Papers

New on arXiv

Mon 22 Jun 2026

513 papers

TelecomGPT-R1: A Unified Open-Source Reasoner for the Telecom StackLLM reasoning → Reasoning benchmarks and evaluation
Toward Safe LLM Agents: A Survey of Specification, Verification, and EnforcementLLM agents and tool use → Agent threats, guardrails and governance
The Hallucination Snowball: Modeling Error Propagation as State Transitions in Multi-Agent LLM PipelinesLLM agents and tool use → Medical and scientific domain agents
TRACE: Business Rule-Grounded Reasoning Curriculum for Knowledge-Preserving Parametric Tool Retrieval in Enterprise LLMsLLM reasoning → Reasoning for agents, planning and code
YUKTI: From Natural-Language Situations to Robust, Verifiable Decisions An Uncertainty-Typed Proposition IR, Assumption-Robust Pareto Frontiers, and a Regret CertificateLLM agents and tool use → General LLM agent frameworks and benchmarks
Model Collapse: On Recursion, Noise, and Uncharted Machine VisionsSynthetic data and self-improvement for LLMs → Model collapse from synthetic data
Learning Dexterous Manipulation Using Contact Wrench Guidance From Human DemonstrationDeep reinforcement learning methods → RL for robot locomotion and manipulation
Data Provenance for Image Auto-Regressive GenerationLLM safety and security → Watermarking generated content and models
RoboGaze: Evaluating Robot World Models via Structured Vision-Language AnalysisWorld models → Embodied and robot world models
The Hitchhiker's Guide to Agentic AI: From Foundations to SystemsLLM reasoning → Reasoning for agents, planning and code +1
GUI vs. CLI: Execution Bottlenecks in Screen-Only and Skill-Mediated Computer-Use AgentsLLM agents and tool use → GUI and mobile agents
Reinforcement Learning Towards Broadly and Persistently Beneficial ModelsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Towards Spec Learning: Inference-Time Alignment from Preference PairsLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Cyclic Denoising Reveals Ultrastable Memories in Diffusion ModelsDiffusion models → Diffusion foundations and sampling
ChartWalker: Benchmarking the Cross-Chart RAG Task with Hierarchical Knowledge GraphsRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
EMAgnet: Parameter-Space EMA Regularization for Policy Gradient Self-Play in Large GamesDeep reinforcement learning methods → RL for games and combinatorial optimization
Learning to Trigger: Reinforcement Learning at the Large Hadron ColliderDeep reinforcement learning methods → RL for energy, networks and operations
Critique of Agent ModelLLM agents and tool use → General LLM agent frameworks and benchmarks
Offline Reinforcement Learning for Warehouse SLAM Throughput ControlOffline RL and imitation learning from data → Offline RL algorithms
The Serialized Bridge: Understanding and Recovering LLM Serving Performance under Blackwell GPU Confidential ComputingEfficient LLM inference and serving → Serving systems and request scheduling
Forget Without Compromise: Nexus Sampling for Streaming KV-Cache Eviction Under Fixed BudgetsLLM agents and tool use → Software engineering coding agents +1
Neuro-Symbolic Drive: Rule-Grounded Faithful Reasoning for Driving VLAsLLM reasoning → Chain-of-thought reasoning for driving +1
Flow-Corrected Thompson Sampling for Non-Stationary Contextual BanditsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
KLip-PPO: A per-sample KL perspective on PPO-ClipDeep reinforcement learning methods → RL theory and sample complexity
RIFT-Bench: Dynamic Red-teaming For Agentic AI SystemsLLM agents and tool use → Agent threats, guardrails and governance +1
Catastrophic Compositional Generation: Why Vanilla Diffusion Models Fail to ExtrapolateDiffusion models → Diffusion foundations and sampling
Do LLM Attribution Metrics Transfer? Auditing Retrieval-Augmented Generation Evaluation Across Datasets and ConstructsRetrieval-augmented generation → RAG robustness and adaptive retrieval
Closing the Loop: Formally Verified Law as a Reward Signal for Self-Improving Legal AISynthetic data and self-improvement for LLMs → Self-play RL for reasoning
ARIA: Adaptive Region-Based Importance Allocation for Conditional Diffusion DistillationDiffusion models → Diffusion foundations and sampling
The Professor: Multi-Teacher Unsupervised Prompt Distillation for Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
REALM: A Unified Red-Teaming Benchmark for Physical-World VLMsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
E-MRL: Cross-view Aligned Evidence-driven Multimodal Reinforcement Learning for Reliable 3D Tumor AnalysisMultimodal LLMs / vision-language models → Medical vision-language models +1
Mind the Heads: Topological Representation Alignment for Multimodal LLMsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
One Year Later...The Harms Persist, But So Do We!LLM safety and security → Safety alignment, refusal and guardrails
MGI: Member vs Generated InferenceLLM safety and security → Membership inference attacks
HANCLIP: A Family of Hyperbolic Angular Negation Vision Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
ABACUS: Adapting Unified Foundation Model for Bridging Image Count Understanding and GenerationMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Cryptographic certificates of validity for trustworthy AILLM agents and tool use → Agent threats, guardrails and governance
Emergent Relational Order in LLM Agent Societies: From Collective Affect to Authority StratificationLLM agents and tool use → Social simulation and multi-agent debate
ESAA-Conversational: An Event-Sourced Memory Layer for Continuity, Handoff, and Curation Across Heterogeneous LLM Coding AgentsLLM agents and tool use → Agent memory systems
Randomized YaRN Improves Length Generalization for Long-Context ReasoningLLM reasoning → Test-time compute and efficient thinking
LIBERO-Safety: A Comprehensive Benchmark for Physical and Semantic Safety in Vision-Language-Action ModelsVision-language-action models → Safety and adversarial attacks
Keep The Essentials: Efficient Reference Conditioned Generation via Token DroppingDiffusion models → Diffusion foundations and sampling +1
AIR: Adaptive Interleaved Reasoning with Code in MLLMsLLM reasoning → Multimodal reasoning +1
MAS-PromptBench: When Does Prompt Optimization Improve Multi-Agent LLM Systems?LLM agents and tool use → General LLM agent frameworks and benchmarks
EnterpriseClawBench: Benchmarking Agents from Real Workplace SessionsLLM agents and tool use → General LLM agent frameworks and benchmarks
Flatness Preserves Instruction Following in Vision-Language-Action ModelsVision-language-action models → Generalist VLA models and benchmarks
Learning Process Rewards via Success Visitation Matching for Efficient RLDeep reinforcement learning methods → Deep RL algorithms and representations
Diffusion Models Adapt to Low-Dimensional Structure Under Flexible Coefficient ChoicesDiffusion models → Diffusion foundations and sampling
DiT-Reward: Generative Representations for Text-to-Image Reward ModelingDiffusion models → Diffusion and flow policies for robotics +1
Learning to See While Learning to Act: Diffusion Models for Active Perception in Robot ImitationDiffusion models → Diffusion and flow policies for robotics +1
dVLA-RL: Reinforcement Learning over Denoising Trajectories for Discrete Diffusion Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation +2
RECALL: Recovery Experience Collection for Active Lifelong Learning in Vision-Language-Action ModelsVision-language-action models → Reinforcement learning and post-training +1
Vera: A Layered Diffusion Model for Content-Preserving Video EditingDiffusion models → Video generation
Causal Discovery in the Era of AgentsLLM agents and tool use → Medical and scientific domain agents
SPIRAL: Learning to Search and AggregateLLM reasoning → Test-time compute and efficient thinking
KEMO: Event-Driven Keyframe Memory for Long-Horizon Robot Manipulation with VLA PoliciesVision-language-action models → Reasoning and chain-of-thought in VLA
Kamera: Unified Position-Invariant Multimodal KV Cache for Training-Free ReuseEfficient LLM inference and serving → Visual token pruning in multimodal models
A Watermark for Vision-Language-Action and World Action ModelsWorld models → Embodied and robot world models +2
Scheduling Thoughts: Learning the Order of Thought in Diffusion Language ModelsLLM reasoning → Diffusion reasoning and test-time scaling +1
HoloAgent-0: A Unified Embodied Agent Framework with 3D Spatial MemoryLLM agents and tool use → Multimodal and embodied agents
Dense Reward for Multi-View 3D Reasoning with Global Maps and Local ViewsLLM reasoning → Spatial and embodied reasoning
VeriEvol: Scaling Multimodal Mathematical Reasoning via Verifiable Evol-InstructLLM reasoning → Multimodal reasoning
Self-Compacting Language Model AgentsLLM agents and tool use → Agent memory systems
Scaling State-Space Models from Lines to Paragraphs: An Ablation of Mamba-based OCRState-space sequence models (S4, Mamba) → Core SSM architectures for language
Concordia: JIT-Compiled Persistent-Kernel Checkpointing for Fault-Tolerant LLM InferenceLLM agents and tool use → Software engineering coding agents +1
TROPT: An Open Framework for Unifying and Advancing Discrete Text OptimizationLLM safety and security → Jailbreaking
MeshFlow: Mesh Generation with Equivariant Flow MatchingDiffusion models → 3D, point cloud and scene generation
CADRE: Stable, Parameter Efficient Adaptation of Medical Vision Language Models with Bounded Forgetting and Prior DriftMultimodal LLMs / vision-language models → Medical vision-language models
SkyJEPA: Learning Long-Horizon World Models for Zero-Shot Sim-to-Real Control of QuadrotorsWorld models → Latent predictive world models
Cross-Architectural Mixture-of-Experts with Adaptive Soft Routing for Plant Leaf Disease ClassificationMixture-of-experts → Vision mixture-of-experts
Flowing With Purpose: Latent Action Guided Flow Matching Policies For Robotic ManipulationDiffusion models → Diffusion and flow policies for robotics
Detecting Malicious Agent Skills in the Wild using AttentionLLM agents and tool use → Agent threats, guardrails and governance
HyperQuant: A Rate-Distortion-Optimal Quantization Pipeline for Large Language and Diffusion ModelsDiffusion models → Diffusion foundations and sampling +1
ReasoningLens: Hierarchical Visualization and Diagnostic Auditing for Large Reasoning ModelsLLM reasoning → Test-time compute and efficient thinking
Litmus: Zero-Label, Code-Driven Metric Specification for Evaluating AI SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
Measuring & Mitigating Over-Alignment for LLMs in Multilingual Criminal Law CourtsLLM safety and security → Safety alignment, refusal and guardrails
TSD: A Physics-Inspired Trajectory Saliency Detector for Efficient Imitation LearningOffline RL and imitation learning from data → Robot manipulation imitation learning
TooBad: Backdoor Diffusion Models with Ultra-Low Poison Rate and Imperceptible TriggerDiffusion models → Robustness, privacy and adversarial aspects +1
Faithful Grounded Visual Reasoning via Learned Proxy-TokensLLM reasoning → Multimodal reasoning +1
Superhuman AI for Generals.io Using Self-Play Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
Group Selection Promotes Prosocial Prompts in Populations of LLM AgentsLLM agents and tool use → Social simulation and multi-agent debate
The Watermark Shortcut: How Provenance Marking Sabotages Audio Deepfake DetectionLLM safety and security → Watermarking generated content and models
Tmax: A simple recipe for terminal agentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Uncertainty-based Debiasing and Unlearning for DecontaminationLLM safety and security → Machine unlearning
EHR-Complex: Benchmarking Medical Agents for Complex Clinical ReasoningLLM reasoning → Clinical and biomedical applications +1
Ocean4D: Generative Underwater 4D Reconstruction via Medium-Aware Video DiffusionDiffusion models → Inverse problems and image restoration
IOI: Decoupling Kinematics and Physics for Interactive World ModelsWorld models → Embodied and robot world models
Flow6D: Discrete-to-Continuous Flow Matching for Efficient and Accurate Category-Level 6D Pose EstimationDiffusion models → 3D, point cloud and scene generation
Causal Reward World Models: Zero-shot Reward Design for Automated Skill GenerationWorld models → Latent dynamics models and model-based RL
GIF: Locally Sound Geometric Information Flow Control for LLMsLLM agents and tool use → Agent threats, guardrails and governance
BoxCtrl: 3D-Aware Visual Prompting for Geometric Image EditingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Safe Few-Step Generation via Velocity EditingDiffusion models → Flow matching, time series and science
SteerVTE: Seamless Video Text Editing with Style and Glyph ControlDiffusion models → Image generation, text-to-image and editing
HOLMES: Evaluating Higher-Order Logical Reasoning in LLMsLLM reasoning → Reasoning benchmarks and evaluation
RS-Gen: A Multi-Stage Agentic Framework for Reasoning and Search-Augmented Image GenerationLLM agents and tool use → Multimodal and embodied agents
MuPPET: A Benchmark for Contextual Privacy of LLM Assistants in Multi-Party ConversationsLLM agents and tool use → Agent threats, guardrails and governance
CFPO: Counterfactual Policy Optimization for Multimodal ReasoningLLM reasoning → Multimodal reasoning +1
Memory Contagion: Cross-Temporal Propagation of Evaluator Bias via Agent MemoryLLM agents and tool use → Agent memory systems
FlowTTS-GRPO: Online Reinforcement Learning with Multi-Objective Reward Optimization for Flow-Matching Based Text-to-SpeechDiffusion models → Audio, speech and music

All 513