Papers

New on arXiv

Fri 14 Aug 2026

354 papers

DeepInstructor: An Agentic AI Instructor for Experience-Driven Idea EvaluationLLM agents and tool use → Web, search, and research agents
Discovering and Preserving Category Correlation Knowledge via Adaptive Reciprocal Knowledge DistillationKnowledge distillation → General knowledge distillation methods
OpenAgentFlow: Enabling System-Wide Safety Boundaries for Heterogeneous AI Agent FleetsLLM agents and tool use → Agent threats, guardrails and governance
BiasMix-Finance: Post-Generation KYC Guardrails for LLM Portfolio AdviceLLM safety and security → Safety alignment, refusal and guardrails
Trust Is Not Enough: Influence Calibration for On-Policy Self-Distillation in Agentic RLLLM agents and tool use → Reinforcement learning for LLM agents +1
LLMs Can Predict Failure Risk, But Struggle to Predict Which Collaboration Protocol Pays Off: Cost-Aware Protocol Routing Across Reasoning TasksLLM reasoning → Reasoning for agents, planning and code +1
SpIn-ViT: Designing a Sparsity-Induced Vision Transformer That Is Mechanistically InterpretableMechanistic interpretability → Interpreting vision and multimodal models
Optimal Watermark Localization in Mixed-Source Large Language Model TextsLLM safety and security → Watermarking generated content and models
How Do Agents Fail on AutoResearch: End-to-End Diagnostic Evaluation on 100 Real-World Frontier Research TasksLLM agents and tool use → Web, search, and research agents
Interpretable Cross-Lingual Alignment in Small Language Models: Probing Cultural and Pragmatic Reasoning in Japanese-English Bilingual LLMsLLM reasoning → Reasoning benchmarks and evaluation +1
Where Does Retrieval Fail? Evaluating RAG Architectures for Agricultural AdvisoryRetrieval-augmented generation → Time series and agricultural RAG
Workspace Topology as an Attack Vector in Agentic Coding AssistantsLLM agents and tool use → Agent threats, guardrails and governance +1
Evaluating Agentic Code Repair Capabilities in Distributed SystemsLLM agents and tool use → Software engineering coding agents
STAR-FL: Secure Federated Learning with Spatial-Temporal Analysis and Robust AggregationLLM safety and security → Federated unlearning and poisoning
What to Forget in Unlearning? Forget Set Curation for Language ModelsLLM safety and security → Machine unlearning
Zero-MELO: Test-Time Evidence Calibration with Multimodal LLMs for Zero-Shot Micro-Gesture RecognitionMultimodal LLMs / vision-language models → Multimodal emotion understanding
MINT: Min-Selection Preference Distillation for Balanced Multi-Objective AlignmentLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Emergent Misaligned Communication in Long-Horizon Multi-Agent LLM CommerceLLM agents and tool use → General LLM agent frameworks and benchmarks
Imagining Recovery: Inference-Time Counterfactual Realignment for Vision-Language-Action ModelsVision-language-action models → Failure recovery and robustness
Generated Context versus Governed State: Functional Conditions for Accountable Longitudinal Clinical ReasoningLLM reasoning → Clinical and biomedical applications
Beyond Tokens: A Survey on Decoding Methods for Large Language and Vision-Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
CEDAR-GRPO: Process-Aware Reinforcement Learning for General Abductive Reasoning in LLMsLLM reasoning → Reasoning benchmarks and evaluation
Instruction-Based Video Editing by Repurposing an Image Editing ModelDiffusion models → Video generation
From Positionwise Confidence to Prefix Scheduling: Verifier Skipping in Speculative DecodingEfficient LLM inference and serving → Speculative decoding
Agentic Data Cleaning Without a Clean Reference: An Experimental Study of Capabilities and Trade-offsLLM agents and tool use → Web, search, and research agents
KHiM-Mamba: Injecting Pathology Knowledge into Mamba via Hidden-State Modulation for Whole Slide Image AnalysisState-space sequence models (S4, Mamba) → Mamba for medical image analysis
Marionette: Predicting World States, Rendering Geometry, Painting AppearanceWorld models → Video-generation world models and simulators
Rollplex: Cross-Phase GPU Spatial Sharing for Vision Language Model Post-TrainingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Generating Benchmark Health Data Using a Tabular Diffusion TransformerDiffusion models → Flow matching, time series and science
Twin: Playing an Unknown Game with a Test-Time Digital TwinWorld models → World models for language agents +1
You Only Pass Once: Answering and Abstaining Together in a Single Forward Pass of a Frozen Language ModelLLM reasoning → Test-time compute and efficient thinking
SheetCompass: Hierarchical Relation Graphs for Agentic Spreadsheet ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning +1
Wyvern: An Agentic Framework for Generating Grounded Multimodal ReportsLLM agents and tool use → Web, search, and research agents
Designing Reinforcement Learning for Diffusion Models: A Unified Path-Space ViewDiffusion models → Diffusion and flow policies for robotics +1
Offline Deep Q* Estimation with Diffusion ModelsDiffusion models → Diffusion and flow policies for robotics +1
Tripwire: Triggering Aligned Refusal via Statistically Certified Safety NeuronsLLM safety and security → Safety alignment, refusal and guardrails
DeaMoE: Efficient MoE Structure for Fast Small-Batch DecodingMixture-of-experts → MoE large language models
AgentRewind: Recoverable Execution for Long-Horizon LLM AgentsLLM agents and tool use → Software engineering coding agents
Reflex: Enabling Fast and Predictive Vision-Language-Action Models for Reaction-Critical ManipulationVision-language-action models → Failure recovery and robustness
ScienceFlow: A long-horizon agent for ML research, scientific discovery and beyondLLM agents and tool use → Web, search, and research agents
ATLAS: Discovering Agent Strategies through LLM-Guided Abstraction and Automata LearningLLM agents and tool use → General LLM agent frameworks and benchmarks
Clearing the Fog: Towards Installing and Refining Proactive Exploration Capabilities in LLM AgentsLLM agents and tool use → Reinforcement learning for LLM agents
Beyond Capacity: Scalable MoE LLM Inference via High-Bandwidth Flash with Direct GPU and HBM PathsMixture-of-experts → Efficient MoE training and inference systems +1
CORAL: Curriculum-Optimized Reward Adaptation for LiDAR-Based Goal-Directed Urban DrivingEnd-to-end autonomous driving → Reinforcement learning for driving +1
Envs-FORGE: Frontier-Optimized Reward-Grounded Environment Synthesis for Agent RLDeep reinforcement learning methods → RL training of language and agent models
Spatial Message Passing in Language Space for Pathology Image InterpretationMultimodal LLMs / vision-language models → Medical vision-language models
Information Spreading in Diffusion Models from Effective Field TheoryDiffusion models → Diffusion foundations and sampling
Seeing Red, Thinking Bad: Color Bias in Vision Language ModelsLLM reasoning → Multimodal reasoning +1
SimpleOPD: Simple Tokenizer-Agnostic On-Policy Distillation for Long-Context ReasoningLLM reasoning → Prompting and chain-of-thought reasoning +1
TimeSage-EV: A Live Benchmark for Agentic Time Series Analysis in Evolving EnvironmentsLLM agents and tool use → Financial trading and forecasting agents
On the Robustness of Temporal Vision-Language Models for Surgical Endoscopy VideosMultimodal LLMs / vision-language models → Medical vision-language models
Personalized Digital Semantic Communication for Image Transmission with Vision-Language ModelsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
Polaris : Multi Agentic System for Conversational Enterprise AnalyticsLLM agents and tool use → Web, search, and research agents
The More Popular, The Harder to Forget: Adaptive Popularity for LLM UnlearningLLM safety and security → Machine unlearning
AutoSchema: Live Schema Grounding for Agentic Text-to-Sparql over Heterogeneous Knowledge GraphsLLM agents and tool use → Medical and scientific domain agents
FreeBalance: Pre-Routing Online Moe Load Balancing via Residual Workload PredictionMixture-of-experts → Efficient MoE training and inference systems
KV Cache Compression Through the Lens of Transform CodingEfficient LLM inference and serving → KV cache optimization
Can Language Models Understand mmWave Data? Benchmarking Large Language Models for mmWave Radar-Based Human UnderstandingLLM reasoning → Multimodal reasoning
Concept Guidance: Precise, Training-Free Latent Control for Text-to-Image GenerationDiffusion models → Image generation, text-to-image and editing
BiasTrace: Linking Reasoning Behaviours to Biased Outputs in LLMsLLM reasoning → Reasoning benchmarks and evaluation
Removing Temporal Note Redundancy Improves Multimodal Reinforcement Learning for MedicineDeep reinforcement learning methods → RL for clinical treatment and surgery
Deep Reinforcement Learning solution for pickup and delivery routing problems with time window and capacity constraintsDeep reinforcement learning methods → RL for games and combinatorial optimization
CSG-Mamba: A Convolutional Scoring Gating Vision State Space Network for Endoscopic Polyp SegmentationState-space sequence models (S4, Mamba) → Mamba for medical image analysis
Self-Supervised Visual On-Policy DistillationKnowledge distillation → Multimodal distillation
AgilePE: Autonomous UAV Pursuit-Evasion via Self-Play Reinforcement LearningDeep reinforcement learning methods → RL for robot navigation and UAV control
Act2Intention: A Benchmark For Developing Active Mobile Agents Through Inferring User Intention from GUI ActionsLLM agents and tool use → GUI and mobile agents
LegacyWorld: Atomicity-Aware Evaluation of GUI Agents for Legacy WorkflowsLLM agents and tool use → GUI and mobile agents
Traj-LeWM: Path-Aware World-Model Planning via Latent Trajectory CostWorld models → Latent predictive world models
Reinforcement Learning-Based Production Scheduling in an Industry-Based Coating Scenario Using the Digital Model PlaygroundDeep reinforcement learning methods → RL for energy, networks and operations
Learning to Run Power Networks: Effective AlphaZero-inspired Topological ControlDeep reinforcement learning methods → RL for energy, networks and operations
A Graph-Based Reinforcement Learning Framework for Structured Drift Diagnosis and Recovery in Autonomous LLM AgentsLLM agents and tool use → Reinforcement learning for LLM agents
Ambisonics Encoding of Room Impulse Responses using a Device-Agnostic Diffusion ModelDiffusion models → Audio, speech and music
P2Skill: Privacy Preserving Skill Distillation for Cloud-Local LLM Inference SystemsLLM reasoning → Reasoning for agents, planning and code +1
AppLooper: An Agentic Application Engineering Loop for Accountable Release with Virtual-User FeedbackLLM agents and tool use → Software engineering coding agents
CoDS: Robust Collaborative Perception via Expert-driven Detection and BEV SegmentationMixture-of-experts → Vision mixture-of-experts
PILOT: Privileged Imitation Learning for End-to-End Motion Planning of Autonomous UAVs under Partial ObservabilityOffline RL and imitation learning from data → Imitation learning methods
Mandato: Protocol-Level Enforcement of Digitally Signed Mandates on AI Agent Actions with Cryptographically Chained Audit TrailsLLM agents and tool use → Agent threats, guardrails and governance
MACS: A Hybrid Multi-Agent Framework for Reliable Conversational E-Commerce RecommendationLLM agents and tool use → Agentic recommender systems
When Denoising Hurts: Rethinking the Terminal Step of Diffusion Time Series Forecasters -- Extended VersionDiffusion models → Flow matching, time series and science
HERMES: a multi-agent framework for structured knowledge extraction from ultra-long documents in geoscienceLLM agents and tool use → Web, search, and research agents
Model-agnostic Retrieval-Augmented Extended Forecasting for time seriesRetrieval-augmented generation → Time series and agricultural RAG
Evolve Vision-Language-Action Model into an Agent with On-the-fly Tool-useVision-language-action models → Generalist VLA models and benchmarks +1
Source-Agnostic Image Translation Based on Latent Aware Adaptive MaskingDiffusion models → Diffusion foundations and sampling
Beyond Text Conditioning: A Systematic Study of MLLM-DiT Fusion for Video GenerationDiffusion models → Video generation
Adversarial Learning of Classifier-Free Guidance SchedulesDiffusion models → Diffusion foundations and sampling
Demystifying Agent Skills: Why They Work-Until They Don'tLLM agents and tool use → Agent skill learning and evolution
Agent-Orchestration in Autonomous Chip DesignLLM agents and tool use → Software engineering coding agents
HAM-RAG: Hierarchy-Aware Multimodal RAG for Structure-Faithful Interleaved GenerationRetrieval-augmented generation → Multimodal RAG
SSP: An Event-Matched Syn2Sim2Phy Cross-Domain Evaluation Framework for Autonomous Driving VLA ModelsEnd-to-end autonomous driving → Reasoning VLA and VLM driving policies
ForgeWM: Progressive Causal Training for Few-Step Action-Conditioned Video World ModelsWorld models → Video-generation world models and simulators
Content Based Video Narration of Gameplay with Vision Language ModelsMultimodal LLMs / vision-language models → Video understanding with MLLMs
MedClaw: Heuristic Agent Harness for Long-Horizon Surgical Video ReasoningLLM reasoning → Clinical and biomedical applications +1
EchoRec: Multi-Item Prediction-Empowered Generative Recommendation via Cycle-Consistent Preference AlignmentLLM alignment / RLHF / preference optimisation → Preference alignment for recommenders and RAG
Batch-wise Adaptive Pruning: Periodic Neuron Activation-Aware Weight Pruning for Language Reasoning ModelLLM reasoning → Test-time compute and efficient thinking
Nanbeige4.2-3B on Apple Silicon: Fixing Deployment Bugs and Decreasing Looped Transformer Memory OverheadLLM agents and tool use → Software engineering coding agents
FIRM: Fine-Grained Intra-Token Representation of Masks for Remote Sensing Reasoning SegmentationLLM reasoning → Spatial and embodied reasoning +1
Rethinking Auxiliary Modalities in Multimodal Zero-shot Anomaly Detection: From Semantic Fusion to Conditional ModulationMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
PPOM: Marginalizing Patch-Grid Phase for CLIP-Based Generalizable Vision-Language Prompt TuningMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
MoE Expert Execution in Disaggregated LLM Serving with a High-Bandwidth ReRAM Near-Memory ArchitectureMixture-of-experts → Efficient MoE training and inference systems +1
Repair, Not Improvement: Decomposing Constrained Decoding in Tool-Call AbstentionLLM agents and tool use → Tool learning and function calling

All 354