Papers

New on arXiv

Fri 17 Apr 2026

393 papers

FineServe: A Fine-Grained Dataset and Characterization of Global LLM Serving WorkloadsEfficient LLM inference and serving → Serving systems and request scheduling
From Consumption to Reflection: Designing Human-AI Relations for Stable ReasoningLLM reasoning → Reasoning benchmarks and evaluation
Improving Heart-Focused Medical Question Answering in LLMs via Variance-Aware Rubric Rewards with GRPOLLM reasoning → Clinical and biomedical applications
EvoSpec: Evolving Speculative Decoding via Real-Time Vocabulary and Parameter AdaptationEfficient LLM inference and serving → Speculative decoding
Governance by Design: Architecting Agentic AI for Organizational Learning and Scalable AutonomyLLM agents and tool use → Software engineering coding agents
M3DocDep: Multi-modal, Multi-page, Multi-document Dependency Chunking with Large Vision-Language ModelsMultimodal LLMs / vision-language models → Document understanding and multimodal RAG +1
ACE: Self-Evolving LLM Coding Framework via Adversarial Unit Test Generation and Preference OptimizationLLM alignment / RLHF / preference optimisation → Process reward models for reasoning +1
Sparse Prefix Caching for Hybrid and Recurrent LLM ServingEfficient LLM inference and serving → KV cache optimization
SAT: Sequential Agent Tuning for Coordinator Free Plug and Play Multi-LLM Training with Monotonic Improvement GuaranteesLLM agents and tool use → Reinforcement learning for LLM agents
MedMamba: Recasting Mamba for Medical Time Series ClassificationState-space sequence models (S4, Mamba) → Physiological signal modeling
Application of a Mixture of Experts-based Foundation Model to the GlueX DIRC DetectorMixture-of-experts → Efficient MoE training and inference systems
Audio2Tool: Speak, Call, Act -- A Dataset for Benchmarking Speech Tool UseLLM agents and tool use → General LLM agent frameworks and benchmarks
Complete Cyclic Subtask Graphs for Tool-Using LLM Agents: Flexibility, Cost, and Bottlenecks in Long-Horizon WorkflowsLLM agents and tool use → Software engineering coding agents
Know When to Trust the Skill: Delayed Appraisal and Epistemic Vigilance for Single-Agent LLMsLLM reasoning → Reasoning for agents, planning and code +1
Don't Start What You Can't Finish: A Counterfactual Audit of Support-State Triage in LLM AgentsLLM agents and tool use → Software engineering coding agents
Why Training-Free Token Reduction Collapses: The Inherent Instability of Pairwise Scoring SignalsEfficient LLM inference and serving → Token pruning and merging for ViTs
When Agents Go Quiet: Output Generation Capacity and Format-Cost Separation for LLM Document SynthesisLLM agents and tool use → Software engineering coding agents
Reducing Peak Memory Usage for Modern Multimodal Large Language Model PipelinesMultimodal LLMs / vision-language models → Visual token pruning and efficiency
Active World-Model with 4D-informed Retrieval for Exploration and AwarenessWorld models → World models for language agents
Multitask LQG Control: Performance and Generalization BoundsDeep reinforcement learning methods → RL for games and combinatorial optimization
Agentic Large Language Models for Training-Free Neuro-Radiological Image AnalysisLLM agents and tool use → Medical and scientific domain agents
Debate as Reward: A Multi-Agent Reward System for Scientific Ideation via RL Post-TrainingLLM agents and tool use → Reinforcement learning for LLM agents
Auditing Automated Evaluation, Error Propagation, and Runtime Mitigation in Tool-Using Language AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Autonomous Vehicle Collision Avoidance With Racing Parameterized Deep Reinforcement LearningEnd-to-end autonomous driving → Reinforcement learning for driving +1
RankGuide: Tensor-Rank-Guided Routing and Steering for Efficient ReasoningLLM reasoning → Test-time compute and efficient thinking
Agentic Risk-Aware Set-Based Engineering DesignLLM agents and tool use → Multimodal and embodied agents
DARLING: Detection Augmented Reinforcement Learning with Non-Stationary GuaranteesDeep reinforcement learning methods → RL theory and sample complexity
Rewind-IL: Online Failure Detection and State Respawning for Imitation LearningOffline RL and imitation learning from data → Language-conditioned policies and VLAs
KAIROS: Stateful, Context-Aware Power-Efficient Agentic Inference ServingLLM agents and tool use → Software engineering coding agents
ReconVLA: An Uncertainty-Guided and Failure-Aware Vision-Language-Action Framework for Robotic ControlVision-language-action models → Failure recovery and robustness
Benign Fine-Tuning Breaks Safety Alignment in Audio LLMsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
FRIGID: Scaling Diffusion-Based Molecular Generation from Mass Spectra at Training and Inference TimeDiffusion models → Flow matching, time series and science
Agentic Frameworks for Reasoning Tasks: An Empirical StudyLLM agents and tool use → General LLM agent frameworks and benchmarks
AVRT: Audio-Visual Reasoning Transfer through Single-Modality TeachersLLM reasoning → Audio-language model reasoning
SafeLM: Unified Privacy-Aware Optimization for Trustworthy Federated Large Language ModelsLLM safety and security → Membership inference attacks
Revisiting a Pain in the Neck: A Semantic Reasoning Benchmark for Language ModelsLLM reasoning → Reasoning benchmarks and evaluation
Human Cognition in Machines: A Unified Perspective of World ModelsWorld models → World models for language agents
Randomized Antipodal Search Done Right for Data Pareto Improvement of LLM UnlearningLLM safety and security → Machine unlearning
Real-Time Visual Attribution Streaming in Thinking ModelLLM reasoning → Multimodal reasoning
The Global Neural World Model: Spatially Grounded Discrete Topologies for Action-Conditioned PlanningWorld models → Latent predictive world models
POLAR: Online Learning for LoRA Adapter Caching and Routing in Edge LLM ServingEfficient LLM inference and serving → Serving systems and request scheduling
Agentic AI for Education: A Unified Multi-Agent Framework for Personalized Learning and Institutional IntelligenceLLM agents and tool use → Educational agents and math problem solving
Reasoning on the Manifold: Bidirectional Consistency for Self-Verification in Diffusion Language ModelsLLM reasoning → Diffusion reasoning and test-time scaling +1
S-GRPO: Unified Post-Training for Large Vision-Language ModelsLLM reasoning → Multimodal reasoning +1
LLM as a Tool, Not an Agent: Code-Mined Tree Transformations for Neural Architecture SearchLLM agents and tool use → General LLM agent frameworks and benchmarks
A Survey on Long-Term Memory Security in LLM Agents: Attacks, Defenses, and Governance Across the Memory LifecycleLLM agents and tool use → Agent threats, guardrails and governance
Conjunctive Prompt Attacks in Multi-Agent LLM SystemsLLM agents and tool use → Agent threats, guardrails and governance +1
TWGuard: A Case Study of LLM Safety Guardrails for Localized Linguistic ContextsLLM safety and security → Safety alignment, refusal and guardrails
Learning to Reason with Insight for Informal Theorem ProvingLLM reasoning → Mathematical reasoning
From Benchmarking to Reasoning: A Dual-Aspect, Large-Scale Evaluation of LLMs on Vietnamese Legal TextLLM reasoning → Reasoning for agents, planning and code
Hero-Mamba: Mamba-based Dual Domain Learning for Underwater Image EnhancementState-space sequence models (S4, Mamba) → Mamba for image restoration
Information Router for Mitigating Modality Dominance in Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Semantic Area Graph Reasoning for Multi-Robot Language-Guided SearchLLM reasoning → Spatial and embodied reasoning
Beyond Distribution Sharpening: The Importance of Task RewardsLLM reasoning → Reinforcement learning for reasoning +1
Do Vision-Language Models Truly Perform Vision Reasoning? A Rigorous Study of the Modality GapLLM reasoning → Multimodal reasoning +1
Where Do Vision-Language Models Fail? World Scale Analysis for Image GeolocalizationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Find, Fix, Reason: Context Repair for Video ReasoningMultimodal LLMs / vision-language models → Video understanding with MLLMs
Detecting and Suppressing Reward Hacking with Gradient FingerprintsLLM alignment / RLHF / preference optimisation → Reward hacking and misspecification
Optimizing Korean-Centric LLMs via Token PruningEfficient LLM inference and serving → KV cache optimization
Beyond Surface Statistics: Robust Conformal Prediction for LLMs via Internal RepresentationsLLM reasoning → Test-time compute and efficient thinking
ChemGraph-XANES: An Agentic Framework for XANES Simulation and CurationLLM agents and tool use → Medical and scientific domain agents
MARCH: Multi-Agent Radiology Clinical Hierarchy for CT Report GenerationLLM agents and tool use → Medical and scientific domain agents
MOMENTA: Mixture-of-Experts Over Multimodal Embeddings with Neural Temporal Aggregation for Misinformation DetectionMixture-of-experts → Multimodal MoE models
AtManRL: Towards Faithful Reasoning via Differentiable Attention SaliencyLLM reasoning → Multimodal reasoning
Motion-Adapter: A Diffusion Model Adapter for Text-to-Motion Generation of Compound ActionsDiffusion models → Video generation
The Amazing Stability of Flow MatchingDiffusion models → Flow matching, time series and science
AEGIS: Anchor-Enforced Gradient Isolation for Knowledge-Preserving Vision-Language-Action Fine-TuningMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +2
Chain-of-Thought Degrades Visual Spatial Reasoning Capabilities of Multimodal LLMsLLM reasoning → Spatial and embodied reasoning +1
Mind's Eye: A Benchmark of Visual Abstraction, Transformation and Composition for Multimodal LLMsLLM reasoning → Multimodal reasoning +1
Elucidating the SNR-t Bias of Diffusion Probabilistic ModelsDiffusion models → Diffusion foundations and sampling
Towards Intrinsic Interpretability of Large Language Models:A Survey of Design Principles and ArchitecturesMechanistic interpretability → Knowledge and representations inside LLMs
Safe Deep Reinforcement Learning for Building Heating Control and Demand-side FlexibilityDeep reinforcement learning methods → RL for energy, networks and operations
Cut Your Losses! Learning to Prune Paths Early for Efficient Parallel ReasoningLLM reasoning → Test-time compute and efficient thinking
AstroVLM: Expert Multi-agent Collaborative Reasoning for Astronomical Imaging Quality DiagnosisLLM reasoning → Knowledge- and tool-augmented reasoning +2
SocialGrid: A Benchmark for Planning and Social Reasoning in Embodied Multi-Agent SystemsLLM reasoning → Reasoning for agents, planning and code +1
AgentV-RL: Scaling Reward Modeling with Agentic VerifierLLM reasoning → Reasoning for agents, planning and code +2
ReactBench: A Benchmark for Topological Reasoning in MLLMs on Chemical Reaction DiagramsLLM reasoning → Clinical and biomedical applications +1
Weak-Link Optimization for Multi-Agent Reasoning and CollaborationLLM reasoning → Reasoning for agents, planning and code +1
When Safe Concepts Become Unsafe: Multi-Concept Compositional Vulnerabilities in Text-to-Image ModelsLLM safety and security → Vision-language and text-to-image safety
A Case Study on the Impact of Anonymization Along the RAG PipelineRetrieval-augmented generation → Privacy and security of RAG
Integrating Graphs, Large Language Models, and Agents: Reasoning and RetrievalLLM reasoning → Knowledge- and tool-augmented reasoning +1
RAGognizer: Hallucination-Aware Fine-Tuning via Detection Head IntegrationRetrieval-augmented generation → Hallucination detection in RAG
VADF: Vision-Adaptive Diffusion Policy Framework for Efficient Robotic ManipulationDiffusion models → Diffusion and flow policies for robotics
Making Image Editing Easier via Adaptive Task Reformulation with Agentic ExecutionsLLM agents and tool use → Multimodal and embodied agents
Efficient Video Diffusion Models: Advancements and ChallengesDiffusion models → Video generation
Experience Compression Spectrum: Unifying Memory, Skills, and Rules in LLM AgentsLLM agents and tool use → Agent memory systems
DPrivBench: Benchmarking LLMs' Reasoning for Differential PrivacyLLM reasoning → Reasoning for agents, planning and code
CiPO: Counterfactual Unlearning for Large Reasoning Models through Iterative Preference OptimizationLLM reasoning → Reasoning for agents, planning and code +2
Disentangling Mathematical Reasoning in LLMs: A Methodological Investigation of Internal MechanismsLLM reasoning → Mathematical reasoning
CoEvolve: Training LLM Agents via Agent-Data Mutual EvolutionLLM agents and tool use → Reinforcement learning for LLM agents
Discover and Prove: An Open-source Agentic Framework for Hard Mode Automated Theorem Proving in Lean 4LLM agents and tool use → Educational agents and math problem solving
Placing Puzzle Pieces Where They Matter: A Question Augmentation Framework for Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning +1
Breaking the Training Barrier of Billion-Parameter Universal Machine Learning Interatomic PotentialsMixture-of-experts → MoE theory and gating functions
Aligning What Vision-Language Models See and Perceive with Adaptive Information FlowMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Beyond a Single Frame: Multi-Frame Spatially Grounded Reasoning Across Volumetric MRILLM reasoning → Clinical and biomedical applications +1
Qwen3.5-Omni Technical ReportMixture-of-experts → Multimodal MoE models
CHOP: Chunkwise Context-Preserving Framework for RAG on Multi DocumentsRetrieval-augmented generation → Multimodal RAG
Self-Distillation as a Performance Recovery Mechanism for LLMs: Counteracting Compression and Catastrophic ForgettingKnowledge distillation → Knowledge distillation for LLMs
Pruning Unsafe Tickets: A Resource-Efficient Framework for Safer and More Robust LLMsLLM safety and security → Jailbreaking
MemEvoBench: Benchmarking Safety Risks from Memory Misevolution in LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance

All 393