Papers

New on arXiv

Thu 24 Sep 2026

527 papers

Diffu-LoRA: A Novel Low-Rank Adaptation for Personalized Diffusion ModelsDiffusion models → Image generation, text-to-image and editing
Synthesis Through Simulation: Generating Coherent Enterprise Data via Scalable Agent-System InteractionLLM agents and tool use → Software engineering coding agents
Task-Oriented Key-Layer KV Communication for Efficient Latent Multi-Agent CollaborationLLM agents and tool use → General LLM agent frameworks and benchmarks
Bounded Autonomy and Verifiable Safety for Agentic AI Enabled AutomationLLM agents and tool use → Agent threats, guardrails and governance
Stable and Counterfactually Robust Physical World Models from Imposed Structure and Learned PhysicsWorld models → Latent predictive world models
Neurosymbolic Routing for Reliable Reasoning on Resource-Constrained Edge DevicesLLM reasoning → Test-time compute and efficient thinking
Cross-Modal Knowledge Distillation for Acoustic Pedestrian DetectionKnowledge distillation → Detection and dense prediction distillation
Panoptic Scene Program Diffusion TransformerDiffusion models → Image generation, text-to-image and editing
SMARtCARE: Privacy-Preserving Agentic AI Systems for Bounded-Autonomy Clinical Decision SupportLLM agents and tool use → Medical and scientific domain agents
Subjects, Not Authors: The Authorship Hazard in Agentic DataspacesLLM agents and tool use → Agent threats, guardrails and governance
MVAgent: Multi-Agent Video Generation via Consistent Condition Construction and Shot-Level Policy OptimizationLLM agents and tool use → Multimodal and embodied agents
Probabilistic Robustness-driven Universal Adversarial Perturbations with Explainability against Deep Reinforcement Learning-based Intrusion Detection SystemDeep reinforcement learning methods → Adversarial robustness and RL in finance
The Hard Part Comes After Search: Benchmarking Web Agents on Synthesizing, Organizing, and Displaying KnowledgeLLM agents and tool use → Web, search, and research agents
Action Forcing: Training World Models on Unsupervised Video by Recovering Underlying Egomotion BasesWorld models → Embodied and robot world models
MOCHA: Multi-Objective Co-Design using Hypernetwork ArchitecturesDeep reinforcement learning methods → RL for games and combinatorial optimization
Atlases Are Already Inside: Recovering Population Templates from Pretrained Diffusion ModelsDiffusion models → Medical imaging and segmentation
Thinking Less to Simulate Better: Intuitive Prompting Improves LLM Agents Simulating Individual Social Media Reactions, Including Unfamiliar ContentLLM agents and tool use → Social simulation and multi-agent debate
SoGuDiff: Socially Guided Diffusion for Steerable, Norm-Grounded Robot NavigationDiffusion models → Diffusion and flow policies for robotics
Auditing Latent-Space Monitors for Autonomous DrivingEnd-to-end autonomous driving → End-to-end system architecture
AutoResearch at Production Scale: Failure Modes and a Multi-Agent FrameworkLLM agents and tool use → Agentic recommender systems
Learning to Replace MCMC in Split-Gibbs Diffusion Posterior Sampling via Deep UnfoldingDiffusion models → Inverse problems and image restoration
Mentored Decoding: Faster Inference meets BoostingEfficient LLM inference and serving → Speculative decoding
CARGO: Context-Aware Retrieval-Gated Evaluation of Agentic AI in ProductionLLM agents and tool use → Web, search, and research agents
RAZOR: Pruning Replaceable Experts in LLMsMixture-of-experts → Expert routing strategies
Policy-Calibrated DAgger: Offline Calibrated Noise Injection for Imitation LearningDiffusion models → Diffusion and flow policies for robotics +1
Spectral Feedback for Test-Time Alignment of Protein Diffusion ModelsDiffusion models → Flow matching, time series and science
Inference-Time Target Speaker Unlearning in LLM-Based Automatic Speech RecognitionLLM safety and security → Machine unlearning
WALT: Learning World-Model-Aligned Latent Trajectories for Autonomous DrivingWorld models → World models for autonomous driving +1
ProCAP: Probabilistic Cross-Attentive Prompt Learning for Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Adaptive Multi-Value Control in LLMs via Causal Activation SteeringMechanistic interpretability → Activation steering methods
POIL: Point-based One-Shot Imitation Learning with Stable Dynamical SystemsOffline RL and imitation learning from data → Robot manipulation imitation learning
CSCWD: Cross-Scale Channel-wise Knowledge Distillation for Lightweight Tiny Object Detection on Edge DevicesKnowledge distillation → Detection and dense prediction distillation
From Weak Data to Strong Policy: Q-Targets Enable Provable In-Context Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Stealth Apart, Harm Together: Skill Cascading Attacks on Skill-Based Agent SystemsLLM agents and tool use → Agent threats, guardrails and governance
Strategic Self-ConsistencyLLM reasoning → Test-time compute and efficient thinking
Coding Agents Aren't Enough! Evaluating an Enterprise Security Brain for Agentic Cloud InvestigationsLLM agents and tool use → Agent threats, guardrails and governance
Bridging LLM Agents and Data Spaces: An Architectural Mediation Approach using the Model Context ProtocolLLM agents and tool use → Web, search, and research agents
Parameters vs. Context: TRACE Fine-Tuning for Robust Retrieval-Augmented GenerationRetrieval-augmented generation → RAG robustness and adaptive retrieval
LLM Agents Can Easily Tamper With Their Own TracesLLM agents and tool use → Agent threats, guardrails and governance
AD-WM: Action-Discriminative World Models for Counterfactual Model Predictive ControlWorld models → Latent predictive world models
Temporal Gradient Inversion for Private Trajectory Reconstruction in Embodied Reinforcement LearningDeep reinforcement learning methods → RL for energy, networks and operations
RAPID: Robot Agentic Programming from DemonstrationsLLM agents and tool use → Multimodal and embodied agents
Rolling-WAM: World Action Models with Rolling ImaginationWorld models → Embodied and robot world models
Coding Agents for Generalized Task and Motion Planning ProblemsLLM agents and tool use → Reinforcement learning for LLM agents
To Trust or Not to Trust: Retrieval-Augmented Fact Checking in SpeechRetrieval-augmented generation → RAG robustness and adaptive retrieval
PoEM: Predicting RL Outcomes from Existing PoliciesDeep reinforcement learning methods → RL training of language and agent models
Instrumental Monitor Evasion Emerges Under Ordinary Task PressureLLM safety and security → Prompt injection and agent security
Underwater C3-JEPA: An Object-Centric Cross-View World Model for ROV SalvageWorld models → Latent predictive world models
The Alignment Illusion in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
SAGE: Mitigating Long-Horizon Reasoning Biases via Topological GuidanceLLM reasoning → Test-time compute and efficient thinking
Jev-Mobile: Jev as an Executor for Mobile GUI AgentsLLM agents and tool use → GUI and mobile agents
Search-Aware Reinforcement Learning for Multi-Component Query Understanding in Roblox Game SearchDeep reinforcement learning methods → RL training of language and agent models
GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AILLM agents and tool use → General LLM agent frameworks and benchmarks
Screen Before You Serve: Simulation for Production Customer Experience AI Agents at 140M ScaleLLM agents and tool use → Financial trading and forecasting agents
Multimodal Thinking with Renderable ProgramsLLM reasoning → Multimodal reasoning +1
Faster Visuomotor Policy Learning on Action Manifolds via Riemannian MeanFlowDiffusion models → Diffusion and flow policies for robotics
Evaluating Agent Skills for Version-Specific Plugin Migration: A Retrospective StudyLLM agents and tool use → Software engineering coding agents
Accelerating Video Diffusion via Training-Free Trajectory RoutingDiffusion models → Diffusion foundations and sampling
Self-Adaptive VLA for Robust Robot DeploymentVision-language-action models → Reinforcement learning and post-training
Self-Play Pretraining with Zero DataSynthetic data and self-improvement for LLMs → Self-play RL for reasoning
Era by Eon: Benchmarking Enterprise Agents on Hidden KnowledgeLLM agents and tool use → Web, search, and research agents
Aim Short to Reach Far: Your Frozen World Model Can Plan Better Than You ThinkWorld models → World models for navigation
Res-HIL: Human-Guided Residual Reinforcement Learning for Sample-Efficient Dexterous ManipulationOffline RL and imitation learning from data → Robot manipulation imitation learning +1
Automated Regulatory Compliance Question Answering in Financial Services with Domain-Adapted Retrieval-Augmented GenerationRetrieval-augmented generation → Legal and financial RAG
Advancing Model Research in AgentX: Long-Horizon Autonomy for Industrial Recommender SystemsLLM agents and tool use → Agentic recommender systems
Let Training Guide Selection: Online Synthetic Data Filtering via Real-Anchored UtilitySynthetic data and self-improvement for LLMs → Model collapse from synthetic data
World Action Agent: Harnessing VLMs for Robot Manipulation via World Action RehearsalLLM reasoning → Spatial and embodied reasoning +2
ENDOPROMPT: Victim-Side Pseudo-References for Utility DegradationLLM safety and security → Prompt injection and agent security
Mind What Matters for Reasoning: Aligning Cross-Modal Attention via Selective Probability Mass ConcentrationLLM reasoning → Multimodal reasoning +1
Beyond Spatial Benchmarks: From Spatial Reasoning to NavigationLLM reasoning → Spatial and embodied reasoning
An Empirical Study of VLM Pipelines for Long-Document QAMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Who Holds the Pen? Let Specifications, Not Agents, Sign OffLLM agents and tool use → General LLM agent frameworks and benchmarks
MILO: Efficient Many-shot In-Context Learning with Block-wise Low-rank CompressionEfficient LLM inference and serving → KV cache optimization
MorphIK: Morphology-Conditioned Neural Inverse Kinematics for Unknown RobotsDiffusion models → Diffusion and flow policies for robotics
Accurate Sampling from Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Working with Agentic `Teammates': When a New Organizational Actor Collides with the Human Ecosystem of WorkLLM agents and tool use → Social simulation and multi-agent debate
Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
From Graphs to Feeders: Constraint-Guided Diffusion for Rule-Compliant Feeder GenerationDiffusion models → Diffusion foundations and sampling
When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context CompressionLLM reasoning → Reasoning for agents, planning and code +2
Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMsLLM reasoning → Test-time compute and efficient thinking
Self-Supervised Anchoring of Fingertip Sensing to Proprioception and Proactive Actions for Robot Imitation LearningOffline RL and imitation learning from data → Robot manipulation imitation learning
S2Planner: Multi-Scale Semantic Planner for End-to-End Autonomous DrivingEnd-to-end autonomous driving → End-to-end system architecture
FlashLoop: Fast and Memory-Efficient Looped Transformers via Lazy UpdatesEfficient LLM inference and serving → Token pruning and merging for ViTs
Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic ExecutionLLM agents and tool use → Agent threats, guardrails and governance
Prefilling the Reasoning Channel: Output-Prefix Attacks on Reasoning LLMsLLM reasoning → Reasoning for agents, planning and code +1
Breaking the Environment Wall: Evolving LLM Agent Environments for Recursive Self-ImprovementLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Depth through recurrence: Looped transformers for flow-matching TTSDiffusion models → Audio, speech and music
SWE-PolyVision: Benchmarking Cross-Image Abductive Reasoning for Repository-Level Software EngineeringLLM reasoning → Multimodal reasoning
SWE-Prometheus: Measuring Engineering Governance Improvements in Real-World RepositoriesLLM agents and tool use → Software engineering coding agents
Industrial Anomaly Detection via Defect-Grounded Reasoning in Visual Latent SpaceLLM reasoning → Multimodal reasoning
Demystifying Agent Skills for Smart Contract Auditing: Design, Effectiveness, Behavioral ImpactLLM agents and tool use → Agent threats, guardrails and governance
IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative SynthesisLLM reasoning → Reinforcement learning for reasoning +1
Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment FailuresLLM safety and security → Safety alignment, refusal and guardrails
Rufus-Air: An Open LLM Post-Training RecipeLLM alignment / RLHF / preference optimisation → Process reward models for reasoning
RACaP: Agentic Reasoning, Acting, and Coding as Policies for Evolvable Robot LearningLLM agents and tool use → Reinforcement learning for LLM agents
Robo-Harness K1: Harnessing Robot-Use Agents via Perception AugmentationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Decoupled Early Exits for Task-Dependent Compute Allocation in Flow-Matching VLAsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +3
Epistemic-Probabilistic Model for Guarded Multi-Agent LLM CoordinationLLM agents and tool use → General LLM agent frameworks and benchmarks
Parts-of-Speech as Emergent Categories in SAE Latent SpaceMechanistic interpretability → SAE features and their applications
Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP

All 527