Papers

New on arXiv

Wed 29 Apr 2026

367 papers

Quantifying Prior Dominance in RAG SystemsRetrieval-augmented generation → RAG robustness and adaptive retrieval
Bidirectional Semantic Complementary Tool Retrieval for Remote Sensing AgentsLLM agents and tool use → Multimodal and embodied agents
Query-Adaptive Semantic Chunking for Retrieval-Augmented Generation: A Dynamic Strategy with Contextual Window ExpansionRetrieval-augmented generation → Multimodal RAG
Autonomous LLM Agents & CTFs: A Second LookLLM agents and tool use → Agent threats, guardrails and governance
A Reproducibility Analysis of PO4ISR: Diagnosing and Mitigating Semantic Drift in LLM-Based Session RecommendationLLM reasoning → LLM reasoning for recommendation
Scaling Mobile Agent Systems: From Capability Density to Collective IntelligenceLLM agents and tool use → General LLM agent frameworks and benchmarks
Decision Evidence Maturity Model for Agentic AI: A Property-Level Method SpecificationLLM agents and tool use → Agent threats, guardrails and governance
The Cost of Consensus: Isolated Self-Correction Prevails Over Unguided Homogeneous Multi-Agent DebateLLM agents and tool use → Social simulation and multi-agent debate
When Good OCR Is Not Enough: Benchmarking OCR Robustness for Retrieval-Augmented GenerationRetrieval-augmented generation → Multimodal RAG
Generalized Category Discovery under Domain Shifts: From Vision to Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
SiriusHelper: An LLM Agent-Based Operations Assistant for Big Data PlatformsLLM agents and tool use → Web, search, and research agents
AdvDMD: Adversarial Reward Meets DMD For High-Quality Few-Step GenerationDiffusion models → Diffusion foundations and sampling +1
Frame Entrepreneurs in an AI Agent Community: Concentrated Identity-Claim Production on MoltbookLLM agents and tool use → Social simulation and multi-agent debate
AutoREC: A reinforcement learning platform for equivalent circuit model generationDeep reinforcement learning methods → Interpretable and logic-guided RL
Self-Evolving Software AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks +1
AutoSurfer -- Teaching Web Agents through Comprehensive Surfing, Learning, and ModelingLLM agents and tool use → Web, search, and research agents
Compliance vs. Sensibility: On the Reasoning Controllability in Large Language ModelsLLM reasoning → Reasoning benchmarks and evaluation
Addressing the Reality Gap: A Three-Tension Framework for Agentic AI AdoptionLLM agents and tool use → Educational agents and math problem solving
RAQG-QPP: Query Performance Prediction with Retrieved Query Variants and Retrieval Augmented Query GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
SafeTune: Mitigating Data Poisoning in LLM Fine-Tuning for RTL Code GenerationLLM safety and security → Safety alignment, refusal and guardrails
Reinforced Agent: Inference-Time Feedback for Tool-Calling AgentsLLM agents and tool use → Tool learning and function calling
When Roles Fail: Epistemic Constraints on Advocate Role Fidelity in LLM-Based Political Statement AnalysisLLM agents and tool use → Social simulation and multi-agent debate
Web2BigTable: A Bi-Level Multi-Agent LLM System for Internet-Scale Information Search and ExtractionLLM agents and tool use → Web, search, and research agents
Indirect Prompt Injection in the Wild: An Empirical Study of Prevalence, Techniques, and ObjectivesLLM safety and security → Prompt injection and agent security
Path-Lock Expert: Separating Reasoning Mode in Hybrid Thinking via Architecture-Level SeparationLLM reasoning → Test-time compute and efficient thinking
Energy-Efficient Plant Monitoring via Knowledge DistillationKnowledge distillation → Detection and dense prediction distillation
How a Cooperative-Override Circuit Suppresses Nash Play in Large Language ModelsMechanistic interpretability → Activation steering methods
Distributional Alignment Games for Answer-Level Fine-TuningLLM reasoning → Reasoning for agents, planning and code
Step-level Optimization for Efficient Computer-use AgentsLLM agents and tool use → GUI and mobile agents
How to Guide Your Flow: Few-Step Alignment via Flow Map Reward GuidanceLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Enhancing Linux Privilege Escalation Attack Capabilities of Local LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
TRUST: A Framework for Decentralized AI Service v.0.1LLM reasoning → Reasoning for agents, planning and code
PALCAS: A Priority-Aware Intelligent Lane Change Advisory System for Autonomous Vehicles using Federated Reinforcement LearningEnd-to-end autonomous driving → Cooperative and V2X driving +1
Think it, Run it: Autonomous ML pipeline generation via self-healing multi-agent AILLM reasoning → LLM reasoning for recommendation +1
Useless but Safe? Benchmarking Utility Recovery with User Intent Clarification in Multi-Turn ConversationsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Detecting Clinical Discrepancies in Health Coaching Agents: A Dual-Stream Memory and Reconciliation ArchitectureLLM agents and tool use → Medical and scientific domain agents
Dynamic Adversarial Fine-Tuning Reorganizes Refusal GeometryLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
When Continual Learning Moves to Memory: A Study of Experience Reuse in LLM AgentsLLM agents and tool use → Agent memory systems
Membership Inference Attacks Against Video Large Language ModelsLLM safety and security → Membership inference attacks
Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language ModelsEfficient LLM inference and serving → Diffusion LLM and MoE inference
Three-Step Nav: A Hierarchical Global-Local Planner for Zero-Shot Vision-and-Language NavigationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Select to Think: Unlocking SLM Potential with Local SufficiencyLLM reasoning → Test-time compute and efficient thinking
World2VLM: Distilling World Model Imagination into VLMs for Dynamic Spatial ReasoningWorld models → World models for language agents +2
ClawGym: A Scalable Framework for Building Effective Claw AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Hot Fixing in the WildLLM agents and tool use → Software engineering coding agents
FaaSMoE: A Serverless Framework for Multi-Tenant Mixture-of-Experts ServingMixture-of-experts → Efficient MoE training and inference systems
MoRFI: Monotonic Sparse Autoencoder Feature IdentificationMechanistic interpretability → SAE features and their applications
Edge AI for Automotive Vulnerable Road User Safety: Deployable Detection via Knowledge DistillationKnowledge distillation → Detection and dense prediction distillation
STARRY: Spatial-Temporal Action-Centric World Modeling for Robotic ManipulationWorld models → Embodied and robot world models +1
Language Diffusion Models are Associative Memories Capable of Retrieving Unseen DataDiffusion models → Diffusion foundations and sampling
Unifying Sparse Attention with Hierarchical Memory for Scalable Long-Context LLM ServingEfficient LLM inference and serving → KV cache optimization
Uncertainty-Aware Predictive Safety Filters for Probabilistic Neural Network DynamicsDeep reinforcement learning methods → Safe and constrained RL
Rule-based High-Level Coaching for Goal-Conditioned Reinforcement Learning in Search-and-Rescue UAV Missions Under Limited-Simulation TrainingDeep reinforcement learning methods → RL for robot navigation and UAV control
Asynchronous Federated Unlearning with Invariance Calibration for Medical ImagingLLM safety and security → Federated unlearning and poisoning
Bian Que: An Agentic Framework with Flexible Skill Arrangement for Online System OperationsLLM agents and tool use → General LLM agent frameworks and benchmarks
Accelerating RL Post-Training Rollouts via System-Integrated Speculative DecodingEfficient LLM inference and serving → Speculative decoding
Decoupling Knowledge and Task Subspaces for Composable Parametric Retrieval Augmented GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
Factorized Latent Reasoning for LLM-based RecommendationLLM reasoning → LLM reasoning for recommendation
GLM-5V-Turbo: Toward a Native Foundation Model for Multimodal AgentsLLM agents and tool use → Multimodal and embodied agents
FutureWorld: A Live Reinforcement Learning Environment for Predictive Agents with Real-World Outcome RewardsLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Unified 4D World Action Modeling from Video Priors with Asynchronous DenoisingWorld models → Embodied and robot world models +1
AgentSim: A Platform for Verifiable Agent-Trace SimulationLLM reasoning → Reasoning for agents, planning and code +1
When to Retrieve During Reasoning: Adaptive Retrieval for Large Reasoning ModelsLLM reasoning → Knowledge- and tool-augmented reasoning +1
SciHorizon-DataEVA: An Agentic System for AI-Readiness Evaluation of Heterogeneous Scientific DataLLM agents and tool use → Web, search, and research agents
When to Vote, When to Rewrite: Disagreement-Guided Strategy Routing for Test-Time ScalingLLM reasoning → Test-time compute and efficient thinking
OCR-Memory: Optical Context Retrieval for Long-Horizon Agent MemoryLLM agents and tool use → Multimodal and embodied agents
TDD Governance for Multi-Agent Code Generation via Prompt EngineeringLLM agents and tool use → Software engineering coding agents
State Beyond Appearance: Diagnosing and Improving State Consistency in Dial-Based Measurement ReadingMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
PAINT: Partial-Solution Adaptive Interpolated Training for Self-Distilled ReasonersLLM reasoning → Prompting and chain-of-thought reasoning +1
Learning to Route Electric Trucks Under Operational UncertaintyDeep reinforcement learning methods → RL for energy, networks and operations
Preserving Disagreement: Architectural Heterogeneity and Coherence Validation in Multi-Agent Policy SimulationLLM agents and tool use → Social simulation and multi-agent debate
DUAL-BLADE: Dual-Path NVMe-Direct KV-Cache Offloading for Edge LLM InferenceEfficient LLM inference and serving → KV cache optimization
PRAG: End-to-End Privacy-Preserving Retrieval-Augmented GenerationRetrieval-augmented generation → Privacy and security of RAG
AGEL-Comp: A Neuro-Symbolic Framework for Compositional Generalization in Interactive AgentsWorld models → World models for language agents +1
GIFGuard: Proactive Forensics against Deepfakes in Facial GIFs via Spatiotemporal WatermarkingLLM safety and security → Watermarking generated content and models
Lyapunov-Guided Self-Alignment: Test-Time Adaptation for Offline Safe Reinforcement LearningOffline RL and imitation learning from data → Safe offline RL
Tatemae: Detecting Alignment Faking via Tool Selection in LLMsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
Progressive Semantic Communication for Efficient Edge-Cloud Vision-Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
Auto-Relational ReasoningLLM reasoning → Reasoning benchmarks and evaluation
SafeReview: Defending LLM-based Review Systems Against Adversarial Hidden PromptsLLM safety and security → Prompt injection and agent security
Delta Score Matters! Spatial Adaptive Multi Guidance in Diffusion ModelsDiffusion models → Diffusion foundations and sampling
PKS4:Parallel Kinematic Selective State Space Scanners for Efficient Video UnderstandingState-space sequence models (S4, Mamba) → Video, motion and generative Mamba models
Delineating Knowledge Boundaries for Honest Large Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
When Hidden States Drift: Can KV Caches Rescue Long-Range Speculative Decoding?Efficient LLM inference and serving → Speculative decoding
Sparsity as a Key: Unlocking New Insights from Latent Structures for Out-of-Distribution DetectionMechanistic interpretability → Sparse autoencoder training and evaluation
SecMate: Multi-Agent Adaptive Cybersecurity Troubleshooting with Tri-Context PersonalizationLLM agents and tool use → Agent threats, guardrails and governance
Topology-Aware Representation Alignment for Semi-Supervised Vision-Language LearningMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Probabilistic data quality assessment for structural monitoring data via outlier-resistant conditional diffusion modelDiffusion models → Flow matching, time series and science
Beyond Fixed Formulas: Data-Driven Linear Predictor for Efficient Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Uncertainty-Aware Reward Discounting for Mitigating Reward HackingLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Shorthand for Thought: Compressing LLM Reasoning via Entropy-Guided SupertokensLLM reasoning → Test-time compute and efficient thinking
UIGaze: How Closely Can VLMs Approximate Human Visual Attention on User Interfaces?Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
ACPO: Anchor-Constrained Perceptual Optimization for Diffusion Models with No-Reference Quality GuidanceDiffusion models → Diffusion foundations and sampling
Adaptive and Fine-grained Module-wise Expert Pruning for Efficient LoRA-MoE Fine-TuningMixture-of-experts → Mixture of LoRA experts
Efficient, VRAM-Constrained xLM Inference on ClientsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Addressing Performance Saturation for LLM RL via Precise Entropy Curve ControlLLM reasoning → Reinforcement learning for reasoning
A Systematic Comparison of Prompting and Multi-Agent Methods for LLM-based Stance DetectionLLM agents and tool use → Social simulation and multi-agent debate
DreamProver: Evolving Transferable Lemma Libraries via a Wake-Sleep Theorem-Proving AgentLLM agents and tool use → Educational agents and math problem solving
Folding Tensor and Sequence Parallelism for Memory-Efficient Transformer Training & InferenceMixture-of-experts → Expert routing strategies
CheXthought: A global multimodal dataset of clinical chain-of-thought reasoning and visual attention for chest X-ray interpretationLLM reasoning → Clinical and biomedical applications +1

All 367