Papers

New on arXiv

Thu 5 Jun 2025

495 papers

Dynamic Context Tuning for Retrieval-Augmented Generation: Enhancing Multi-Turn Planning and Tool AdaptationRetrieval-augmented generation → Conversational and speech RAG
Customizing Speech Recognition Model with Large Language Model FeedbackLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Confidence Is All You Need: Few-Shot RL Fine-Tuning of Language ModelsLLM reasoning → Reinforcement learning for reasoning
Exploring Adversarial Watermarking in Transformer-Based Models: Transferability and Robustness Against Defense Mechanism for Medical ImagesLLM safety and security → Watermarking generated content and models
Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature EngineeringLLM safety and security → Prompt injection and agent security
List-Level Distribution Coupling with Applications to Speculative Decoding and Lossy CompressionEfficient LLM inference and serving → Speculative decoding
Beyond RLHF and NLHF: Population-Proportional Alignment under an Axiomatic FrameworkLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Toward Greater Autonomy in Materials Discovery Agents: Unifying Planning, Physics, and ScientistsLLM agents and tool use → Medical and scientific domain agents
When Maximum Entropy Misleads Policy OptimizationDeep reinforcement learning methods → RL theory and sample complexity
SynthesizeMe! Inducing Persona-Guided Prompts for Personalized Reward Models in LLMsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
SoK: Are Watermarks in LLMs Ready for Deployment?LLM safety and security → Watermarking generated content and models
UTSA-NLP at ArchEHR-QA 2025: Improving EHR Question Answering via Self-Consistency PromptingLLM reasoning → Clinical and biomedical applications
MMTU: A Massive Multi-Task Table Understanding and Reasoning BenchmarkLLM reasoning → Knowledge- and tool-augmented reasoning
Policy Search, Retrieval, and Composition via Task Similarity in Collaborative Agentic SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion TransformersDiffusion models → 3D, point cloud and scene generation
ScaleRTL: Scaling LLMs with Reasoning Data and Test-Time Compute for Accurate RTL Code GenerationLLM reasoning → Reasoning for agents, planning and code
When Semantics Mislead Vision: Mitigating Large Multimodal Models Hallucinations in Scene Text Spotting and UnderstandingMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Agentomics-ML: Autonomous Machine Learning Experimentation Agent for Genomic and Transcriptomic DataLLM agents and tool use → Medical and scientific domain agents
Avoiding Death through Fear Intrinsic ConditioningDeep reinforcement learning methods → Deep RL algorithms and representations
MORSE-500: A Programmatically Controllable Video Benchmark to Stress-Test Multimodal ReasoningLLM reasoning → Multimodal reasoning +1
StealthInk: A Multi-bit and Stealthy Watermark for Large Language ModelsLLM safety and security → Watermarking generated content and models
FocusDiff: Advancing Fine-Grained Text-Image Alignment for Autoregressive Visual Generation through RLDiffusion models → Image generation, text-to-image and editing
Towards Reliable Identification of Diffusion-based Image ManipulationsDiffusion models → Image generation, text-to-image and editing
Zeroth-Order Optimization Finds Flat MinimaDeep reinforcement learning methods → RL theory and sample complexity
MLLM-CL: Continual Learning for Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Interpretation Meets Safety: A Survey on Interpretation Methods and Tools for Improving LLM SafetyLLM safety and security → Safety alignment, refusal and guardrails
Sentinel: SOTA model to protect against prompt injectionsLLM safety and security → Prompt injection and agent security
Causal Policy Learning in Reinforcement Learning: Backdoor-Adjusted Soft Actor-CriticDeep reinforcement learning methods → Deep RL algorithms and representations
Structured Labeling Enables Faster Vision-Language Models for End-to-End Autonomous DrivingLLM reasoning → Chain-of-thought reasoning for driving +2
BYO-Eval: Build Your Own Dataset for Fine-Grained Visual Assessment of Multimodal Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
LLMs Can Compensate for Deficiencies in Visual RepresentationsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Prefix Grouper: Efficient GRPO Training through Shared-Prefix ForwardLLM reasoning → Reinforcement learning for reasoning
Robustness Evaluation for Video Models with Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Coordinated Robustness Evaluation Framework for Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Mixture-of-Experts Meets In-Context Reinforcement LearningMixture-of-experts → MoE in reinforcement learning and robotics +1
SIV-Bench: A Video Benchmark for Social Interaction Understanding and ReasoningLLM reasoning → Reasoning benchmarks and evaluation
Constructive Symbolic Reinforcement Learning via Intuitionistic Logic and Goal-Chaining InferenceDeep reinforcement learning methods → Interpretable and logic-guided RL
Dream to Generalize: Zero-Shot Model-Based Reinforcement Learning for Unseen Visual DistractionsWorld models → Latent dynamics models and model-based RL +1
Self-Predictive Dynamics for Generalization of Vision-based Reinforcement LearningDeep reinforcement learning methods → RL for robot locomotion and manipulation
Contrastive Flow MatchingDiffusion models → Flow matching, time series and science
VideoMathQA: Benchmarking Mathematical Reasoning via Multimodal Understanding in VideosLLM reasoning → Multimodal reasoning
Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning DatasetsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Inference-Time Hyper-Scaling with KV Cache CompressionLLM reasoning → Test-time compute and efficient thinking +1
SparseMM: Head Sparsity Emerges from Visual Concept Responses in MLLMsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial ReasoningLLM reasoning → Spatial and embodied reasoning
Exploring Diffusion Transformer Designs via GraftingDiffusion models → Diffusion foundations and sampling
Flattery, Fluff, and Fog: Diagnosing and Mitigating Idiosyncratic Biases in Preference ModelsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Kinetics: Rethinking Test-Time Scaling LawsLLM reasoning → Test-time compute and efficient thinking
Unleashing Hour-Scale Video Training for Long Video-Language UnderstandingMultimodal LLMs / vision-language models → Video understanding with MLLMs
MINT-CoT: Enabling Interleaved Visual Tokens in Mathematical Chain-of-Thought ReasoningLLM reasoning → Multimodal reasoning
AV-Reasoner: Improving and Benchmarking Clue-Grounded Audio-Visual Counting for MLLMsLLM reasoning → Audio-language model reasoning
Does Your 3D Encoder Really Work? When Pretrain-SFT from 2D VLMs Meets 3D VLMsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Improving Data Efficiency for LLM Reinforcement Fine-tuning Through Difficulty-targeted Online Data Selection and Rollout ReplayLLM reasoning → Reinforcement learning for reasoning
Constrained Entropic Unlearning: A Primal-Dual Framework for Large Language ModelsLLM safety and security → Machine unlearning
Time to Talk: LLM Agents for Asynchronous Group Communication in Mafia GamesLLM agents and tool use → General LLM agent frameworks and benchmarks
ProRefine: Inference-Time Prompt Refinement with Textual FeedbackLLM agents and tool use → General LLM agent frameworks and benchmarks
SeedVR2: One-Step Video Restoration via Diffusion Adversarial Post-TrainingDiffusion models → Inverse problems and image restoration
DM-SegNet: Dual-Mamba Architecture for 3D Medical Image Segmentation with Global Context ModelingState-space sequence models (S4, Mamba) → Mamba for medical image analysis
Control Tax: The Price of Keeping AI in CheckLLM agents and tool use → Agent threats, guardrails and governance
Sample Complexity and Representation Ability of Test-time Scaling ParadigmsLLM reasoning → Test-time compute and efficient thinking
A Smooth Sea Never Made a Skilled SAILOR: Robust Imitation via Learning to SearchOffline RL and imitation learning from data → Imitation learning methods
EOC-Bench: Can MLLMs Identify, Recall, and Forecast Objects in an Egocentric World?Multimodal LLMs / vision-language models → Embodied and 3D vision-language models
Video World Models with Long-term Spatial MemoryWorld models → Video-generation world models and simulators
Micro-Act: Mitigating Knowledge Conflict in LLM-based RAG via Actionable Self-ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning +1
LeanPO: Lean Preference Optimization for Likelihood Alignment in Video-LLMsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Just Enough Thinking: Efficient Reasoning with Adaptive Length Penalties Reinforcement LearningLLM reasoning → Test-time compute and efficient thinking
CLATTER: Comprehensive Entailment Reasoning for Hallucination DetectionLLM reasoning → Hallucination detection and mitigation
Evaluating Sparse Autoencoders: From Shallow Design to Matching PursuitMechanistic interpretability → Sparse autoencoder training and evaluation
Progressive Tempering Sampler with DiffusionDiffusion models → Diffusion foundations and sampling
DSG-World: Learning a 3D Gaussian World Model from Dual State VideosWorld models → Video-generation world models and simulators
LLM-First Search: Self-Guided Exploration of the Solution SpaceLLM reasoning → Test-time compute and efficient thinking
Optimal-PhiBE: A PDE-based Model-free framework for Continuous-time Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
Follow-Your-Motion: Video Motion Transfer via Efficient Spatial-Temporal Decoupled FinetuningDiffusion models → Video generation
RELIC: Evaluating Complex Reasoning via the Recognition of Languages In-ContextLLM reasoning → Reasoning benchmarks and evaluation
OGGSplat: Open Gaussian Growing for Generalizable Reconstruction with Expanded Field-of-ViewDiffusion models → 3D, point cloud and scene generation
Quantifying Cross-Modality Memorization in Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Counterfactual reasoning: an analysis of in-context emergenceLLM reasoning → Causal reasoning in LLMs
TreeRPO: Tree Relative Policy OptimizationLLM reasoning → Reinforcement learning for reasoning
ECoRAG: Evidentiality-guided Compression for Long Context RAGRetrieval-augmented generation → Generation and context integration
Dissecting Bias in LLMs: A Mechanistic Interpretability PerspectiveMechanistic interpretability → Probing deception, bias and beliefs
LiPo: A Lightweight Post-optimization Framework for Smoothing Action Chunks Generated by Learned PoliciesOffline RL and imitation learning from data → Robot manipulation imitation learning
Resisting Contextual Interference in RAG via Parametric-Knowledge ReinforcementRetrieval-augmented generation → RAG robustness and adaptive retrieval
CIVET: Systematic Evaluation of Understanding in VLMsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
DiCoRe: Enhancing Zero-shot Event Detection via Divergent-Convergent LLM ReasoningLLM reasoning → Causal reasoning in LLMs
Membership Inference Attacks on Sequence ModelsLLM safety and security → Membership inference attacks
Realizing Text-Driven Motion Generation on NAO Robot: A Reinforcement Learning-Optimized Control PipelineDeep reinforcement learning methods → RL for robot locomotion and manipulation
Whole-Body Constrained Learning for Legged Locomotion via Hierarchical OptimizationDeep reinforcement learning methods → RL for robot locomotion and manipulation
Truly Self-Improving Agents Require Intrinsic Metacognitive LearningSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
Astraea: A Token-wise Acceleration Framework for Video Diffusion TransformersDiffusion models → Video generation +1
Interpretable Multimodal Framework for Human-Centered Street Assessment: Integrating Visual-Language Models for Perceptual Urban DiagnosticsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
RIVAL: Reinforcement Learning with Iterative and Adversarial Optimization for Machine TranslationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Reason-to-Recommend: Using Interaction-of-Thought Reasoning to Enhance LLM RecommendationLLM reasoning → LLM reasoning for recommendation
DemoSpeedup: Accelerating Visuomotor Policies via Entropy-Guided Demonstration AccelerationOffline RL and imitation learning from data → Robot manipulation imitation learning
Toward Automated Robustness Evaluation of Mathematical ReasoningLLM reasoning → Mathematical reasoning
Mathematical Reasoning for Unmanned Aerial Vehicles: A RAG-Based Approach for Complex Arithmetic ReasoningLLM reasoning → Spatial and embodied reasoning
Towards Storage-Efficient Visual Document Retrieval: An Empirical Study on Reducing Patch-Level EmbeddingsEfficient LLM inference and serving → Visual token pruning in multimodal models
TextVidBench: A Benchmark for Long Video Scene Text UnderstandingMultimodal LLMs / vision-language models → Video understanding with MLLMs
Agentic AI for Intent-Based Industrial AutomationLLM agents and tool use → General LLM agent frameworks and benchmarks
Evaluating Chinese Large Language Models: The Influence of Persona Assignment on Stereotypes and SafeguardsLLM safety and security → Safety alignment, refusal and guardrails
APVR: Hour-Level Long Video Understanding with Adaptive Pivot Visual Information RetrievalMultimodal LLMs / vision-language models → Video understanding with MLLMs

All 495