Papers

New on arXiv

Thu 16 Jul 2026

350 papers

LaSeD: Label-Semantic Self-Distillation for Visual-Only Surgical Phase RecognitionKnowledge distillation → Distillation for medical imaging
MechSparse: Mechanism-Guided Sparse PEFT Selection Is Task-ShapedMechanistic interpretability → Circuits and mechanistic analysis
Does Moral Reasoning Training Help or Hurt? Red-Teaming RL-Trained Ethical Agents with Persona AttacksLLM agents and tool use → Reinforcement learning for LLM agents +1
Physics-Aware Video Generation via Agentic Planning and Graph-Guided OptimizationDiffusion models → Video generation +1
PersonaMem-v3: Toward Omni-Platform Personal Intelligence for Holistic User Understanding, Recommendation, and Agentic TasksLLM agents and tool use → Agent memory systems
Learning Agent Execution for KV-Cache Management in Agentic ServingLLM agents and tool use → Software engineering coding agents +1
AEVAL: From Anecdotal to Deterministic Testing for Agentic Skill WorkflowsLLM agents and tool use → Agent skill learning and evolution
VarRate: Training-Free Variable-Rate KV Cache Compression for Long-Context LLMsEfficient LLM inference and serving → KV cache optimization
Diffusion models recover accurate mixture weights despite score function insensitivityDiffusion models → Diffusion foundations and sampling
ADS-C: Antidistillation Sampling for ClassificationKnowledge distillation → General knowledge distillation methods
From Black Box to Executable Logic: Explainable Reinforcement Learning through Prolog Expert SystemsDeep reinforcement learning methods → Interpretable and logic-guided RL
Robust Peak-cost Constrained Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
Prediction-Only Distillation in Linear and Logistic RegressionKnowledge distillation → General knowledge distillation methods
Beyond a Joke: Multi-Angle Reasoning for Detecting and Explaining Harmful Humor in MemesLLM reasoning → Multimodal reasoning
Do Coding Agents Need Executable World Models, Simplification, and Verification to Solve ARC-AGI-3?World models → World models for language agents +1
Coercion and Deception in AI-to-AI Management: An Agentic Benchmark of Unprompted EscalationLLM agents and tool use → Social simulation and multi-agent debate
DrawingVQA: A Real-World Benchmark for Multi-Depth Visual-Textual Reasoning on Construction DrawingsLLM reasoning → Spatial and embodied reasoning +1
Precise but Uncoupled: Reviewer Precision Does Not Guarantee Critique Uptake in Multi-Agent Math ReasoningLLM reasoning → Reasoning for agents, planning and code
Reasoning-Guided Part-Level Visual Grounding via Reinforcement LearningLLM reasoning → Multimodal reasoning +1
AnovaX: A Local, Multi-Agent Voice Assistant with LLM Planning, Typed Executors, and Adaptive RecoveryLLM agents and tool use → General LLM agent frameworks and benchmarks
Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World TrajectoriesVision-language-action models → Generalist VLA models and benchmarks
Co-Design of Aeroelastic Systems with Deep Reinforcement LearningDeep reinforcement learning methods → RL with diffusion and flow models
Hierarchical Denoising For Multi-Step Visual ReasoningLLM reasoning → Multimodal reasoning
Partition, Prompt, Aggregate: Statistical Self-Consistency in Language ModelsLLM reasoning → Test-time compute and efficient thinking
SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent CollaborationLLM agents and tool use → Web, search, and research agents
HoloGeo: Mitigating Landmark Bias in Geo-localization via Evidence-Driven ReasoningLLM reasoning → Spatial and embodied reasoning +1
Bridge Evidence: Static Retrieval Utility Does Not Predict Causal Utility in Multi-Step Agentic SearchLLM agents and tool use → Web, search, and research agents
AutoSynthesis: An agentic system for automated meta-analysisLLM agents and tool use → Web, search, and research agents
ARMOR++: Agentic Orchestration of a Multi-Domain Primitive Set for Transferable Attacks on Deepfake DetectorsLLM agents and tool use → Agent threats, guardrails and governance
When Words Are Safe But Actions Kill: Probing Physical Jailbreak Beyond Textual Jailbreak in Hidden-State Risk SpaceLLM safety and security → Jailbreaking
Symbal: Detecting Systematic Misalignments in Model-Generated CaptionsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
BadWAM: When World-Action Models Dream Right but Act WrongWorld models → World model security and GUI agents
Self-Evolving Human-Centered Framework for Explainable Depression Symptom AnnotationSynthetic data and self-improvement for LLMs → Medical synthetic data and agents
Mask-Aware Policy Gradients for Diffusion Language ModelsLLM reasoning → Diffusion reasoning and test-time scaling +1
Plover: Steering GUI Agents through Plan-Centric InteractionLLM agents and tool use → GUI and mobile agents
Benchmarking Multimodal Large Language Models for Scientific Visualization LiteracyMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
On-Policy Delta DistillationLLM alignment / RLHF / preference optimisation → Process reward models for reasoning +1
Setup Complete, Now You Are Compromised: Weaponizing Setup Instructions Against AI Coding AgentsLLM agents and tool use → Software engineering coding agents
Improving Weak World Models Behind Strong Agents in Atari PongWorld models → Latent dynamics models and model-based RL
Digital Pantheon: Simulating and Auditing Coalition Formation with LLM AgentsLLM agents and tool use → Social simulation and multi-agent debate +1
Evaluating covariate balance for long time horizon Markov decision processesOffline RL and imitation learning from data → Offline RL algorithms +1
BrainPilot: Automating Brain Discovery with Agentic ResearchLLM agents and tool use → Medical and scientific domain agents
DriftWorld: Fast World Modeling through DriftingWorld models → Embodied and robot world models
Beyond Single Expert: Harmonizing Diverse Visual Priors in MLLMs for Spatial UnderstandingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Weakly-Supervised RGB-D Salient Object Detection via SAM-driven Pseudo Annotation and State Space Interaction-based DiffusionDiffusion models → Image generation, text-to-image and editing
CosFly-VLA: A Spatially Aware Vision-Language-Action Model for UAV TrackingVision-language-action models → Aerial and drone VLA
SMC-ES: Automated synthesis of formally verified control policiesDeep reinforcement learning methods → Safe and constrained RL
AeroAct: Action-Centered World-Action Models for Language-Conditioned Quadrotor FlightWorld models → World models for navigation
OmniaBench: Benchmarking General AI Agents Across Diverse ScenariosLLM agents and tool use → General LLM agent frameworks and benchmarks
On Success and Simplicity: A Second Look at Transferable Vision-Language Attack PipelineMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
U-shaped Multi-granularity Learning for Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Multi-Axis Max@K Reinforcement Learning for Representative Diversity in Text-to-Image GenerationDeep reinforcement learning methods → RL with diffusion and flow models
Optimal Self-Distillation for Rectified Flow via Linear ProbingDiffusion models → Flow matching, time series and science +1
Steering Robustness into World Action Models via Mechanistic Interpretability and Optimal ControlWorld models → Embodied and robot world models +1
TanGO: Training-Free 3D Editing via Tangent-Space Guidance and OptimizationDiffusion models → 3D, point cloud and scene generation
Human-Robot Interaction in GenAI Architectures via the Agent-Client ProtocolLLM agents and tool use → General LLM agent frameworks and benchmarks
Show Me How You Reason and I'll Tell You Who You Are: Reasoning Graphs for Robust LLM Authorship AttributionLLM reasoning → Knowledge- and tool-augmented reasoning
FirmPilot: Evidence-Guided Multi-Agent Environment Recovery for IoT Firmware RehostingLLM agents and tool use → Software engineering coding agents
FlashDecoder: Real-Time Latent-to-Pixel Streaming Decoder with TransformersDiffusion models → Video generation
StructureClaw: Traceable LLM Agents and an Executable Benchmark for Structural Engineering WorkflowsLLM agents and tool use → Multimodal and embodied agents
Leveraging Instruction Tuning and Merging for Reasoning Model AdaptationLLM reasoning → Reinforcement learning for reasoning
Proof-or-Stop: Don't Trust the Agent, Trust the Evidence -- Loop Engineering for Verifiable Evidence-Gated Lifecycle ControlLLM agents and tool use → Agent threats, guardrails and governance
PAC Learning in Turn-Based Stochastic Games with Reachability Objectives: A Decentralized Private Approach via Expected Conditional DistanceDeep reinforcement learning methods → RL for games and combinatorial optimization
The Energy Society: A Simulation Environment for Studying Agent Cooperation under Survival PressureLLM agents and tool use → Social simulation and multi-agent debate
Towards Human-like Physical Intelligence: Lifelong Vision-Language-Action Learning for Robotic ManipulationVision-language-action models → Reinforcement learning and post-training
Is External Database Protection Static in Retrieval-Augmented Generation? Rethinking Privacy Preservation under Dynamic QueriesRetrieval-augmented generation → Privacy and security of RAG
Transcoders for Investigating Deception in Language ModelsMechanistic interpretability → Probing deception, bias and beliefs
SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents +2
Rare Concept Generation via Counterfactual Inference in Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
AI vs Human Expert Reasoning: Assessing Agreements in Building Typology Predictions based on Street View ImageryLLM reasoning → Spatial and embodied reasoning +1
FoMoVLA: Bridging Visual Foresight and Motion Guidance for Vision-Language-Action ModelsVision-language-action models → World model-based VLA
GeoDetect: Geometric Adversarial Detection for VLPsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
CoTu at EXACT 2026: Neuro-Symbolic Reasoning for Transparent Educational QALLM reasoning → Reasoning benchmarks and evaluation
WorkDrive: Roadwork Chain of Causation for Autonomous DrivingMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection +1
BridgeFlow: Fast and Robust SE(2)-Equivariant Motion Planning with Flow MatchingDiffusion models → Diffusion and flow policies for robotics
Does Multi-Agent Debate Improve AI Feedback on Research Papers?LLM agents and tool use → Web, search, and research agents
VideoSEMA: a scalable and efficient Mamba-like attention for video understandingState-space sequence models (S4, Mamba) → Video, motion and generative Mamba models
Gold-Guided Programmatic Distillation for Financial Reasoning over Hybrid Tables and TextLLM reasoning → Reasoning for agents, planning and code
Reinforcement Learning for the Full Strawberry Harvesting Process: Obstacle Separation, Detachment, and PlacementDeep reinforcement learning methods → RL for robot locomotion and manipulation
Lights, Camera, Malfunction: When Illumination Robustness Leaves VLA Models Blind to ColorVision-language-action models → Safety and adversarial attacks
Reflex: Real-Time VLA Control through Streaming InferenceVision-language-action models → Efficient VLA inference and adaptation +1
ReBind: Multi-Reference Video Editing via Structured Instructions with Explicit Reference RelationshipsDiffusion models → Video generation
TopoAgent: A Self-Evolving Topological Agent for Multimodal Scientific ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning +2
Trajectory-Aware Flow Matching for Topology OptimisationDiffusion models → Flow matching, time series and science
MemPoison: Uncovering Persistent Memory Threats and Structural Blind Spots in LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
D-cut: Adaptive Verification Depth Pruning for Batched Speculative DecodingMixture-of-experts → Efficient MoE training and inference systems +1
NavCMPO: Critic-Guided MeanFlow Policy Optimization for Adaptive NavigationDeep reinforcement learning methods → RL with diffusion and flow models
MCPEvol-Bench: Benchmarking LLM Agent Performance Across Dynamic Evolutions of MCP ServersLLM agents and tool use → Tool learning and function calling
Analytic Abduction: Causal Decomposition and Governed Commitment for Human--AI CoordinationLLM reasoning → Reasoning benchmarks and evaluation
Action QFormer: Structured Representation Shaping under Action Supervision in Vision-Language-Action ModelsVision-language-action models → VLA architectures and generalization
Bad Memory: Evaluating Prompt Injection Risks from Memory in Agentic SystemsLLM agents and tool use → Agent threats, guardrails and governance +1
MagicPrompt: Ultra-Lightweight Prompt Tuning for Video GenerationDiffusion models → Video generation
Multi-LLM Collaborative MRI Report Generation for Visual Instruction Tuning in Brain OncologyMultimodal LLMs / vision-language models → Medical vision-language models
Advanced Image Generation: Negative Prompt Optimization and Latent Classifier GuidanceDiffusion models → Diffusion foundations and sampling
Sharp Stability Threshold and Certification for Designing Stable Residual ArchitecturesState-space sequence models (S4, Mamba) → Core SSM architectures for language
Collaborative Spatial Learning with Multi-LLM Agents in Networked Social ExperimentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Alipay-PIBench: A Realistic Payment Integration Benchmark for Coding AgentsLLM agents and tool use → Software engineering coding agents
A Modern Multimodal Assistant on a 6 GB 2011 GPU: Stage-Validated, All-GPU CUDA Inference for FermiMixture-of-experts → Efficient MoE training and inference systems
MARS: Multi-hop Adaptive Retrieval and SPARQL Generation for KGQALLM reasoning → Knowledge- and tool-augmented reasoning
Seeing the End at Step Zero: Accelerating Diffusion MLLMs via MLP Sparsity-Aware TruncationLLM reasoning → Diffusion reasoning and test-time scaling +1

All 350