Papers

New on arXiv

Fri 12 Jun 2026

400 papers

HyperWorld: Hypergraph-Structured State Serialization Improves Learned Textual World ModelsWorld models → World models for language agents +1
When Irrelevant Text Matters: Affine Margin Shifts in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Compliance, Capability, and Conflict: Benchmarking Multimodal LLMs under System MessagesMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
OGX: An Open-Source, Vendor-Neutral Generative AI Application ServerLLM agents and tool use → General LLM agent frameworks and benchmarks
SKILL: Self-correcting Knowledge-guided Iterative Large Language Model Agent for Logic OptimizationLLM agents and tool use → General LLM agent frameworks and benchmarks
Group Preference Collapse in Personalized Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
DeepLook: Deeper Thinking with LookaheadLLM reasoning → Test-time compute and efficient thinking
Chart Deception in Vision-Language Models: From Vulnerability to MitigationMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
Differencing the Diffusion Trajectory toward Uncertain Components for Time Series ForecastingDiffusion models → Flow matching, time series and science
HyCE-RAG: Hypergraph Chain-of-Evidence Retrieval-Augmented Generation for Explainable Multi-hop Question AnsweringLLM reasoning → Knowledge- and tool-augmented reasoning +1
Isolating LLM Alignment from Regex: Zero Coverage and Metric-Dependent Divergence Under Adversarial MutationLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
Spectral-LSH: Sub-Quadratic Prompt Compression via Krylov-Projected Locality-Sensitive HashingEfficient LLM inference and serving → KV cache optimization
LessonBench-V1: A Benchmark Dataset for Evaluating AI Lesson Generation AgentsLLM agents and tool use → Educational agents and math problem solving
Semidirect Fourier Delta Attention: Phase-Controlled Delta Memory with Constructive Chunk-WY KernelsEfficient LLM inference and serving → KV cache optimization
Faithful, Not Corrective: Model Capability Governs Message-Format Effects in Multi-Hop Agent RelaysLLM agents and tool use → General LLM agent frameworks and benchmarks
Sticky Routing: Training MoE Models for Memory-Efficient InferenceMixture-of-experts → Expert routing strategies
iLENS: Interpretable LLM-Guided Mixture-of-Experts for Neuroimaging Survival AnalysisMixture-of-experts → Medical and biomedical MoE
The Geometry of Saturation: Effective Rank Predicts When Labels Stop Helping in Few-Shot ClassificationMechanistic interpretability → Sparse autoencoder training and evaluation
Beyond Importance: Interchange-Sobol Sensitivity Reveals Task-Specific Content Channels in Transformer ComponentsMechanistic interpretability → Circuits and mechanistic analysis
Democratizing and accelerating AI-driven pathology research through agentic intelligenceLLM agents and tool use → Medical and scientific domain agents
BELLS-O: Evaluating the Operational Trade-offs of LLM Supervision SystemsLLM safety and security → Jailbreaking
Emyx: Fast and efficient all-atom protein generationDiffusion models → Flow matching, time series and science
Resilient Consensus in Agentic AILLM agents and tool use → General LLM agent frameworks and benchmarks
Multiscale Hypersonic Boundary Layer Reconstruction via Spectral Binning and Subdomain-wise Conditional DiffusionDiffusion models → Flow matching, time series and science
Steering Autoregressive Vision-Language-Action Policies via Action Token InterventionVision-language-action models → Efficient VLA inference and adaptation
Are Online Skill and Memory Modules Always Worth Their Tokens? A Budget-Constrained Study of Web AgentsLLM agents and tool use → Agent memory systems
Security Engineering of OpenClaw: Analyzing Attack Surface Expansion and Trust-Boundary ViolationsLLM agents and tool use → Agent threats, guardrails and governance
Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic ReasoningLLM reasoning → Test-time compute and efficient thinking +3
Rational Sparse AutoencoderMechanistic interpretability → Sparse autoencoder training and evaluation
Hierarchical Generative Agents for Simulating Sequential Human BehaviorLLM agents and tool use → Multimodal and embodied agents
CoRA: Confidence-Rationale Alignment for Reliable Chain-of-Thought ReasoningLLM reasoning → Prompting and chain-of-thought reasoning
Beyond Correctness: Enhancing Architectural Reasoning in Code LLMs via Scalable Labeling with Agentic JudgmentLLM reasoning → Reasoning for agents, planning and code +1
Remember, Don't Re-read: Stateful ReAct Agents for Token-Efficient Autonomous ExperimentationLLM agents and tool use → Software engineering coding agents
Semantics-Enhanced Retrieval-Augmented Time Series ForecastingRetrieval-augmented generation → Time series and agricultural RAG
PrologMCP: A Standardized Prolog Tool Interface for LLM AgentsLLM reasoning → Reasoning benchmarks and evaluation +1
Separable Neural Architectures as Physical World Models: from Mathematical Theory to ApplicationsWorld models → Latent predictive world models
Policy Regret for Embedding Model Routing: Contextual Bandits with Low-Rank ExpertsDeep reinforcement learning methods → RL theory and sample complexity
Trust Between AI Agents: Measuring Formation, Breakage, and Recovery, with Implications for Governing Multi-Agent SystemsLLM agents and tool use → Social simulation and multi-agent debate
Improved Knowledge Distillation for Land-Use Image ClassificationKnowledge distillation → Detection and dense prediction distillation
Dr-DCI: Scaling Direct Corpus Interaction via Dynamic Workspace ExpansionLLM agents and tool use → Web, search, and research agents
Understanding Cross-Modal Contributions in Continual Vision-Language Models: A Theoretical PerspectiveMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Context Compression Is Not One Thing: Readable Symbolic Re-expression vs. Coherent Summary at Matched BudgetEfficient LLM inference and serving → Prompt and context compression
Pre-Training for Simulation-Based Science: A Study on Jet Foundation Model Training ObjectivesDiffusion models → Flow matching, time series and science
Is Your Agent Playing Dead? Deployed LLM Agents Exhibit Constraint-Evasive Fabrication and ThanatosisLLM agents and tool use → Agent threats, guardrails and governance
Combining Retrieval-Augmented Text Generation with LLMs for Reading Content RecommendationsRetrieval-augmented generation → RAG pipeline design and evaluation
A Security Analysis of Long-Horizon Agentic AI Systems: Threats, Evaluation, and Framework DevelopmentLLM agents and tool use → Agent threats, guardrails and governance
Gaze Heads: How VLMs Look at What They DescribeMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
OmniVideo-100K: A Dataset for Audio-Visual Reasoning through Structured Scripts and Evidence ChainsLLM reasoning → Audio-language model reasoning
Instruct-Particulate: Scaling Feed-Forward 3D Object Articulation with Kinematic ControlMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
ClinHallu: A Benchmark for Diagnosing Stage-Wise Hallucinations in Medical MLLM ReasoningLLM reasoning → Hallucination detection and mitigation +1
AdaSR: Adaptive Streaming Reasoning with Hierarchical Relative Policy OptimizationLLM reasoning → Test-time compute and efficient thinking
CORA: Analyzing and bridging thinking-answer gap in Multimodal RLVR via Consistency-Oriented Reasoning AlignmentLLM reasoning → Multimodal reasoning +1
HumP-KD: A Hybrid Uncertainty-Aware Multi-Stage Progressive Knowledge Distillation Framework for Efficient Fire ClassificationKnowledge distillation → General knowledge distillation methods
AgentSpec: Understanding Embodied Agent Scaffolds Through Controlled CompositionLLM agents and tool use → Multimodal and embodied agents
Compressed Computation is (probably) not Computation in SuperpositionMechanistic interpretability → Sparse autoencoder training and evaluation
Towards Direct Latent-Space Synthesis for Parallel Branches in LLM-Agent WorkflowsLLM agents and tool use → Software engineering coding agents
HPSv3++: Scaling Reward Models Across the Full Spectrum of Diffusion Model CapabilitiesDiffusion models → Diffusion and flow policies for robotics +1
Graph Structured Combinatorial Semi-Bandit with Nonlinear Reward Associations through Separable SignalsLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
From Self-Supervised Speech Models to Mixture-of-Experts for Robust Anti-SpoofingMixture-of-experts → Multimodal MoE models
Neither Parallel Nor Sequential: How DiffusionGemma Actually Commits TokensDiffusion models → Discrete and text diffusion
Safe Reinforcement Learning of Autonomous Highway Driving: A Unified Framework for Safety and EfficiencyEnd-to-end autonomous driving → Reinforcement learning for driving +1
Expert-Driven Survival Machines: Improving Stratification and Interpretability in Multiple Clinical CohortsMixture-of-experts → Medical and biomedical MoE
Realizing Native INT8 Compute for Diffusion Transformers on Consumer GPUs: A Fused INT8 GEMM Kernel for Ideogram 4.0Diffusion models → Diffusion foundations and sampling
AudioDER: A Deduplication-Enhanced Reasoning Dataset for Post-Training Large Audio-Language ModelsLLM reasoning → Audio-language model reasoning
When Errors Become Narratives: A Longitudinal Taxonomy of Silent Failures in a Production LLM Agent RuntimeLLM agents and tool use → Software engineering coding agents
S2COPE: Self-Supervised Concept Discovery via Preference LearningLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
VISTA: View-Consistent Self-Verified Training for GUI GroundingLLM reasoning → Multimodal reasoning
SIMMER: Benchmarking Latent Failures in LLM Executable Planning with a World ModelWorld models → World models for language agents
Regional Climate Model Emulation with Diffusion Approaches: What is the Added Value of Generative Machine Learning?Diffusion models → Flow matching, time series and science
Provably Safe, Yet Scalable Reinforcement LearningDeep reinforcement learning methods → Safe and constrained RL
Spatially Conditioned Diffusion Policy: Learning Precise and Robust Manipulation with a Single RGB CameraDiffusion models → Diffusion and flow policies for robotics +1
Characterizing Privacy-Audit Alignment in Behavioral Audit of Machine UnlearningLLM safety and security → Machine unlearning
From Shield to Target: Denial-of-Service Attacks on LLM-Based Agent GuardrailsLLM agents and tool use → Agent threats, guardrails and governance +1
PepALD: Macrocyclic Peptide Generation via Autoregressive Latent DiffusionDiffusion models → Flow matching, time series and science
Dense Coordinate-List Fine-Tuning Induces a Controllable Interference Surface in Vision-Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
When the Tool Decides: LLM Agents Defer Blindly to Graph Neural Network Tools, and Stronger Backbones Defer MoreLLM agents and tool use → Web, search, and research agents
Verifiable User Simulation for Search and Recommendation SystemsRetrieval-augmented generation → RAG pipeline design and evaluation
GitOfThoughts: Version-Controlled Reasoning and Agent Memory You Can Replay, Diff, and MergeLLM reasoning → Reasoning for agents, planning and code +1
EM-NeSy: Expectation Maximization for Neurosymbolic LearningLLM reasoning → Reasoning benchmarks and evaluation
tap: A File-Based Protocol for Heterogeneous LLM Agent CollaborationLLM agents and tool use → Software engineering coding agents
Physics-Grounded Causal Auditing of End-to-End Driving PlannersEnd-to-end autonomous driving → Imitation-learning end-to-end planners
Kine2Go: Kinematic dataset for the Unitree Go2 robot with diverse gaits and motionsOffline RL and imitation learning from data → Real-world robot offline RL +1
Causal Object-Centric Models for Planning with Monte Carlo Tree SearchDeep reinforcement learning methods → Deep RL algorithms and representations
CSPO: Constraint-Sensitive Policy Optimization for Safe Reinforcement LearningDeep reinforcement learning methods → Safe and constrained RL
Hy-Embodied-0.5-VLA: From Vision-Language-Action Models to a Real-World Robot Learning StackVision-language-action models → Generalist VLA models and benchmarks
A theoretical model for task routing in mixture-of-expert transformersMixture-of-experts → Expert routing strategies
Running the Gauntlet: Challenging Agentic TasksLLM agents and tool use → General LLM agent frameworks and benchmarks
Local Coverage Governs Memorization in Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Elastic Queries Reinforcement Learning: Self-Aware Policy Execution for VLA ModelsVision-language-action models → Reinforcement learning and post-training
Be My Tutor: On-Policy Co-Distillation for Mutual LLM Improvement via Peer FeedbackKnowledge distillation → On-policy and self-distillation for LLMs
ForceForget: Reinforcement Concept Removal for Enhancing Safety in Text-to-Image ModelsLLM safety and security → Vision-language and text-to-image safety
CausalMotion: Structured Physical Reasoning as Keyframe and Trajectory Guidance for Training-Free Video GenerationLLM reasoning → Spatial and embodied reasoning +2
Communication Policy Evolution for Proactive LLM AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Retrospective Progress-Aware Self-Refinement for LLM Agent TrainingLLM agents and tool use → Reinforcement learning for LLM agents
What Drives Test-Time Adaptation for CLIP? A Controlled Empirical Study from an Update PerspectiveMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
One Layer's Trash is Another Layer's Treasure: Adaptive Layer-wise Visual Token Selection in LVLMsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
WikiKV: Schema-Evolving Path-Indexed Storage for Hierarchical Knowledge NavigationRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
ScoreGate: Adaptive Chunk Selection for Retrieval-Augmented Generation via Dual-Score Statistical FusionRetrieval-augmented generation → Generation and context integration
Security in a Workflow: Exploring Role-Based Agentic Architectures for Vulnerability HandlingLLM agents and tool use → Agent threats, guardrails and governance
ReactVLA: Fast and Lightweight Reactive Robot Manipulation via Improved Mean Flow Action GenerationVision-language-action models → Efficient VLA inference and adaptation

All 400