Papers

New on arXiv

Thu 10 Sep 2026

441 papers

Molecular representation shapes the balance between target fidelity and exploration in flow based polymer generationDiffusion models → Flow matching, time series and science
Decoupling Error Attribution in Cloud-Native Graph-RAG: A Data Integrity Diagnostic FrameworkRetrieval-augmented generation → Privacy and security of RAG
Attention-DP3: Spatially Object-aware 3D Diffusion Policy via Geometry-aligned Attentional ConditioningDiffusion models → Diffusion and flow policies for robotics
Task-Based CT Protocol Optimization Using Reinforcement Learning and Virtual Imaging TrialsDeep reinforcement learning methods → RL for clinical treatment and surgery
GroundBench: A Factorized, Counterfactual Benchmark for Locating VLM Affordance FailuresMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
From Experiments to Decisions: Reusing Evidence in Autonomous Coding ResearchLLM agents and tool use → Software engineering coding agents
AnchorVLN: Geometry-Anchored Vision-Language Grounding Reasoning for Open-Vocabulary NavigationLLM reasoning → Spatial and embodied reasoning
Amortized Low-Rank Adaptation for Model-Based Reinforcement LearningWorld models → Latent dynamics models and model-based RL +1
EAR: Entity-Aware Partitioning Approach for Retrieval-Augmented Generation DevelopmentRetrieval-augmented generation → RAG robustness and adaptive retrieval
GTA: Graph Theory Agent and Benchmark for Algorithmic Graph Reasoning with LLMsLLM reasoning → Knowledge- and tool-augmented reasoning +1
Adaptive Chemotherapy Control under Tumor Heterogeneity via Reinforcement LearningDeep reinforcement learning methods → RL for clinical treatment and surgery
DIA: Denoising Intermediate Advantage for Diffusion Policy OptimizationDiffusion models → Diffusion and flow policies for robotics
Chopthin-Consensus Power Sampling: A Diversity-Preserving Approach to LLM DecodingLLM reasoning → Test-time compute and efficient thinking
Repair Before Reinforce: Context-Augmented Knowledge Graph Reasoning for Multi-Hop Question AnsweringLLM reasoning → Knowledge- and tool-augmented reasoning
Vortex: Bridging Extreme Compression and Efficient LLM InferenceEfficient LLM inference and serving → Serving systems and request scheduling
Plans They Abandon, Reports They Author: The Narrative Layer of Autonomous AgentsLLM agents and tool use → Software engineering coding agents
GAUGE: When Not to Trust LLM-as-a-Judge in User-Simulated Evaluation of Task-Oriented AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Retrieval-Augmented Generation for Scientific Code UnderstandingRetrieval-augmented generation → RAG for code
Physics as the label for measuring and correcting materials reasoning in multimodal modelsLLM reasoning → Clinical and biomedical applications
WinSyn: An Automated Pipeline for Realistic Enterprise Question-Answering EvaluationRetrieval-augmented generation → Agentic RAG, agents and memory
NDT Factory: Synthesizing Verified Network Digital Twins from Semantic Models via Multi-Agent LLMLLM agents and tool use → Agentic AI for wireless networks
Certifying Concept Unlearning in Text-to-Image Diffusion ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
Can LLMs in Draft-Verify-Revise Pipelines Resolve Deictic Ambiguity?LLM reasoning → Test-time compute and efficient thinking
Score-based Outlier Generation via Controlling the Radon-Nikodym DerivativeDiffusion models → Diffusion foundations and sampling
Language Is an Insufficient Substrate for Quantitative Reasoning, and Consequential Domains Need Large Quantitative ModelsLLM reasoning → Reasoning benchmarks and evaluation
RodForesight: A World Model Enhanced Diffusion Policy for Slender Rod InsertionWorld models → Embodied and robot world models
Hierarchical Prototype Emergence in Modern Hopfield ModelsDiffusion models → Diffusion foundations and sampling
Efficient Vision-Language-Action Management and Serving for Robot FactoriesMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +2
Reality Is the Final Verifier: On Two Key Gaps in Agentic Software EngineeringLLM agents and tool use → Software engineering coding agents
Pelican-Sim 1.0: A General World Model Simulator for Embodied IntelligenceWorld models → World models for navigation +1
Reinforcement Learning for Syndrome ExtractionDeep reinforcement learning methods → Quantum reinforcement learning
Inverting Self-Triggered Control: Adversarial Reinforcement Learning for Sparse Denial-of-Service AttacksDeep reinforcement learning methods → Adversarial robustness and RL in finance
Certified Safety Curation: Distribution-Free Guarantees for Safe Offline Reinforcement LearningOffline RL and imitation learning from data → Safe offline RL
Test-Driven Approaches to Software Engineering with Large Language Models: A Survey of Phases, Tasks, and Agent SkillsLLM agents and tool use → Software engineering coding agents
Is Bash All You Need? An Empirical Study of Tool Interfaces for Enterprise Digital Worker AgentsLLM agents and tool use → Software engineering coding agents
SenseNova-U1.5: Towards Native Unified Visual IntelligenceKnowledge distillation → Multimodal distillation
Data Scarcity and Model Sparsity: Mixtures-of-Experts Overfit More to Repeated DataMixture-of-experts → MoE large language models
Can Edge-Deployable Vision-Language Models Identify Species?Multimodal LLMs / vision-language models → Agricultural vision-language models
Artificial Id: Drive and Persistent Alignment in Agentic AILLM agents and tool use → General LLM agent frameworks and benchmarks
MindTopo: Can Foundation Models Reason in Topological Space?LLM reasoning → Spatial and embodied reasoning
Caption-once, Frames-on-Demand: Visual-Need Routing for Budget-Aware Agentic Long Video UnderstandingLLM agents and tool use → Multimodal and embodied agents
CausalArena: Benchmarking Causal Discovery in the Foundation Model EraLLM reasoning → Causal reasoning in LLMs
Guided Super-Resolution of Digital Elevation Models with Diffusion-Based Image GeneratorsDiffusion models → Inverse problems and image restoration
UniMPA: A Unified Memory-Prediction-Action Model via Action-Grounded Transition ModelingVision-language-action models → World model-based VLA
The Last AI Built by Humans: Toward Genuine Recursive Self-ImprovementSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
RetroThinker: Enabling Retrospective Thinking in Speech LLMsLLM reasoning → Audio-language model reasoning
BlueSTAR: Tiered Agentic Architecture for Autonomous Cyber DefenseLLM agents and tool use → Agent threats, guardrails and governance
Model-Aware Schedules Improve Generation via Fiberwise Optimal TransportDiffusion models → Flow matching, time series and science
Atlas: Efficient Verifiable Semantic SearchRetrieval-augmented generation → RAG robustness and adaptive retrieval
Predefined-Time Resilient Integral Reinforcement Learning for Input-Constrained Unknown Nonlinear Systems Under FDI Attacks and Disturbances: A Fully Data-Driven ApproachDeep reinforcement learning methods → RL theory and sample complexity
Near-Optimal Reinforcement Learning with Multi-Step Transition LookaheadDeep reinforcement learning methods → RL theory and sample complexity
SpecGuard: Inference-Time Backdoor Detection For FreeEfficient LLM inference and serving → Speculative decoding
Predicting Privacy Leakage from Weight Spectral DensityLLM safety and security → Membership inference attacks
A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias CoefficientsKnowledge distillation → On-policy and self-distillation for LLMs
RAG-Safety-Bench: Reliable Evaluation of Retrieval-Augmented LLM SafetyRetrieval-augmented generation → Privacy and security of RAG +1
Building py-kvcache: A Performance Characterization of External KV Caching for vLLM with NVMe SSDsEfficient LLM inference and serving → KV cache optimization
Reproducibility in the Age of Agentic AI: Context Engineering at the Timescale of a CodebaseLLM agents and tool use → Web, search, and research agents
When Agents Disagree: Bayesian Backward Reasoning as a Label-Free Anchor for Multi-Agent Collective Decision-MakingLLM reasoning → Reasoning for agents, planning and code +1
Negative Self-Distillation: Learning to Reason by Avoiding FlawsLLM reasoning → Prompting and chain-of-thought reasoning +1
ActSafeGuard: Differentiable and Training-Aligned Constraint Enforcement for Flow-Matching PoliciesWorld models → Embodied and robot world models +2
Stress-Testing Dynamical and Generative Downscaling Using Subseasonal Extreme Precipitation ForecastsDiffusion models → Flow matching, time series and science
COBRA-Skills: Contextual Bandit-Guided Evolution for Agent Skill OptimizationLLM agents and tool use → Agent skill learning and evolution
Ecdysis: Efficient and Effective Training of Runtime Harnesses for LLM AgentsLLM agents and tool use → Software engineering coding agents
SSEMG-Net: A Spectrogram-Based Mamba Network for Surface Electromyography DenoisingState-space sequence models (S4, Mamba) → Physiological signal modeling
RDDMPI: Residual Denoising Diffusion Model for Probabilistic Multivariate Time Series ImputationDiffusion models → Flow matching, time series and science
Your Retriever Already Knows: Distribution-Shape QPP for RAG Retrieval SufficiencyRetrieval-augmented generation → RAG robustness and adaptive retrieval
Distributed Optimization of Modular Production Systems using Model-based Reinforcement Learning with Inverse ModelsDeep reinforcement learning methods → RL for energy, networks and operations
OmniKVQuant: KV Cache Quantization for Omni-LLMsEfficient LLM inference and serving → KV cache quantization
TimelyRAG: Semantic-Temporal Hybrid Retrieval for Time-Critical Question Answering in Overlapping-Evolving DocumentsRetrieval-augmented generation → RAG robustness and adaptive retrieval
Memory as Plans: World-Action Modeling with Memory-Grounded PlanningWorld models → Embodied and robot world models
CAP: Continuously Adaptive Perception-Blind Humanoid Locomotion via Learned DenoisingWorld models → Humanoid and legged locomotion world models
World in World: Explore the World with World ModelsWorld models → Video-generation world models and simulators
Ethics Training Agents: Facilitating Group-Based Ethics Education with Role-Playing and Discussion for Ethical Reflection and ExplorationLLM agents and tool use → Educational agents and math problem solving
Breaking the Central Bias: Spatially Partitioned Experts for Coordinate-Based NeuroevolutionMixture-of-experts → MoE theory and gating functions
ChurnBench: A Drift-Aware Benchmark Demonstrating That Refresh Scheduling, Not Cache Age, Governs Staleness in Agentic AILLM agents and tool use → Software engineering coding agents
Recursive Code World Models: Building Complex Worlds through Recursive Scene ProgramsWorld models → Video-generation world models and simulators
From Document Silos to Process Intelligence: A Multi-Layer Knowledge Graph for CMC Process DevelopmentLLM agents and tool use → Web, search, and research agents
Published Unlearning Numbers Move Per Checkpoint, and Not Because the Removed Data Survives: An Audit of 263 Released Batch-Normalized CheckpointsLLM safety and security → Machine unlearning
BridgeMatch: Conditional Transport Bridges in Matching Matrix Space for 3D Deformable RegistrationDiffusion models → 3D, point cloud and scene generation
FARM: Reading Failure Signals from the Internal Predictive States of a Frozen Robotic World ModelWorld models → Latent dynamics models and model-based RL
Safety-aware Skill Adaptation for Reinforcement Learning in Dynamic EnvironmentsDeep reinforcement learning methods → Safe and constrained RL
X-AuT: Progressive Audio-Encoder Compression for Speech LLMs with Cross-Scale DistillationKnowledge distillation → Speech and audio distillation
Beyond Confidence: Stability-Aware Test-Time Adaptation for LLM ReasoningLLM reasoning → Test-time compute and efficient thinking
PATTON: Enabling Commodity PIM for Production LLM ServingEfficient LLM inference and serving → Serving systems and request scheduling
VikingRAG: Accurate and Token-efficient Retrieval-augmented Generation over Structured DocumentsRetrieval-augmented generation → RAG pipeline design and evaluation +1
RAMamba-Net: A Reliability-Aware and Mamba-Based Multimodal Fusion Network for Auditory Attention DetectionState-space sequence models (S4, Mamba) → Speech and audio SSMs
Exploring Diffusion Transformers for Cross-Modal Augmentation in Multimodal Brain State DecodingDiffusion models → Image generation, text-to-image and editing
CoSTAR: Data Synthesis-Driven Constraint-Aware COBOL Section Summarization for Legacy System ModernizationSynthetic data and self-improvement for LLMs → Code and instruction data synthesis
Magenta: Closing the Loop Between Mathematical Reasoning and Lean VerificationLLM reasoning → Mathematical reasoning
Mr.LHDR: A Benchmark for Multimodal Real-World Long-Horizon Deep Research AgentsLLM agents and tool use → Web, search, and research agents
Routing by Reasoning Need: Trajectory-Aware Decoding Control for Diffusion Vision-Language ModelsLLM reasoning → Multimodal reasoning +1
Your Model Already Knows Don't Teach It, Learn to Ask It: Soft Prompting for Few-Shot Adaptation of Vision-Language ModelsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Memory Compression for High-Fanout Agent SandboxesLLM agents and tool use → Agent memory systems
Beyond Noise Steering: Dual-Latent Space Reinforcement Learning for Generative Robot PolicyDeep reinforcement learning methods → RL for robot locomotion and manipulation
Generative Replay Mitigates Sample Starvation in Quantum Architecture SearchDeep reinforcement learning methods → Quantum reinforcement learning
OmniHallu: Unified Hallucination Detection for Cross-Modal Comprehension and Generation in Multimodal Large Language ModelsMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Sci-MMR: Benchmarking Multi-Step Evidence-Grounded Scientific Reasoning in Multimodal AgentsLLM reasoning → Knowledge- and tool-augmented reasoning +1
HALDETECT at ImageEval 2026 Shared Tasks: Answer-First Contrastive Grounding with QLoRAMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
A Voice-Interactive Multi-Agent System for Smart Operating Rooms: Architecture Design and Key TechnologiesLLM agents and tool use → Medical and scientific domain agents
REVA: Reusable Evidence View Aggregation for Context-Efficient RAG ServingRetrieval-augmented generation → Generation and context integration

All 441