Papers

New on arXiv

Mon 10 Aug 2026

507 papers

WatchPoint: Executable User Feedback for Real-World Agentic Web DevelopmentLLM agents and tool use → Software engineering coding agents
Dynamic Deep Prompt Optimization for Defending Against Jailbreak Attacks on LLMsLLM safety and security → Jailbreaking
GDLAM: Group-Disentangled Latent Action Model for Highly Disentangled Embodied PretrainingWorld models → Embodied and robot world models
Vision-Language Models as copilots for Autonomous UAV Navigation: Analysis of Latency and Reliability in Degraded EnvironmentsLLM reasoning → Spatial and embodied reasoning +1
Self-Evolving Skills via Surrogate-Guided Solve-and-ReproduceSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
Mamba-based Selective State Space Modeling Improves the Accuracy-Complexity Tradeoff of SmolVLA Vision-Language-Action ExpertsVision-language-action models → Generalist VLA models and benchmarks +1
Deep Analog: Open-Set Film Emulation with Reference-Conditioned 3D LUTsMixture-of-experts → Vision mixture-of-experts
Toward a Theory of Value in AI AlignmentLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Do Personalized Skills Help Coding Agents? An Empirical Study of Developer Interaction HistoriesLLM agents and tool use → Software engineering coding agents
From Detection to Understanding: TAR and TAR-Bench for Multi-Task Traffic Anomaly ReasoningLLM reasoning → Chain-of-thought reasoning for driving
Co-Evolution in Agentic Systems: Toward Self-Directed Evolution Beyond Human DesignLLM agents and tool use → General LLM agent frameworks and benchmarks
TRACE-GS: On-Policy Trajectory Distillation with Privileged Geometric Conditioning for Sparse-View 3DGS RestorationDiffusion models → 3D, point cloud and scene generation
From Prompt Injection to Web Exploitation: Revisiting Classic Vulnerabilities in LLM-Integrated ApplicationsLLM agents and tool use → Agent threats, guardrails and governance +1
Withholding the Completing Chunk: Exact Release-Boundary Equivalence for Production Streaming GuardrailsLLM safety and security → Safety alignment, refusal and guardrails
Chain of Spatial Thoughts: Modality-Agnostic Spatial Grounding for Vision Language ModelsLLM reasoning → Spatial and embodied reasoning +1
Not a Monolith: Lab-Level Divergence in the Cooperative Equilibria of Chinese Frontier LLM AgentsLLM agents and tool use → Social simulation and multi-agent debate
FACT: Failure-Aware Causal Training for World-Action ModelsWorld models → Embodied and robot world models
Self-evolving Agentic Customer Support System at LinkedInLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Mind Viruses: Self-Propagating Ideas in Multi-Agent LLM SystemsLLM agents and tool use → Agent threats, guardrails and governance
Mitigating Bus Bunching with Reinforcement Learning Enhanced by Semantic Stop EmbeddingDeep reinforcement learning methods → RL for autonomous driving and traffic
Boundary-Seeking Policy Gradient for Safe Reinforcement LearningDeep reinforcement learning methods → Safe and constrained RL
Post-Hoc Sparse Coding of Latent Communication Between Vision-Language Model AgentsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning +1
TRACE: Trustworthy Retrieval-Augmented Conversational EngineRetrieval-augmented generation → RAG robustness and adaptive retrieval
DoseBridge: Denoising Diffusion Bridge Model for Dose Prediction in Lung Intensity-Modulated Proton TherapyDiffusion models → Medical imaging and segmentation
Intrinsic Structure: Spectral Identifiability for Mechanistic InterpretabilityMechanistic interpretability → Circuits and mechanistic analysis
Generating Attacks for LLMs with GFlowNetsLLM safety and security → Prompt injection and agent security
MarkNull: Model-Agnostic Watermark Removal in AI-Generated Images via On-Manifold Latent ManipulationLLM safety and security → Watermarking generated content and models
MAD-HOI: Masked Autoregressive Diffusion for Generating Articulated Hand Object Interactions from TextDiffusion models → Diffusion and flow policies for robotics
SBCO: Self-Supervised, Verifier-Grounded Harness Optimization For Planning AgentsSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
Multimodal Item Parameter Estimation using Simulated Response ProbabilitieMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
The CASE Framework: A Multi-Disciplinary Control Architecture for Governing Enterprise Agentic AILLM agents and tool use → Agent threats, guardrails and governance
REATS: LLM Reasoning-based Ensemble Learning for Adaptive Time Series ForecastingLLM reasoning → Time series reasoning
The Evaluation Protocol Determines the Result: An Independent Reproduction of LeWorldModel on TwoRoomWorld models → Embodied and robot world models
Procedural Fairness Failures in RLHF from Preference AveragingLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
4D-WAM: 4D Consistent World Modeling for Autonomous DrivingWorld models → World models for autonomous driving +1
UserToolBench: A User-Profile-Hidden Benchmark for Personalized Decision Making in Tool-Use LLMsLLM agents and tool use → Tool learning and function calling
FlowScout: From Execution Feedback to Reliable Tool-Using Agent WorkflowsLLM agents and tool use → Software engineering coding agents
DOCSCHISEL: Adaptive Tool Documentation Optimization Framework for LLM AgentsLLM agents and tool use → Tool learning and function calling
Perception Before Supervision: Self-Contained Visual Distillation from Counterfactual Blind SpotsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
Multimodal Model Diffing for Feature Discovery and ControlMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
Learning How the World Evolves: Extrapolative Video World Models via Latent Dynamics ReasoningWorld models → Latent predictive world models +1
DistMoE: Private-data Rehearsal-free Routing in Mixture-of-Experts for Distributed Instruction TuningMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning +1
Decoding-Level Taboo: A Diagnostic Stress Test for LLM RobustnessLLM safety and security → Safety alignment, refusal and guardrails
Consilience for Verifier-Free Test-Time ScalingLLM reasoning → Test-time compute and efficient thinking
Fusion Training for Mathematical Generalization in Large Language ModelsLLM reasoning → Mathematical reasoning
BDH-CQ: In-Context Learning with Recurrent Latent ReasoningLLM reasoning → Test-time compute and efficient thinking
SHE: Trajectory-driven Safety Harness Evolution for LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
Energy-Structured Latent World Models with Neural Time Fields for Physically Constistent Open-World Motion PlanningWorld models → Latent predictive world models
ArchAgent v2: A Case Study with the Data Prefetching ChampionshipLLM agents and tool use → Software engineering coding agents
Sci-VBench: Evaluating Knowledge- and Reasoning-Intensive Video Generation in Science DomainsLLM reasoning → Multimodal reasoning
Stealing Reasoning Traces from Proprietary LLM APIsLLM reasoning → Reasoning for agents, planning and code +1
Agentic Harnesses: LLM-Driven Verification Layers for Robot AutonomyLLM agents and tool use → Agent threats, guardrails and governance
The Greatness of Science Cannot Be Planned: Agentic Auto-Research is Fuzz TestingLLM agents and tool use → Web, search, and research agents
RynnValue: Scaling Robotic Value Foundation Models with Temporal DistanceLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Mismatch Matters: On-Policy Distillation Beyond Token AgreementKnowledge distillation → On-policy and self-distillation for LLMs
Distill Skills into Weights, Not Prompts: Abstract Skills as Privileged Signals for On-Policy Self-DistillationKnowledge distillation → On-policy and self-distillation for LLMs
MedPixel: A Unified Pixel-Language Model for Medical Reasoning and SegmentationLLM reasoning → Clinical and biomedical applications
Hierarchical Fast-Slow ReAct Agent for Zero-Shot Object-Goal NavigationLLM agents and tool use → Multimodal and embodied agents
SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code RefactoringLLM agents and tool use → Software engineering coding agents
SpecPath: Testing Coding Agents Across Contract-Equivalent Specification HistoriesLLM agents and tool use → Software engineering coding agents
CARD: Controlled Agentic Reddit Discussions for Credit Card SimulationLLM agents and tool use → Social simulation and multi-agent debate
ADOPD: Reference-Privileged On-Policy Distillation for MLLM-Based Industrial Anomaly DetectionKnowledge distillation → Detection and dense prediction distillation
KGCaRe: Explainable Complex Conditional Question Answering using Automatic Knowledge Graph Construction and Context Retrieval with LLMsRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
PragMatch: Separating Pragmatic Incongruity from Cross-Modal Mismatch in Large Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
SLIM-0.5B: Learning Action-Grounded Predictive Latents for Robot ManipulationWorld models → Embodied and robot world models +1
Efficient Real-World Online Reinforcement Learning for Robot Manipulation via Centralized Training and Critic DecompositionDeep reinforcement learning methods → RL for robot locomotion and manipulation
Disentangling Co-Occurring Retinal Pathologies with Saliency-Guided Sparse Expert RoutingMixture-of-experts → Medical and biomedical MoE
SR-OPSD: Self-Referenced On-Policy Self-DistillationKnowledge distillation → On-policy and self-distillation for LLMs +1
Rethinking Factor Sharing in Federated LoRA: A Rank-Aware Adaptive ApproachLLM reasoning → Test-time compute and efficient thinking
ColluSkill: Adversarial Cross-Skill Composition for Evading Agent Skill ScannersLLM agents and tool use → Agent threats, guardrails and governance
World Tokens: Enhancing Embodied Policies with Training-Time World ModelingWorld models → Embodied and robot world models +1
Test-Time Scaling for CAD Generation via Verifier-Free Consensus SelectionLLM reasoning → Spatial and embodied reasoning
Matryoshka Language Model SuitesEfficient LLM inference and serving → Speculative decoding
Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world modelsWorld models → World models for language agents +1
Thinking With Tools, Not With Pixels: Tool Calls as Text Scaffolds for Visual ReasoningLLM reasoning → Multimodal reasoning +2
Hallucination-Free GUI Grounding via Regression-Free Layout-Aware MatchingMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Listwise Cross-Encoder Fine-Tuning vs. Agentic Instruction Tuning for LLM Rerankers: A Systematic Study in Medical Procedure RerankingLLM agents and tool use → Medical and scientific domain agents
Activation Probes Surface Code-Security Signals that the Model's Output MissesLLM agents and tool use → Software engineering coding agents
DUET: A Diversity-Quality Duet of Distillation Experts for Two-Step Video GenerationDiffusion models → Video generation
Rethinking Self-Evolving Agents: Do We Still Need Prescribed Optimization Pipelines?LLM agents and tool use → General LLM agent frameworks and benchmarks +1
Satellite Trajectory Optimization via Proximal Policy Optimization for Space Debris AvoidanceDeep reinforcement learning methods → RL for robot navigation and UAV control
Measuring the Wrong Thing: Internal Harmfulness Scores Anti-Rank Successful JailbreaksLLM safety and security → Jailbreaking
Bayesian Symbolic Regression with Entropic Reinforcement LearningDeep reinforcement learning methods → Interpretable and logic-guided RL
Marrying Optimal Transport and ODEs for Unified Continuous-Time 4D Reconstruction and TrackingDiffusion models → Flow matching, time series and science
TSPORec: Token Selection via Preference Optimization for LLM-Based Sequential RecommendationLLM alignment / RLHF / preference optimisation → Preference alignment for recommenders and RAG
FactorDrive: Adaptive Multi-Step Reasoning Driven by Planning-Critical Factors for End-to-End Autonomous DrivingLLM reasoning → Chain-of-thought reasoning for driving +2
You Only Flow Once: Calibrated and Real-Time Radar Pose Estimation with Multi-Hypothesis Normalizing FlowsDiffusion models → 3D, point cloud and scene generation
ElasticBack: Stealthy Conditional Backdoor in LLM-Agent Skills via Coupled Trigger-Rule OptimizationLLM agents and tool use → Agent threats, guardrails and governance +1
The Politician, the Liar, and the Obedient Worker: Emerging Behavior of LLM Agents in Hierarchical GamesLLM agents and tool use → Social simulation and multi-agent debate
SonicWeave: Chunk-Routed Mixture-of-Experts for Unified Audio Scene GenerationMixture-of-experts → Multimodal MoE models
Se-DPO: Self-Evolving Token Credit for Direct Preference OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
From Runnable to Verifiable: An Independent Reproducibility Study of LLM/Agent-Driven Vulnerability Validation ArtifactsLLM agents and tool use → Agent threats, guardrails and governance
Bidirectional Context Self-Distillation for Reinforcement Learning of Skill-Based LLM AgentsLLM agents and tool use → Reinforcement learning for LLM agents +1
Pragmatic Attack Surface: Vulnerabilities of Implicit Context in Large Language ModelsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Dual-Adversarial Safety Alignment: Cultivating Intrinsic Threat Comprehension in LRMsLLM reasoning → Reasoning for agents, planning and code +2
verdi: retrieval is not transfer for continual world model optimizationWorld models → World models for language agents
Carnot: Interpretable, Interactive, and Optimized Execution of Deep Research QueriesLLM agents and tool use → Web, search, and research agents
RangeFactory: Scalable Construction of Multi-Hop Cyber RangesLLM agents and tool use → Agent threats, guardrails and governance
STAIR: Effective Incident Response Using an End-to-End Agentic Planning FrameworkLLM agents and tool use → Agent threats, guardrails and governance
One Adapter Pair per Model: A Universal Activation Interface for Language ModelsMechanistic interpretability → Activation steering methods

All 507