Papers

New on arXiv

Thu 5 Mar 2026

437 papers

S-Path-RAG: Semantic-Aware Shortest-Path Retrieval Augmented Generation for Multi-Hop Knowledge Graph Question AnsweringRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
RBF-Solver: A Multistep Sampler for Diffusion Probabilistic Models via Radial Basis FunctionsDiffusion models → Diffusion foundations and sampling
Scaling Agentic Capabilities, Not Context: Efficient Reinforcement Finetuning for Large ToolspacesLLM agents and tool use → Reinforcement learning for LLM agents
On the Generalization Capacities of MLLMs for Spatial IntelligenceMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Breaking the Geometric Bottleneck: Contrastive Expansion in Asymmetric Cross-Modal DistillationKnowledge distillation → General knowledge distillation methods
Thinking with Gaze: Sequential Eye-Tracking as Visual Reasoning Supervision for Medical VLMsLLM reasoning → Clinical and biomedical applications
TML-Bench: Benchmark for Data Science Agents on Tabular ML TasksLLM agents and tool use → Web, search, and research agents
Score-Guided Proximal Projection: A Unified Geometric Framework for Rectified Flow EditingDiffusion models → Image generation, text-to-image and editing
MIRACL: A Diverse Meta-Reinforcement Learning for Multi-Objective Multi-Echelon Combinatorial Supply Chain OptimisationDeep reinforcement learning methods → RL for energy, networks and operations
Safe-Night VLA: Seeing the Unseen via Thermal-Perceptive Vision-Language-Action Models for Safety-Critical ManipulationVision-language-action models → Safety and adversarial attacks
CodeScout: Contextual Problem Statement Enhancement for Software AgentsLLM agents and tool use → Software engineering coding agents
Designing Explainable Conversational Agentic Systems for Guaran'i SpeakersLLM agents and tool use → General LLM agent frameworks and benchmarks
Revisiting the (Sub)Optimality of Best-of-N for Inference-Time AlignmentLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Agentic AI -- Physicist Collaboration in Experimental Particle Physics: A Proof-of-Concept Measurement with LEP Open DataLLM agents and tool use → Web, search, and research agents
LTLGuard: Formalizing LTL Specifications with Compact Language Models and Lightweight Symbolic ReasoningLLM reasoning → Reasoning for agents, planning and code
Interpretable Perception and Reasoning for Audiovisual GeolocationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Reasoning Models Struggle to Control their Chains of ThoughtLLM reasoning → Prompting and chain-of-thought reasoning
Towards Robust Retrieval-Augmented Generation Based on Knowledge Graph: A Comparative AnalysisRetrieval-augmented generation → RAG robustness and adaptive retrieval
MultiHaystack: Benchmarking Multimodal Retrieval and Reasoning over 40K Images, Videos, and DocumentsLLM reasoning → Multimodal reasoning +1
Autonomous Algorithm Discovery for Ptychography via Evolutionary LLM ReasoningLLM reasoning → Clinical and biomedical applications
Longitudinal Lesion Inpainting in Brain MRI via 3D Region Aware DiffusionDiffusion models → Medical imaging and segmentation
SecureRAG-RTL: A Retrieval-Augmented, Multi-Agent, Zero-Shot LLM-Driven Framework for Hardware Vulnerability DetectionRetrieval-augmented generation → Privacy and security of RAG +1
Reinforcement Learning for Power-Flow Network AnalysisDeep reinforcement learning methods → RL for energy, networks and operations
TransMASK: Masked State Representation through Learned TransformationOffline RL and imitation learning from data → Robot manipulation imitation learning
Keeping the Evidence Chain: Semantic Evidence Allocation for Training-Free Token Pruning in Video Temporal GroundingEfficient LLM inference and serving → Visual token pruning in multimodal models
When Rubrics Fail: Error Enumeration as Reward in Reference-Free RL Post-Training for Virtual Try-OnLLM alignment / RLHF / preference optimisation → Reward hacking and misspecification
Relational Semantic Reasoning on 3D Scene Graphs for Open World Interactive Object SearchLLM reasoning → Spatial and embodied reasoning
Real Faults in Model Context Protocol (MCP) Software: a Comprehensive TaxonomyLLM agents and tool use → Agent threats, guardrails and governance
Making Reconstruction FID Predictive of Diffusion Generation FIDDiffusion models → Diffusion foundations and sampling
Rethinking Concept Bottleneck Models: From Pitfalls to SolutionsLLM reasoning → Multimodal reasoning
RACAS: Controlling Diverse Robots With a Single Agentic SystemLLM agents and tool use → Multimodal and embodied agents
Safer Reasoning Traces: Measuring and Mitigating Chain-of-Thought Leakage in LLMsLLM reasoning → Reasoning for agents, planning and code
Real-Time AI Service Economy: A Framework for Agentic Computing Across the ContinuumLLM agents and tool use → General LLM agent frameworks and benchmarks
Thinking with Spatial Code for Physical-World Video ReasoningLLM reasoning → Spatial and embodied reasoning
A Novel Hybrid Heuristic-Reinforcement Learning Optimization Approach for a Class of Railcar Shunting ProblemsDeep reinforcement learning methods → RL for autonomous driving and traffic
Why Depth Matters in Parallelizable Sequence Models: A Lie Algebraic ViewState-space sequence models (S4, Mamba) → Core SSM architectures for language
CBR-to-SQL: Rethinking Retrieval-based Text-to-SQL using Case-based Reasoning in the Healthcare DomainLLM reasoning → Clinical and biomedical applications +1
EigenData: A Self-Evolving Multi-Agent Platform for Function-Calling Data Synthesis, Auditing, and RepairLLM agents and tool use → Tool learning and function calling +1
AutothinkRAG: Complexity-Aware Control of Retrieval-Augmented Reasoning for Image-Text InteractionLLM reasoning → Multimodal reasoning +1
RoboPocket: Improve Robot Policies Instantly with Your PhoneOffline RL and imitation learning from data → Robot manipulation imitation learning
Accelerating Text-to-Video Generation with Calibrated Sparse AttentionDiffusion models → Video generation
Reasoning Theater: Disentangling Model Beliefs from Chain-of-ThoughtLLM reasoning → Prompting and chain-of-thought reasoning
Observing and Controlling Features in Vision-Language-Action ModelsVision-language-action models → VLA architectures and generalization
Towards Multimodal Lifelong Understanding: A Dataset and Agentic BaselineLLM agents and tool use → Multimodal and embodied agents
Leveraging LLM Parametric Knowledge for Fact Checking without RetrievalLLM agents and tool use → Web, search, and research agents
HALP: Detecting Hallucinations in Vision-Language Models without Generating a Single TokenMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Beyond Scattered Acceptance: Fast and Coherent Inference for DLMs via Longest Stable PrefixesDiffusion models → Discrete and text diffusion
Latent Wasserstein Adversarial Imitation LearningOffline RL and imitation learning from data → Imitation learning methods
Planning in 8 Tokens: A Compact Discrete Tokenizer for Latent World ModelWorld models → Latent predictive world models
CRISP: Compressed Reasoning via Iterative Self-Policy DistillationLLM reasoning → Prompting and chain-of-thought reasoning
DARK: Diagonal-Anchored Repulsive Knowledge Distillation for Vision-Language Models under Extreme CompressionMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
PhysiFlow: Physics-Aware Humanoid Whole-Body VLA via Multi-Brain Latent Flow Matching and Robust TrackingVision-language-action models → Tactile, force-aware and humanoid VLA
An Exploration-Analysis-Disambiguation Reasoning Framework for Word Sense Disambiguation with Low-Parameter LLMsLLM reasoning → Prompting and chain-of-thought reasoning
DiSCTT: Consensus-Guided Self-Curriculum for Efficient Test-Time Adaptation in ReasoningLLM reasoning → Test-time compute and efficient thinking
InfoFlow KV: Information-Flow-Aware KV Recomputation for Long ContextRetrieval-augmented generation → Generation and context integration
Building Effective AI Coding Agents for the Terminal: Scaffolding, Harness, Context Engineering, and Lessons LearnedLLM agents and tool use → Software engineering coding agents
Frequency-Aware Error-Bounded Caching for Accelerating Diffusion TransformersDiffusion models → Diffusion foundations and sampling
Latent Policy Steering through One-Step Flow PoliciesOffline RL and imitation learning from data → Offline RL algorithms +1
WebChain: A Large-Scale Human-Annotated Dataset of Real-World Web Interaction TracesLLM agents and tool use → Web, search, and research agents
STRUCTUREDAGENT: Planning with AND/OR Trees for Long-Horizon Web TasksLLM agents and tool use → Web, search, and research agents
Knowledge Divergence and the Value of Debate for Scalable OversightLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment +1
Online Self-Training for Co-Adaptation in Hierarchical Diffusion PoliciesDiffusion models → Diffusion and flow policies for robotics
X-RAY: Mapping LLM Reasoning Capability via Formalized and Calibrated ProbesLLM reasoning → Reasoning benchmarks and evaluation
Whispering to a Blackbox: Bootstrapping Frozen OCR with Visual PromptsOffline RL and imitation learning from data → Robot manipulation imitation learning +1
Wiki-R1: Incentivizing Multimodal Reasoning for Knowledge-based VQA via Data and Sampling CurriculumLLM reasoning → Multimodal reasoning
GCAgent: Enhancing Group Chat Communication through Dialogue Agents SystemLLM agents and tool use → General LLM agent frameworks and benchmarks
Boosting ASR Robustness via Test-Time Reinforcement Learning with Audio-Text Semantic RewardsLLM alignment / RLHF / preference optimisation → Preference alignment for recommenders and RAG +1
The Geometric Inductive Bias of Grokking: Bypassing Phase Transitions via Architectural TopologyMechanistic interpretability → Circuits and mechanistic analysis
KARL: Knowledge Agents via Reinforcement LearningLLM agents and tool use → Web, search, and research agents +1
Balancing Coverage and Draft Latency in Vocabulary Trimming for Faster Speculative DecodingEfficient LLM inference and serving → Speculative decoding
Core-based Hierarchies for Efficient GraphRAGRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
Masked diffusion LLMs can use EoS tokens for hidden reasoningLLM reasoning → Diffusion reasoning and test-time scaling +1
Critic in the Loop: A Tri-System VLA Framework for Robust Long-Horizon ManipulationVision-language-action models → Efficient VLA inference and adaptation
Limited-Angle CT Reconstruction Using Multi-Volume Latent Consistency ModelDiffusion models → Medical imaging and segmentation
Mario: Multimodal Graph Reasoning with Large Language ModelsLLM reasoning → Multimodal reasoning
C2-Faith: Benchmarking LLM Judges for Causal and Coverage Faithfulness in Chain-of-Thought ReasoningLLM reasoning → Prompting and chain-of-thought reasoning
Act, Think or Abstain: Complexity-Aware Adaptive Inference for Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation
Feature Resemblance: Towards a Theoretical Understanding of Analogical Reasoning in TransformersLLM reasoning → Reasoning benchmarks and evaluation
Particle-Guided Diffusion for Gas-Phase Reaction KineticsDiffusion models → Flow matching, time series and science
LBM: Hierarchical Large Auto-Bidding Model via Reasoning and ActingLLM reasoning → Reasoning for agents, planning and code
Bidirectional Curriculum Generation: A Multi-Agent Framework for Data-Efficient Mathematical ReasoningLLM reasoning → Reinforcement learning for reasoning +1
SeedPolicy: Horizon Scaling via Self-Evolving Diffusion Policy for Robot ManipulationDiffusion models → Diffusion and flow policies for robotics +1
Decoupling Task and Behavior: A Two-Stage Reward Curriculum in Reinforcement Learning for RoboticsDeep reinforcement learning methods → Deep RL algorithms and representations
Diff-ES: Stage-wise Structural Diffusion Pruning via Evolutionary SearchDiffusion models → Diffusion foundations and sampling
ARC-TGI: Human-Validated Task Generators with Reasoning Chain Templates for ARC-AGILLM reasoning → Multimodal reasoning
Orthogonal Spatial-temporal Distributional Transfer for 4D GenerationDiffusion models → 3D, point cloud and scene generation
Reward-Conditioned Reinforcement LearningLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
NeuronMoE: Neuron-Guided Mixture-of-Experts for Efficient Multilingual LLM ExtensionMixture-of-experts → MoE large language models
WebFactory: Automated Compression of Foundational Language Intelligence into Grounded Web AgentsLLM agents and tool use → Web, search, and research agents
Enhancing Zero-shot Commonsense Reasoning by Integrating Visual Knowledge via Machine ImaginationLLM reasoning → Multimodal reasoning
Good-Enough LLM Obfuscation (GELO)Efficient LLM inference and serving → Early exit and edge inference
Survive at All Costs: Exploring LLM's Risky Behaviors under Survival PressureLLM agents and tool use → Agent threats, guardrails and governance
Competitive Multi-Operator Reinforcement Learning for Joint Pricing and Fleet Rebalancing in AMoD SystemsDeep reinforcement learning methods → RL for energy, networks and operations
ThaiSafetyBench: Assessing Language Model Safety in Thai Cultural ContextsLLM safety and security → Safety alignment, refusal and guardrails
Think, Then Verify: A Hypothesis-Verification Multi-Agent Framework for Long Video UnderstandingLLM reasoning → Multimodal reasoning
3D-RFT: Reinforcement Fine-Tuning for Video-based 3D Scene UnderstandingLLM reasoning → Multimodal reasoning
VRM: Teaching Reward Models to Understand Authentic Human PreferencesLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Mixture of Universal Experts: Scaling Virtual Width via Depth-Width TransformationMixture-of-experts → Expert routing strategies
When Weak LLMs Speak with Confidence, Preference Alignment Gets StrongerLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
VisionPangu: A Compact and Fine-Grained Multimodal Assistant with 1.7B ParametersMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning

All 437