoning +1
From Black Boxes to Transparent Minds: Evaluating and Enhancing the Theory of Mind in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
AgentSynth: Scalable Task Generation for Generalist Computer-Use AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
MAS-LitEval : Multi-Agent System for Literary Translation Quality AssessmentLLM agents and tool use → General LLM agent frameworks and benchmarks
AMPLIFY: Actionless Motion Priors for Robot Learning from VideosWorld models → Embodied and robot world models
Meta-SurDiff: Classification Diffusion Model Optimized by Meta Learning is Reliable for Online Surgical Phase RecognitionDiffusion models → Medical imaging and segmentation
GRAM: A Generative Foundation Reward Model for Reward GeneralizationLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Light Aircraft Game : Basic Implementation and training results analysisDeep reinforcement learning methods → RL for robot navigation and UAV control
S4C: Speculative Sampling with Syntactic and Semantic Coherence for Efficient Inference of Large Language ModelsLLM reasoning → Test-time compute and efficient thinking +1
DCRM: A Heuristic to Measure Response Pair Quality in Preference OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Leveraging Predictive Equivalence in Decision TreesLLM reasoning → Test-time compute and efficient thinking

All 364