Papers

New on arXiv

Thu 17 Sep 2026

463 papers

Sieve and Sage: Efficient Distraction Filtering for Reliable RALM AbstentionRetrieval-augmented generation → Generation and context integration
Adapting Vision-Language Models for Human-Readable XAI in Industrial Object DetectionMultimodal LLMs / vision-language models → Chart understanding and anomaly detection
COVER: Codec-Robust Video Watermarking with Generative Video PriorsLLM safety and security → Watermarking generated content and models
Forecasting Intrathecal Tracer Enhancement from Pre-Contrast Brain MRI: Direct Regression versus Flow MatchingDiffusion models → Medical imaging and segmentation
Can Agents Design Better Chips with a Higher Level Abstraction?LLM agents and tool use → Software engineering coding agents
CoLearn: An Agentic Tutor that Learns its Learner in a Human--AI Co-Learning LoopLLM agents and tool use → Educational agents and math problem solving
A Multi-Engine Dataflow for MoE Decoding on Scratchpad-Based Tensor AcceleratorsMixture-of-experts → Efficient MoE training and inference systems
REFINEPPO: Learning Continuous Control Policies by Iterative Action RefinementDeep reinforcement learning methods → Deep RL algorithms and representations
Geometry of Values: Task Vector Composition for Ethical Preference Alignment in Language ModelsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Origin Is All You Need: Provenance-Aware Transformers for Structural Trust-Boundary SeparationLLM safety and security → Prompt injection and agent security
Aligning with Lived Experience: Heterogeneous Benefits of Fine Tuning in Mental Health Support GenerationLLM alignment / RLHF / preference optimisation → Medical LLM alignment
Physically Based Rendering in the Latent SpaceDiffusion models → 3D, point cloud and scene generation
Scaling Discovery through Test-Time CommunicationLLM agents and tool use → General LLM agent frameworks and benchmarks
Catch Me If You Can: Real-Time Feedback Denoising for Responsive VLAsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
ASGARD: Action-Space Guard for UAV Resilience via Reinforcement LearningDeep reinforcement learning methods → RL for robot navigation and UAV control
CaLR: Causal Latent Revision for Robust Diffusion ReasoningLLM reasoning → Diffusion reasoning and test-time scaling +1
ForeTac-VLA: A Forecasting-Based Tactile-Vision-Language-Action Model for Contact-Rich Robotic ManipulationVision-language-action models → Tactile, force-aware and humanoid VLA
Attention-Aware Routing: Coupling Routing and Attention in MoEsMixture-of-experts → Expert routing strategies
Efficient Bayes-Adaptive Reinforcement Learning with Temporal Logic SpecificationsDeep reinforcement learning methods → Interpretable and logic-guided RL
μ2-Bench: A Multilingual Machine Unlearning BenchmarkLLM safety and security → Machine unlearning
Continuous Delayed-Memory Stochastic Gradient Descent and Continuous-Time Reinforcement Learning from History of Astrophysical Time Series StudiesDeep reinforcement learning methods → RL theory and sample complexity
SpaceDiffusion: Over-the-Orbit Diffusion for Space Generate-and-Forward CommunicationsDiffusion models → Diffusion for compression and wireless
WM-VS: Progress-Aligned World Models for Closed-Loop Visual ServoingWorld models → Embodied and robot world models
Coding Agents with an Obstacle-Aware Harness for Safe Robot ManipulationLLM agents and tool use → Agent threats, guardrails and governance
Quantifying Overclaiming Propensity in Frontier LLM AgentsLLM agents and tool use → Software engineering coding agents
Score Centering Stabilizes Off-policy Reinforcement LearningDeep reinforcement learning methods → RL training of language and agent models
An Empirical Study of Harness Design for Coding AgentsLLM agents and tool use → Software engineering coding agents
JEPA-Anything: Learning Predictive Models across Different WorldsWorld models → Latent predictive world models
StageGuard: Learning Stage Transitions for Long-Horizon Robot Tasks via Agentic DistillationLLM agents and tool use → Multimodal and embodied agents
RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents +1
GeoAAC: Geometry-Based Adaptive Action Chunking from Denoising Trajectories in VLA PoliciesVision-language-action models → Efficient VLA inference and adaptation
FlowSGS: Improving Flow Matching Priors for Inverse Imaging with Stochastic InterpolantsDiffusion models → Inverse problems and image restoration
Semantic Action Graph: A Shared Representation for Agent Grounding and Human Interpretation of Sports HighlightsLLM agents and tool use → Multimodal and embodied agents
Agile-WAM: An Agile Tactile World Action Model for Contact-Rich Robot ControlWorld models → Embodied and robot world models
OPTED: On-Policy Fine-Tuning for End-to-End Driving using a Render-Free TeacherEnd-to-end autonomous driving → Imitation-learning end-to-end planners
RAFT: A Stateful Retrieval-Augmented Framework for Troubleshooting AgentsRetrieval-augmented generation → Agentic RAG, agents and memory
dQwen3.5: Hybrid-Attention Diffusion Language ModelsDiffusion models → Discrete and text diffusion
MILER: Semantic Mid-Level Representation for Sim-to-Real Reinforcement Learning in Unstructured Autonomous DrivingEnd-to-end autonomous driving → Imitation-learning end-to-end planners +1
Video DeltaNet: A Video-Native Hybrid Attention for Livestream Video GenerationDiffusion models → Video generation
On-Demand Attention: Language Models Know When to RecallLLM agents and tool use → Agent memory systems
Q&A on Any Spreadsheet Requires Interpreting Its Grid StructureMechanistic interpretability → Knowledge and representations inside LLMs
Deep Noir: Autonomous Steering Discovery via Architectural Chronometry in Transformer ModelsMechanistic interpretability → Activation steering methods
MoWAM: Explicit Future Motion Prediction for Efficient World Action ModelsWorld models → Embodied and robot world models
Learning Foresight without Explicit Trajectories for 3D Diffusion PoliciesDiffusion models → Diffusion and flow policies for robotics
HIL-UMI: Bringing Human-in-the-Loop Post-Training of Vision-Language-Action Models to Universal Manipulation InterfaceVision-language-action models → Human-video and latent-action pretraining
DexTouch-WM: Learning Action-Conditioned Tactile World Models from Human Touch for Dexterous Robot ManipulationWorld models → Embodied and robot world models
SkipVLA: Skipping VLA Steps with Classical Planning for Fast Robot ManipulationVision-language-action models → Generalist VLA models and benchmarks
TraceFlow: Guiding Frozen Flow-Matching Robot Policies with Success and Failure TracesDiffusion models → Diffusion and flow policies for robotics
Chronicle: Cut-Point Replay for Regression Testing of LLM AgentsLLM agents and tool use → Software engineering coding agents
What Does Privileged Information Add to On-Policy Self-Distillation?Knowledge distillation → On-policy and self-distillation for LLMs
Limits of Confidence in DiffusionDiffusion models → Diffusion foundations and sampling
Reasoning Quality Matters: Combating Reasoning Collapse in LLM-based Embedding LearningLLM reasoning → Knowledge- and tool-augmented reasoning
Learning Slope-Adaptive Whole-Body Locomotion for Humanoid Robots in Roofing ConstructionDeep reinforcement learning methods → RL for robot locomotion and manipulation
Mitigating Retaliatory Algorithmic Collusion in Repeated GamesDeep reinforcement learning methods → RL for games and combinatorial optimization
Language-model groups overstate consensus when replaying human deliberation on a reasoning taskLLM reasoning → Reasoning benchmarks and evaluation
Parallelism, critical windows, and separations among diffusion language modelsDiffusion models → Discrete and text diffusion
SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent HarnessLLM agents and tool use → Software engineering coding agents
When EOS Tokens Disagree: Understanding Length Inflation in On-Policy DistillationKnowledge distillation → On-policy and self-distillation for LLMs
Fingerprinting Multimodal Large Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Spotlights: Discovering Improvement Opportunities in Software RepositoriesLLM agents and tool use → Software engineering coding agents
The Bias of Nonlinear Two-Time-scale Stochastic Approximation under Constant Step-SizesDeep reinforcement learning methods → RL theory and sample complexity
Schema-Anchored Latent Reasoning for Semantic Parsing-Based Knowledge Base Question AnsweringLLM reasoning → Knowledge- and tool-augmented reasoning
MM-Future: Multi-Mode Joint World-Action Modeling for Autonomous DrivingWorld models → World models for autonomous driving +1
Generating Heterogeneous 3D Geological Microstructures from 2D Images via a Stable Diffusion-Adversarial ModelDiffusion models → 3D, point cloud and scene generation
A Qualitative Model for Reasoning about Path and SupportLLM reasoning → Spatial and embodied reasoning
STR-Agent: An LLM-Driven Agent for QoS-Aware Routing in LEO Satellite NetworksLLM agents and tool use → Agentic AI for wireless networks
RoboFind: Multi-Agent Personalized Object Search for People Who Are Blind or Have Low VisionLLM agents and tool use → Multimodal and embodied agents
To Copy or Not to Copy: Controlling Speculative Decoding via Intrinsic Model SignalsEfficient LLM inference and serving → Speculative decoding
MTVA-Bench: Evaluating the Language Model Inside Cascaded Voice AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Cross-Modal Attention Acts as a Frequency Filter: Why Verbose Prompts Improve Robustness in Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Think Thrice Before Reranking: Multi-perspective Evidence and Reasoning Integration for Text RerankingLLM reasoning → Knowledge- and tool-augmented reasoning
Local Sparsity Enables Unsupervised LLM Safety DetectionLLM safety and security → Safety alignment, refusal and guardrails
Alignment-Path Distillation from Non-streaming ASR-LLMs for Streaming Speech RecognitionKnowledge distillation → Speech and audio distillation
AnyviewMeter: Adapting Robotic Reward Models with Camera Geometry and Multi-View AttentionLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Design of the IBM Granite 5.0 TurboCTC ASR ModelEfficient LLM inference and serving → Early exit and edge inference
CARE-VI: Conservative Adaptive Reliability Estimation for Value Improvement in Off-Policy Actor-Critic LearningDeep reinforcement learning methods → RL theory and sample complexity
UnifiedPlayers: Enhance Tool-Integrated Reasoning in Agentic Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning +1
MATCH: Model-Aware Tool Learning with Curriculum Scheduling and Hierarchically Gated RewardsLLM agents and tool use → Tool learning and function calling
A Proposal for an Agentic AI Architecture to Support Multi-Domain Decision-Making in the Brazilian Armed ForcesLLM agents and tool use → Agentic AI for wireless networks
A Free Lunch? Adapting PP-OCRv6 for Historical Text RecognitionMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
DART: Distillation-Aware Reparameterization for Training-Free LoRA Reuse in Few-Step Video Diffusion ModelsDiffusion models → Video generation
The Missing Complement: State-Conditioned Minimal Sufficient Evidence for Coding AgentsLLM agents and tool use → Web, search, and research agents
Correct Now, Insufficient Later: Auditing Update Sufficiency in Context CompressionEfficient LLM inference and serving → KV cache optimization
Astronex-World 1.0: Real-Time Interactive World Model FoundationWorld models → Video-generation world models and simulators
EPIG-Tree: Compute-Optimal Branching for Gradient-Efficient Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning +1
AVTrace: Diagnosing Audio-Visual Temporal Reasoning in Omni ModelsLLM reasoning → Audio-language model reasoning
QCPruner: Query-Conditioned Population Coverage for Visual Token PruningMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache CompressionMixture-of-experts → Efficient MoE training and inference systems +1
Enhanced Knowledge Distillation for Detection Transformer via Teacher Prediction RefinementKnowledge distillation → Detection and dense prediction distillation
Hybrid Residual Reinforcement Learning for Contact-Rich Robotic Book InsertionDeep reinforcement learning methods → RL for robot locomotion and manipulation
Neuro-Symbolic Agentic AI for Networked Low-Altitude UAVsLLM agents and tool use → Agentic AI for wireless networks
Not All AI Agents Are Equal: Characterizing Resource and Performance DynamicsLLM agents and tool use → General LLM agent frameworks and benchmarks
MaSCoD: A Multi-Agent Framework for Structural-Context-Guided Candidate Causal Graph GenerationLLM agents and tool use → Medical and scientific domain agents
Co-VLA: Consensus-based Federated Training for Vision-Language-Action ModelsVision-language-action models → Generalist VLA models and benchmarks
CitySTAR: Structured and Topology-Aware Reasoning for Open-Vocabulary Urban 3D GroundingLLM reasoning → Spatial and embodied reasoning
REARL: A Closed-loop Autonomous Driving Simulation Enhancement Framework with Real Traffic Data and Large Language ModelsEnd-to-end autonomous driving → Large language models for driving
TRACE: Accountable Agentic Retrieval for Source Discovery in Digital ArchivesLLM agents and tool use → Web, search, and research agents
Learning Reliable Parking Policies via Offline Reinforcement Learning with Quantized Action RepresentationsOffline RL and imitation learning from data → Imitation and offline RL for driving
PetriBench: Benchmarking LLM Reasoning over Dynamic State SpacesLLM reasoning → Reasoning for agents, planning and code
D-Quant: Driftable Entropy Coding for KV Cache QuantizationEfficient LLM inference and serving → KV cache quantization

All 463