Papers

New on arXiv

Sun 17 May 2026

283 papers

Runtime Governance for Agentic AI: Action-Boundary Control with Trusted Provenance and Fail-Closed ExecutionLLM agents and tool use → Agent threats, guardrails and governance
Delay-Adaptive Speculation Control for Low-Latency Edge-Cloud LLM InferenceEfficient LLM inference and serving → Early exit and edge inference
Nano World Models: A Minimalist Implementation of Future Video PredictionWorld models → Video-generation world models and simulators
Towards trustworthy agentic AI: a comprehensive survey of safety, robustness, privacy, and system securityLLM agents and tool use → Agent threats, guardrails and governance
VAMP-Diff: VampPrior Latent Diffusion for Photoplethysmography ModelingDiffusion models → Medical imaging and segmentation
AnimeAdapter: A Modular Adapter for Appearance-Consistent Anime Character GenerationDiffusion models → Image generation, text-to-image and editing
Reasoning Portability: Guiding Continual Learning for MLLMs in the RLVR EraLLM reasoning → Multimodal reasoning +1
Variational Diffusion Channel DecoderDiffusion models → Diffusion for compression and wireless
Auditing Reasoning-Trace Memorization Claims after Unlearning with Head-Conditioned CanariesLLM reasoning → Test-time compute and efficient thinking +1
Agent Bazaar: Enabling Economic Alignment in Multi-Agent MarketplacesLLM agents and tool use → Financial trading and forecasting agents
Do LLM Agents Mirror Socio-Cognitive Effects in Power-Asymmetric Conversations?LLM agents and tool use → Social simulation and multi-agent debate
Fine-tuning Pocket-Aware Diffusion Models via Denoising Policy OptimizationDiffusion models → Flow matching, time series and science
GEM: Gaussian Evolution Model for Occupancy Forecasting and Motion PlanningWorld models → World models for autonomous driving
Stop When Reasoning Converges: Semantic-Preserving Early Exit for Reasoning ModelsLLM reasoning → Test-time compute and efficient thinking +1
PEIRA: Learning Predictive Encoders through Inter-View Regressor AlignmentKnowledge distillation → General knowledge distillation methods
When a Zero-Shooter Cheats: Improving Age Estimation via Activation SteeringLLM safety and security → Safety alignment, refusal and guardrails
Causal Intervention-Based Memory Selection for Long-Horizon LLM AgentsLLM agents and tool use → Agent memory systems
MARQUIS: A Three-Stage Pipeline for Video Retrieval-Augmented GenerationRetrieval-augmented generation → Multimodal RAG
WebGameBench: Requirement-to-Application Evaluation for Coding Agents via Browser-Native GamesLLM agents and tool use → Software engineering coding agents
AI Agents May Always Fall for Prompt InjectionsLLM safety and security → Prompt injection and agent security
SparseSAM: Structured Sparsification of Activations in Segment Anything ModelsEfficient LLM inference and serving → Token pruning and merging for ViTs
SegRAG: Training-Free Retrieval-Augmented Semantic SegmentationRetrieval-augmented generation → Multimodal RAG
Verifier-Guided Code Translation via Meta-Step DecodingLLM reasoning → Reasoning for agents, planning and code
Episodic-Semantic Memory Architecture for Long-Horizon Scientific AgentsLLM agents and tool use → Agent memory systems
GraphMind: From Operational Traces to Self-Evolving Workflow AutomationSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
VeriCache: Turning Lossy KV Cache into Lossless LLM InferenceEfficient LLM inference and serving → KV cache optimization
SafeLens: Deliberate and Efficient Video Guardrails with Fast-and-Slow ScreeningLLM safety and security → Safety alignment, refusal and guardrails
Adaptive Generate-Rank-Verify: Inference-Time Search with Costly VerificationLLM reasoning → Test-time compute and efficient thinking
Venom: A PyTorch Generative Modeling ToolkitDiffusion models → Diffusion foundations and sampling
AutoRubric-T2I: Robust Rule-Based Reward Model for Text-to-Image AlignmentLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
From a Single Demonstration to a General Policy for Contact-Rich ManipulationOffline RL and imitation learning from data → Robot learning from demonstration
Mixture of Experts for Low-Resource LLMsMixture-of-experts → Expert routing strategies
NeuSymMS: A Hybrid Neuro-Symbolic Memory System for Persistent, Self-Curating LLM AgentsLLM agents and tool use → Agent memory systems
Form and Function: Machine Unlearning as a Problem of Misaligned StatesLLM safety and security → Machine unlearning
ECG-WM: A Physiology-Informed ECG World Model for Clinical Intervention SimulationDiffusion models → Medical imaging and segmentation
TAME: Test-Time Adversarial Prompt Tuning via Mixture-of-Experts for Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +2
Stable Routing for Mixture-of-Experts in Class-Incremental LearningMixture-of-experts → Expert routing strategies
How Off-Policy Can GRPO Be? Mu-GRPO for Efficient LLM Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning
Firefly: Illuminating Large-Scale Verified Tool-Call Data Generation from Real APIsLLM agents and tool use → Tool learning and function calling
PFlow-T: A Persistence-Driven Forward Process for Topology-Controlled GenerationDiffusion models → Flow matching, time series and science
Evaluating Deep Research Agents on Expert Consulting Work: A Benchmark with Verifiers, Rubrics, and Cognitive TrapsLLM agents and tool use → Web, search, and research agents
Rethinking Code Review in the Age of AI: A Vision for Agentic Code ReviewLLM agents and tool use → Software engineering coding agents
Accelerating Redshift-Conditioned Galaxy Image Synthesis with One-step Generative ModelingDiffusion models → Inverse problems and image restoration
Self-supervised Hierarchical Visual Reasoning with World ModelWorld models → World models for language agents +1
AgentModernize: Preserving Business Logic in Legacy Modernization with Multi-Agent LLMs and Behavioral Specification GraphsLLM agents and tool use → Software engineering coding agents
CasualSynth: Generating Structurally Sound Synthetic DataSynthetic data and self-improvement for LLMs → LLM synthetic data generation
Designing streetscapes from street-view imagery using diffusion modelsDiffusion models → 3D, point cloud and scene generation
SaaSBench: Exploring the Boundaries of Coding Agents in Long-Horizon Enterprise SaaS EngineeringLLM agents and tool use → Software engineering coding agents
RoboFlow4D: A Lightweight Flow World Model Toward Real-Time Flow-Guided Robotic ManipulationWorld models → Embodied and robot world models
AffordVLA: Injecting Affordance Representations into Vision-Language-Action Models via Implicit Feature AlignmentVision-language-action models → Reasoning and chain-of-thought in VLA
A Distributional View for Visual Mechanistic Interpretability: KL-Minimal Soft-Constraint PrincipleMechanistic interpretability → Interpreting vision and multimodal models
The Silent Brush: Evaluating Artistic Style Leakage in AI Art GenerationDiffusion models → Image generation, text-to-image and editing
Self-Supervised On-Policy Distillation for Reasoning Language ModelsLLM reasoning → Prompting and chain-of-thought reasoning +1
Beyond Linear Superposition: Discovering Climate Features in AI Weather Models with KAN-SAEMechanistic interpretability → Interpreting time-series and weather models
Employing Vision-Language Models for Face Image Quality AssessmentMultimodal LLMs / vision-language models → Social bias and fairness in VLMs
DyGRO-VLA: Cross-Task Scaling of Vision-Language-Action Models via Dynamic Grouped Residual OptimizationVision-language-action models → Reinforcement learning and post-training
The Capability Paradox: How Smarter Auditors Make Multi-Agent Systems Less SecureLLM agents and tool use → Agent threats, guardrails and governance
Mamba-VGGT: Persistent Long-Sequence Video Geometry Grounded Transformer via External Sliding Window Mamba MemoryState-space sequence models (S4, Mamba) → Video, motion and generative Mamba models
Event-B Agent: Towards LLM Agent for Formal Model Synthesis and RepairLLM agents and tool use → Educational agents and math problem solving
VerifyMAS: Hypothesis Verification for Failure Attribution in LLM Multi-Agent SystemsLLM agents and tool use → Software engineering coding agents
ClaHF: A Human Feedback-inspired Reinforcement Learning Framework for Improving Classification TasksLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool FeedbackLLM agents and tool use → Agent threats, guardrails and governance
DeTrack: A Benchmark and Altitude-Aware Dual World Model for Drone-embodied TrackingWorld models → World models for navigation
FastOCR: Dynamic Visual Fixation via KV Cache Pruning for Efficient Document ParsingMultimodal LLMs / vision-language models → Document understanding and multimodal RAG +1
MemRepair: Hierarchical Memory for Agentic Repository-Level Vulnerability RepairLLM agents and tool use → Agent threats, guardrails and governance
DiagEval: Trajectory-Conditioned Diagnosis for Reliable Software Evaluation with GUI AgentsLLM agents and tool use → GUI and mobile agents
Medical Context Distorts Decisions in Clinical Vision Language ModelsMultimodal LLMs / vision-language models → Medical vision-language models
Progressive Generalization Augmentation with Deeply Coupled RND-PPO and Domain-Prioritized Noise Injection for Robust Crop Management Reinforcement LearningDeep reinforcement learning methods → RL for energy, networks and operations
S2Accompanist: A Semantic-Aware and Structure-Guided Diffusion Model for Music Accompaniment GenerationDiffusion models → Audio, speech and music
Ablating Safety: Mechanisms for Removing Alignment in Language Models for Security ApplicationsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks
Rethinking Side-Channel Analysis: Automated Discovery and Analysis of Side-Channel Leakage with LLM-Assisted AgentsLLM agents and tool use → Agent threats, guardrails and governance
ADR: An Agentic Detection System for Enterprise Agentic AI SecurityLLM agents and tool use → Agent threats, guardrails and governance +1
FML-bench: A Controlled Study of AI Research Agent Strategies from the Perspective of Search DynamicsLLM agents and tool use → Web, search, and research agents
MasFACT: Continual Multi-Agent Topology Learning via Geometry-Aware Posterior TransferLLM agents and tool use → General LLM agent frameworks and benchmarks
Learning Transferable Topology Priors for Multi-Agent LLM Collaboration Across DomainsLLM agents and tool use → General LLM agent frameworks and benchmarks
You Can't Fool Us: Understanding the Resilience of LLM-driven Agent Communities to MisinformationLLM agents and tool use → Social simulation and multi-agent debate
AMATA: Adaptive Multi-Agent Trajectory Alignment for Knowledge-Intensive Question AnsweringLLM agents and tool use → Web, search, and research agents +1
Taming "Zombie'' Agents: A Markov State-Aware Framework for Resilient Multi-Agent EvolutionLLM agents and tool use → General LLM agent frameworks and benchmarks
Transitivity Meets Cyclicity: Explicit Preference Decomposition for Dynamic Large Language Model AlignmentLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Single-Sample Black-Box Membership Inference Attack against Vision-Language Models via Cross-modal Semantic AlignmentMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Leveraging Error Diversity in Group Rollouts for Reinforcement LearningDeep reinforcement learning methods → RL training of language and agent models
LPG: Balancing Efficiency and Policy Reasoning in Latent Policy GuardrailsLLM reasoning → Reasoning for agents, planning and code +1
Noise scheduling and linear dynamics in diffusion models on Lie groupsDiffusion models → Diffusion foundations and sampling
ASPI: Seeking Ambiguity Clarification Amplifies Prompt Injection Vulnerability in LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
TClone: Low-Latency Forking of Live GUI Environments for Computer-Use AgentsLLM agents and tool use → GUI and mobile agents
Weak-to-Strong Elicitation via Mismatched Wrong DraftsLLM reasoning → Reinforcement learning for reasoning
VISTA: Triplet-Supervised Video Style Transfer with Diffusion TransformersDiffusion models → Image generation, text-to-image and editing
Attention Hijacking: Response Manipulation Across Queries in Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Reasoning Before Diagnosis: Physician-Inspired Structured Thinking for ECG ClassificationLLM reasoning → Clinical and biomedical applications
Deep Reinforcement Learning Framework for Diversified Portfolio Management Across Global Equity MarketsDeep reinforcement learning methods → Adversarial robustness and RL in finance
Compress the Context, Keep the Commitments: A Formal Framework for Verifiable LLM Context CompressionEfficient LLM inference and serving → Prompt and context compression
ConflictRAG: Detecting and Resolving Knowledge Conflicts in Retrieval Augmented GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
MetaCogAgent: A Metacognitive Multi-Agent LLM Framework with Self-Aware Task DelegationLLM agents and tool use → General LLM agent frameworks and benchmarks
Step-wise Rubric Rewards for LLM ReasoningLLM reasoning → Reinforcement learning for reasoning
CLAP: Contrastive Latent-space Prompt Optimization for End-to-end Autonomous DrivingVision-language-action models → VLA for autonomous driving +1
OProver: A Unified Framework for Agentic Formal Theorem ProvingLLM agents and tool use → Educational agents and math problem solving
ContractBench: Can LLM Agents Preserve Observation Contracts?LLM agents and tool use → Agent threats, guardrails and governance
A2RBench: An Automatic Paradigm for Formally Verifiable Abstract Reasoning Benchmark GenerationLLM reasoning → Reasoning benchmarks and evaluation
Scalable Bi-causal Optimal Transport via KL Relaxation and Policy GradientsDeep reinforcement learning methods → RL theory and sample complexity
Is VLA Reasoning Faithful? Probing Safety of Chain-of-Causation in Autonomous Driving ModelsVision-language-action models → VLA for autonomous driving +1

All 283