Papers

New on arXiv

Wed 17 Jun 2026

462 papers

From SQL to Knowledge Graphs: An LLM-Driven Multi-Agent Approach with Data Schema ImprovementLLM agents and tool use → Web, search, and research agents
Reviewing Model Collapse and CountermeasuresSynthetic data and self-improvement for LLMs → Model collapse from synthetic data
Multilingual Verifier Bias in RLVR: Benchmark, Rollout Diagnosis, and the Cross-Lingual Selection BottleneckLLM reasoning → Reinforcement learning for reasoning
Self-Speculation for Faster Reasoning ModelsLLM reasoning → Test-time compute and efficient thinking +1
Can Large Language Models Reason about Event-Time Stream-Processing Semantics?LLM reasoning → Causal reasoning in LLMs
Human-Simulation Interaction: From Prediction to Exploration in LLM Agent Simulations for PolicyLLM agents and tool use → Social simulation and multi-agent debate
HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online AdvertisingLLM agents and tool use → Financial trading and forecasting agents +1
TokenMem: Faithful Knowledge Injection for Frozen LLMsRetrieval-augmented generation → RAG robustness and adaptive retrieval
Learning When to Reason for Text-to-SQL via SFT and DPOLLM reasoning → Knowledge- and tool-augmented reasoning +1
MawForge: Memory-Bounded Expert Materialization for Local Mixture-of-Experts InferenceMixture-of-experts → Efficient MoE training and inference systems +1
Recursive Self-Evolving Agents via Held-Out SelectionLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Self-Improvement Can Self-Regress: The Rise-and-Collapse Failure Mode of LLM Self-TrainingLLM reasoning → Reinforcement learning for reasoning
Repeated post-training is not Self-improving: Diagnosing Scientific Amnesia in Continual DPO PipelinesLLM alignment / RLHF / preference optimisation → LLM preference optimization methods +1
Amplify, Don't Create: Temporal Accumulation for Slow-Burn Prompt InjectionLLM safety and security → Prompt injection and agent security
VeriBound: PAC-Bayesian Generalization Bounds for Process Reward Models Trained with Formal Verification ToolsLLM reasoning → Reinforcement learning for reasoning +1
An approach with Visual and Tabular Mamba to multimodal medical data using Mixed FusionState-space sequence models (S4, Mamba) → Mamba for medical image analysis
A Layered Security Framework Against Prompt Injection in RAG-Based ChatbotsLLM safety and security → Prompt injection and agent security
SAGE-OPD: Selective Agent-Guided Intervention for Multi-Turn On-Policy DistillationLLM agents and tool use → Reinforcement learning for LLM agents +1
BrainG3N: A Dual-Purpose Tokenizer for Controllable 3D Brain MRI GenerationDiffusion models → Medical imaging and segmentation
SAFE-Cascade: Cost-Adaptive Vision-Language Routing for Chart Question AnsweringMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Scaling Self-Play for End-to-End DrivingEnd-to-end autonomous driving → Imitation-learning end-to-end planners
CTS-MoE: Implicit Terrain Adaptation via Mixture-of-Experts for Perceptive LocomotionMixture-of-experts → MoE in reinforcement learning and robotics
Before the Pull Request: Mining Multi-Agent CoordinationLLM agents and tool use → Software engineering coding agents
StaminaBench: Stress-Testing Coding Agents over 100 Interaction TurnsLLM agents and tool use → Software engineering coding agents
Which Pairs to Compare for LLM Post-Training?LLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Comparing Linear Probes with Mahalanobis Cosine SimilarityMechanistic interpretability → Probing deception, bias and beliefs
Configurable Clinical Information Extraction with Agentic RAG: What Works, What Breaks, and WhyRetrieval-augmented generation → Medical and biomedical RAG +1
IHBench: Evaluating Post-Interruption Recovery in Voice Agents with Structured WorkflowsLLM agents and tool use → General LLM agent frameworks and benchmarks
Analyzing the Narration Gap in LLM-Solver LoopsLLM reasoning → Reasoning benchmarks and evaluation
Mix-QVLA: Task-Evidence-Aware Mixed-Precision Quantization of Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation
Uncertainty Decomposition for Clarification Seeking in LLM AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
FloatDoor: Platform-Triggered Backdoors in LLMsLLM safety and security → Backdoor and data poisoning attacks
PerceptionDLM: Parallel Region Perception with Multimodal Diffusion Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?World models → Embodied and robot world models
Emergent AlignmentLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
DeXposure-Claw: An Agentic System for DeFi Risk SupervisionLLM agents and tool use → Financial trading and forecasting agents
Hidden Anchors in Multi-Agent LLM DeliberationLLM agents and tool use → Social simulation and multi-agent debate
Concept Flow Models: Anchoring Concept-Based Reasoning with Hierarchical BottlenecksLLM reasoning → Multimodal reasoning
Dyna-DINO: Efficient ViT Distillation Via Adaptive Representation AnchoringKnowledge distillation → General knowledge distillation methods
Insulin4RL: Real-Time Insulin Management in the Intensive Care Unit for Offline Reinforcement LearningOffline RL and imitation learning from data → Safe offline RL
Can In-Context Learning Support Intrinsic Curiosity?World models → Latent dynamics models and model-based RL
Diffusion Language Models: An Experimental AnalysisDiffusion models → Discrete and text diffusion
Deontic Policies for Runtime Governance of Agentic AI SystemsLLM agents and tool use → Agent threats, guardrails and governance
Scaling Generative Foundation Models for Chest Radiography with Rectified Flow TransformersDiffusion models → Medical imaging and segmentation
Playful Agentic Robot LearningLLM agents and tool use → Multimodal and embodied agents
MortarBench: Evaluating Mortgage Loan Origination AgentsLLM agents and tool use → Financial trading and forecasting agents
Thermodynamic Signatures of Reasoning: Free-Energy and Spectral-Form-Factor Diagnostics for Hallucination Detection in Large Language ModelsLLM reasoning → Hallucination detection and mitigation
DiffusionVS: A Generative Framework for Robust Visual Servoing Based on Diffusion PolicyDiffusion models → Diffusion and flow policies for robotics
ZeroDex: Zero-Shot Long-Horizon Dexterous Manipulation via Multi-View 3D-Grounded VLM ReasoningLLM reasoning → Spatial and embodied reasoning
Beyond the Current Observation: Evaluating Multimodal Large Language Models in Controllable Non-Markov GamesMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
UBP2: Uncertainty-Balanced Preference Planning for Efficient Preference-based Reinforcement LearningLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
Rethinking Reward Supervision: Rubric-Conditioned Self-DistillationLLM reasoning → Reinforcement learning for reasoning +1
Data Intelligence Agents: Interpreting, Modeling, and Querying Enterprise Data via Autonomous Coding AgentsLLM agents and tool use → Web, search, and research agents
Diffusion-Proof: Recipe for Formal Theorem Proving Beyond Auto-Regressive GenerationLLM reasoning → Mathematical reasoning
Enhancing Decision-Making with Large Language Models through Multi-Agent Fictitious PlayLLM agents and tool use → Reinforcement learning for LLM agents
Does VLA Even Know the Basics? Measuring Commonsense and World Knowledge Retention in Vision-Language-Action ModelsVision-language-action models → Reasoning and chain-of-thought in VLA
TextRich: A Multi-Domain Benchmark for Detecting AI-Generated Text-Rich Images from GPT-Image-2Multimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
CABLE: Cloud-Assisted Bandwidth-efficient LMM-based Encoding for V2X SystemsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
DreamReasoner-8B: Block-Size Curriculum Learning for Diffusion Reasoning ModelsLLM reasoning → Diffusion reasoning and test-time scaling +1
OneCanvas: 3D Scene Understanding via Panoramic ReprojectionLLM reasoning → Spatial and embodied reasoning +1
TxBench-PP: Analyzing AI Agent Performance on Small-Molecule Preclinical PharmacologyLLM agents and tool use → Medical and scientific domain agents
Runtime Compliance Verification for AI AgentsLLM safety and security → Prompt injection and agent security
STARE: Surprisal-Guided Token-Level Advantage Reweighting for Policy Entropy StabilityLLM reasoning → Reinforcement learning for reasoning
CodeSentinel: A Three-Layer Defense Against Indirect Prompt Injection in Code ContextsLLM safety and security → Prompt injection and agent security
Mechanism-Guided Selective Unlearning for RLVR-Induced ReasoningLLM safety and security → Machine unlearning
Machine Unlearning for the XGBoost Model with Network Intrusion DatasetsLLM safety and security → Machine unlearning
DASH: Dual-View Self-Distillation with Multi-Layer Hidden Representations for Robust Speech RecognitionKnowledge distillation → Speech and audio distillation
Invertible Neural Network Adapter for One-Step Flow Matching in Robot ManipulationDiffusion models → Diffusion and flow policies for robotics
PhantomSkill: Malicious Code Injection in Agent Skill EcosystemsLLM agents and tool use → Agent threats, guardrails and governance
When AUC Misleads: Polarization-Aware Evaluation of Deepfake Detectors under Domain ShiftDiffusion models → Image generation, text-to-image and editing
User as Engram: Internalizing Per-User Memory as Local Parametric EditsLLM reasoning → Reasoning for agents, planning and code
Beyond Safe Data: Pretraining-Stage Alignment with Regular Safety ReflectionLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RLDiffusion models → Diffusion and flow policies for robotics +2
The Market in the Model: Latent Diffusion as Neural EconomyDiffusion models → Diffusion foundations and sampling
A Technical Taxonomy of LLM Agent Communication ProtocolsLLM agents and tool use → General LLM agent frameworks and benchmarks
Pareto Q-Learning with Reward MachinesDeep reinforcement learning methods → Interpretable and logic-guided RL
Seeing Before Reasoning: Decoupling Perception and Reasoning for Shortcut-Resilient Multimodal On-Policy Self-DistillationLLM reasoning → Multimodal reasoning +1
Leadership as Coordination Control: Behavioral Signatures and the Recovery-Advantage Boundary in Multi-Agent LLM TeamsLLM agents and tool use → General LLM agent frameworks and benchmarks
ReSiReg: Towards Spatially Consistent Semantics in Language-Conditioned Robotic TasksMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Taming I2V models for Image HOI Editing: A Cognitive Benchmark and Agentic Self-Correcting FrameworkLLM agents and tool use → Multimodal and embodied agents
Model-Free Reinforcement Learning Control for Resilient Cyber-Physical SystemsDeep reinforcement learning methods → Safe and constrained RL
DREAM: Extending Vision-Language Models with Dual-Objective Encoding for Cross-Modal RetrievalMultimodal LLMs / vision-language models → Video understanding with MLLMs
Benchmarking Large Vision-Language Models on Fine-Grained Image Tasks: From Evaluation to DiagnosisLLM reasoning → Multimodal reasoning +1
RODS: Reward-Driven Online Data Synthesis for Multi-Turn Tool-Use AgentsSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
Geometric and Stochastic Analysis of Discontinuities in Sparse Mixture-of-ExpertsMixture-of-experts → Expert routing strategies
FoMoE: Breaking the Full-Replica Barrier with a Federation of MoEsMixture-of-experts → Efficient MoE training and inference systems
FlowObject: Flow Steering for Bridging Generative Priors and Reconstruction FidelityDiffusion models → 3D, point cloud and scene generation
Sumi: Open Uniform Diffusion Language Model from ScratchDiffusion models → Discrete and text diffusion
Spotlight: Synergizing Seed Exploration and Spot GPUs for DiT RL Post-TrainingDiffusion models → Diffusion foundations and sampling
Enhancing Multilingual Reasoning via Steerable Model MergingLLM reasoning → Reasoning benchmarks and evaluation
ThinkDeception: A Progressive Reinforcement Learning Framework for Interpretable Multimodal Deception DetectionLLM reasoning → Prompting and chain-of-thought reasoning
CAPRA: Scaling Feedback on Software Architecture Deliverables with a Multi-Agent LLM SystemLLM agents and tool use → Software engineering coding agents
Visual-OPSD: Cross-Modal On-Policy Self-Distillation for Efficient Unified Multimodal ReasoningLLM reasoning → Multimodal reasoning +1
EfficientRollout: System-Aware Self-Speculative Decoding for RL RolloutsLLM reasoning → Reinforcement learning for reasoning +2
Be Your Own Teacher: Steering Protein Language Models via Unsupervised Reward OptimizationLLM alignment / RLHF / preference optimisation → Preference optimization for molecular design
Mem-World: Memory-Augmented Action-Conditioned World Models for Persistent Robot ManipulationWorld models → Embodied and robot world models
Motion-Focused Latent Action Enables Cross-Embodiment VLA Training from Human EgoVideosVision-language-action models → Human-video and latent-action pretraining
GraphPO: Graph-based Policy Optimization for Reasoning ModelsLLM reasoning → Reinforcement learning for reasoning
Object-Centric Residual RL for Zero-Shot Sim-to-Real VLA EnhancementVision-language-action models → Reinforcement learning and post-training
RTSGameBench: An RTS Benchmark for Strategic Reasoning by Vision-Language ModelsLLM reasoning → Reasoning for agents, planning and code +1

All 462