Papers

New on arXiv

Tue 7 Jul 2026

421 papers

PhysMent: An Interactive Approach For LLM Reasoning In Physics ProblemsLLM reasoning → Spatial and embodied reasoning
Token Merging for Multilingual Speech Recognition: A Systematic Study Across Model Scale and Fine-TuningEfficient LLM inference and serving → Visual token pruning in multimodal models
World-Time Compute with Verified Code World ModelsWorld models → World models for language agents
Self-Generated Text Recognition: Quality Heuristics, Cross-Task Transfer, and Downstream Bias in LLM EvaluationSynthetic data and self-improvement for LLMs → LLM synthetic data generation
From Prediction to Intervention: Personalized Meal-Level Glucose Regulation via an LLM AgentLLM agents and tool use → Medical and scientific domain agents
An adaptive and evolvable deep reinforcement learning framework for weather predictionDeep reinforcement learning methods → RL for energy, networks and operations
Discrete Action Space as a Prerequisite for GRPO Convergence in Small-Model Continuous ControlLLM reasoning → Reinforcement learning for reasoning
Trajectory-Aware Retrieval Agents for Temporal Decision- MakingLLM agents and tool use → Web, search, and research agents
Multimodal CoLRAG-TF: Triple-Filtered Retrieval for Complex PDFsLLM reasoning → Multimodal reasoning +1
Reliability-Aware LLM Alignment from Inconsistent Human FeedbackLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
NMR Elucidation as an Agentic Search Problem, Not a Modeling ProblemLLM agents and tool use → Medical and scientific domain agents
Reproducing Recurrent Transformers: The CoTFormerLLM reasoning → Prompting and chain-of-thought reasoning
From Intent to Infrastructure: LLM-Driven Agent Compilers for ISAC NetworksLLM agents and tool use → Agentic AI for wireless networks
Robust 4D Driving Scene Reconstruction from Imperfect Visual PriorsWorld models → World models for autonomous driving
EvoCUA-1.5: Online Reinforcement Learning for Multi-turn Computer-Use AgentsLLM agents and tool use → Reinforcement learning for LLM agents +1
A Risk-Field Enhanced Closed-Loop Digital Twin Framework for Autonomous Driving Safety ValidationEnd-to-end autonomous driving → Adversarial scenarios and safety testing +1
Verification of Adaptive Agentic Controllers through Finite Rule RevisionLLM agents and tool use → Agent threats, guardrails and governance
Norm Enforcement for AI Agents: Robustly Shaping Behavior in Multi-Agent SystemsLLM agents and tool use → Social simulation and multi-agent debate
OmniSCS: Omni Safety-Critical Scenario Synthesis for Autonomous Driving via a Fully Editable Driving WorldWorld models → World models for autonomous driving +1
Knowledge-Constrained Shape Optimization with a Mixture-of-Experts Neural Operator for High-Confidence DesignMixture-of-experts → MoE theory and gating functions
Auditable Machine Unlearning for Privacy-Compliant Ransomware Detection Using Multi-Shard SISA and Deep Reinforcement LearningLLM safety and security → Machine unlearning
Gen4U: Unifying Video Generation and Understanding via DiffusionDiffusion models → Video generation
Geometric Self-Distillation for Reasoning GeneralizationLLM reasoning → Prompting and chain-of-thought reasoning +1
A Gold-Standard Study of What Makes a Lightweight Game-Playing Agent StrongLLM agents and tool use → Reinforcement learning for LLM agents
LLMs Silently Correct African American English: Auditing and Mitigating Dialect Bias via Activation SteeringMechanistic interpretability → Probing deception, bias and beliefs
Retrieving and Refining Winning Noise Tickets for Diffusion-Based Motion GenerationDiffusion models → Video generation
Tensor Train Diffusion: Leveraging Low-Rank Structures for High-Dimensional Score-Based SamplingDiffusion models → Diffusion foundations and sampling
Generative Diffusion Models of Stochastic Graph SignalsDiffusion models → Flow matching, time series and science
Evaluating SageMath-Augmented LLM Agents for Computational and Experimental MathematicsLLM agents and tool use → Educational agents and math problem solving
Behavioral Privacy Leakage in Agentic Negotiation: Formalizing and Mitigating Inference Attacks via Randomized PoliciesLLM agents and tool use → Agent threats, guardrails and governance
When Agents Go Rogue: Activation-Based Detection of Malicious Behaviors in Multi-Agent SystemsLLM agents and tool use → Agent threats, guardrails and governance
What Predicts Correctness in Text-to-SQL? A Selective-Prediction StudyLLM reasoning → Knowledge- and tool-augmented reasoning
Enhancing deep learning models for time series classification via knowledge distillationKnowledge distillation → General knowledge distillation methods
From Agentic to Autogenic Network Management for AI-Native 6G and Beyond: A Standards PerspectiveLLM agents and tool use → Agentic AI for wireless networks
Cost-Effective Agent Harnesses for Abstract Reasoning and Generalization on ARC-AGI-1LLM reasoning → Multimodal reasoning
Trees from Marginals: Autoregressive drafting with factorized priorsEfficient LLM inference and serving → Speculative decoding
LLM-powered reasoning in agent-based modelingLLM reasoning → Reasoning for agents, planning and code +1
When Does In-Context Search Help? A Sampling-Complexity Theory of Reflection-Driven ReasoningLLM reasoning → Test-time compute and efficient thinking
Reliable and Developer-Aligned Evaluation of Agents for Software EngineeringLLM agents and tool use → Software engineering coding agents
Vision Language Action (VLA) Models for Unmanned Aerial Robotics and Bimanual Manipulation: A ReviewVision-language-action models → Aerial and drone VLA
NativeMEM: Native Memory Compression for Long-Horizon Robotic ManipulationVision-language-action models → Reasoning and chain-of-thought in VLA
Pelican-VLA 0.5: Attending Before Acting Benefits GeneralizationVision-language-action models → Generalist VLA models and benchmarks
POPS: Recovering Unlearned Multi-Modality Knowledge in MLLMs with Prompt-Optimized Parameter ShakingMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
ORAN-DEFEND: Subspace Detection and Sanitization of Backdoor DRL xApps in Open RANDeep reinforcement learning methods → Adversarial robustness and RL in finance
Healthier LLMs: Retrieval-Augmented Generation for Public Health Question AnsweringRetrieval-augmented generation → Medical and biomedical RAG
What a World Model Represents Is Three QuestionsWorld models → World models for language agents
ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing ModalitiesMixture-of-experts → Medical and biomedical MoE
Dynamic-in-Few-Step: Unifying Dynamic Computation and Few-Step Distillation for Efficient Video GenerationDiffusion models → Video generation
Open-Ended Scenario Reasoning for Specialist Model AdaptationLLM reasoning → Reasoning for agents, planning and code
AgentLens: Production-Assessed Trajectory Reviews for Coding Agent EvaluationLLM agents and tool use → Software engineering coding agents
SpaR3D-MoE: Adaptive 3D Spatial Reasoning from Sparse Views Meets Geometry-Inductive Mixture-of-ExpertsLLM reasoning → Spatial and embodied reasoning +2
HiFuzz: Hierarchical Reinforcement Learning for Semantic-Aware and Adaptive CPU FuzzingDeep reinforcement learning methods → RL training of language and agent models
Deep Reinforcement Learning for Reliability Based Bi-Objective Portfolio OptimizationDeep reinforcement learning methods → Adversarial robustness and RL in finance
D2PO: Optimizing Diffusion Samplers via Dynamic PreferenceDiffusion models → Diffusion and flow policies for robotics +1
Security and Privacy in Agentic AI: Grand Challenges and Future DirectionsLLM agents and tool use → Agent threats, guardrails and governance
NEST: Tackling Dataset-Level Distribution Shifts via Regime-Oriented Mixture-of-ExpertsMixture-of-experts → MoE for time series and graphs
TriRoute: Unified Learned Routing for Joint Adaptive Attention, Experts, and KV-Cache AllocationMixture-of-experts → Expert routing strategies +1
Lift3D-VLA: Lifting VLA Models to 3D Geometry and Dynamics-Aware ManipulationVision-language-action models → 3D and spatial-aware VLA
RynnWorld-4D: 4D Embodied World Models for Robotic ManipulationWorld models → Embodied and robot world models
RynnWorld-Teleop: An Action-Conditioned World Model for Digital TeleoperationWorld models → Embodied and robot world models
Graph Convolutional Attention: A Spectral Perspective on Graph Denoising and DiffusionDiffusion models → Flow matching, time series and science
CAIRN: Cross-Room 3D Scene Understanding with Topology-Aware Large Multimodal ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache CompressionEfficient LLM inference and serving → KV cache optimization
Bridging Physical Reasoning and Task Generalization via Visual Action Outcome Reasoning AlignmentLLM reasoning → Spatial and embodied reasoning +1
FreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM InferenceEfficient LLM inference and serving → KV cache optimization
Point as Skeleton: Accumulated Point Cloud Enhanced Autoregressive Generation for Closed-Loop Autonomous Driving SimulationWorld models → World models for autonomous driving +1
DynaKRAG: A Unified Framework for Learnable Evidence Control in Multi-Hop Retrieval-Augmented GenerationRetrieval-augmented generation → Knowledge-graph and structured-knowledge RAG
Doomed from the Start: Early Abort of LLM Agent Episodes via a Recall-Controlled Probe CascadeLLM agents and tool use → Agent memory systems
AirflowAttack: Thermal-Airflow Adversarial Perturbations against Infrared Remote-Sensing Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Assessing the Operational Impact of Poisoning Attacks over Augmented 3D Point Cloud Public Datasets for Connected and Autonomous VehiclesLLM safety and security → Backdoor and data poisoning attacks
From Voting to Agent Collaboration: Answer-Type-Aware LLM Pipelines for BioASQ 14bLLM agents and tool use → Medical and scientific domain agents
Danus: Orchestrating Mathematical Reasoning Agents with Fact-Graph MemoryLLM reasoning → Mathematical reasoning +1
Analysis-by-Proxy: Localization Signals in VLMs Operating as Condition EncodersLLM reasoning → Spatial and embodied reasoning +2
SIEVE: Structure-Aware Data Selection for Imitation Learning with VLA ModelsVision-language-action models → Human-video and latent-action pretraining +1
TILDE: TILt-based Distributional Erasure for Concept UnlearningDiffusion models → Robustness, privacy and adversarial aspects +1
HoloCount: A Holistic Visual Counting Benchmark for MLLMsLLM reasoning → Multimodal reasoning
An Experimental Design Approach to Evaluating Agentic AI's Autonomous Model DiscoveryLLM agents and tool use → Web, search, and research agents
RuBench: A Repository-Level Agentic Coding Benchmark with Natively Authored Russian Task SpecificationsLLM agents and tool use → Software engineering coding agents
From Foundation to Application: Improving VLA Models in PracticeVision-language-action models → Generalist VLA models and benchmarks
A Definition and Roadmap for World ModelsWorld models → World models for language agents
FADRA: Frequency-Aware Diffusion with Residual Adaptation for Video Face RestorationDiffusion models → Inverse problems and image restoration
VaseMuseum: Digital Intelligent Museum for Ancient Greek PotteryMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
ActionCache: Training-Free Acceleration for Vision-Language-Action Models with Action Caching and RefinementVision-language-action models → Efficient VLA inference and adaptation
Harnessing Code Agents for Automatic Software VerificationLLM agents and tool use → Educational agents and math problem solving
Bridging Diffusion Pruning and Step Distillation with Teacher-Aligned RepairDiffusion models → Diffusion foundations and sampling
Driving the Wrong Way: Leveraging Interpretability in End2End Autonomous Driving ModelsEnd-to-end autonomous driving → End-to-end system architecture
Estimating Uncertainty from Reasoning: A Large-Scale Study of Multi- and Crosslingual MCQA Performance in LLMsLLM reasoning → Test-time compute and efficient thinking
DT-Guard: Intent-Driven Reasoning-Active Training for Reasoning-Free LLM Safety GuardrailLLM reasoning → Reasoning for agents, planning and code +1
AlayaWorld: Long-Horizon and Playable Video World GenerationWorld models → Video-generation world models and simulators
Task Decomposition-Guided Reranking for Adaptive Agent Skill RetrievalLLM agents and tool use → Agent skill learning and evolution
Straight-Path Flow Matching for Incomplete Multi-View ClusteringDiffusion models → Flow matching, time series and science
AgentTether: Graph-Guided Diagnosis and Runtime Intervention for Reliable LLM Agent OperationLLM agents and tool use → Software engineering coding agents
Optimal Transport Q-Learning for Flow Policy Steering and AccelerationVision-language-action models → Efficient VLA inference and adaptation +1
Diagnosing Semantic Handoff Failures in Agent-Orchestrated Vision-Language-Action Skill CompositionVision-language-action models → Failure recovery and robustness +1
VendorBench-100: A Unified Cross-Paradigm Benchmark for Deepfake Image DetectionMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Demonstrating TOFFEE: A Learned System for Synthesizing Data Agent Trajectories at ScaleLLM agents and tool use → Web, search, and research agents +1
Entanglement as a Structural Complexity Axis: A PAC-Bayesian View of Generalization in Quantum Policies and Value FunctionsDeep reinforcement learning methods → Quantum reinforcement learning
Information Gain-based Rollout Policy Optimization: An Adaptive Tree-Structured Rollout Approach for Multi-Turn LLM AgentsLLM agents and tool use → Reinforcement learning for LLM agents
MoWorld: A Flash World ModelWorld models → Video-generation world models and simulators
UBEP: Re-architecting Expert Parallelism Communication Library for Production SuperpodsMixture-of-experts → Efficient MoE training and inference systems

All 421