Papers

New on arXiv

Fri 7 Aug 2026

418 papers

StepTrigger: Contact-State-Triggered Backdoor Attacks on VLM-Powered Legged RobotsLLM safety and security → Prompt injection and agent security
Enhancing Audio Reasoning via Semantic Summary PredictionLLM reasoning → Audio-language model reasoning
StableEval Arena: A Cost-Aware Agentic Benchmark for Stablecoin Price Stability PredictionLLM agents and tool use → Financial trading and forecasting agents
Selective Cross-View Consistency for World Action Models: Held-Out Viewpoint Robustness Without Test-Time Camera InformationWorld models → Embodied and robot world models
Glance, Scrutinize, and Think: Advancing Video Anomaly Detection from Training-Free to Agentic ReasoningLLM reasoning → Chain-of-thought reasoning for driving
Preserving Item Semantics for Free: Rethinking Token Initialization in LLM-Based Generative RecommendationLLM reasoning → LLM reasoning for recommendation
Shape Mutating Expert Compression:LorExperts and BTExpertsMixture-of-experts → MoE large language models
When the Judge Should Not Decide: Evidence-Locked, Non-Compensatory Selection Bounds LLM-Judge Failure in Reasoning PipelinesLLM reasoning → Test-time compute and efficient thinking
CausalNav: Reliability-Certified Causal World Models for Control under Physical-Parameter ShiftWorld models → Latent dynamics models and model-based RL
The Anatomy of a Prompt Injection: A Component Model for Structured AnalysisLLM safety and security → Prompt injection and agent security
CliniCARE-Bench: Clinical Calibrated Audit of Medical Reasoning in EHRLLM reasoning → Clinical and biomedical applications
AndroidReality: How Far Are Mobile Agents from the Real World?LLM agents and tool use → GUI and mobile agents
The Sample Complexity of Policy Learning with Mu-ResetsDeep reinforcement learning methods → RL theory and sample complexity
Jako Tako or Fluent? Presenting PoVisLE: A Polish Vision-Language EvaluationMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
LUCID: Latent-Skill Unified Control via Imagined Dynamics for Long-Horizon Humanoid Loco-ManipulationWorld models → Humanoid and legged locomotion world models
QuantumMind: Constraint-Grounded Agentic Reasoning for Speedup Analysis in Quantum ComputingLLM agents and tool use → Medical and scientific domain agents
BRUCE: Benchmarking Robustness Under Corruption Escalation for Scientific Vision-Language ReasoningLLM reasoning → Multimodal reasoning +1
Finite Constant Frontiers and Auditable Regret Certificates for Average-Reward Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
CODS: Iterative Bellman-Residual Data Selection for Reusable Offline Reinforcement LearningOffline RL and imitation learning from data → Offline RL algorithms
SpikeWorld: Fast-State Adaptation for Frozen Spiking World ModelsWorld models → Embodied and robot world models
Keep It Simple: Multi-Key Episodic Memory Retrieval for Ultra-Long Video UnderstandingMultimodal LLMs / vision-language models → Video understanding with MLLMs
An Agentic AI Framework Overcomes Fundamental Limitations of Large Language Models for Glaucoma Detection from Fundus PhotographyLLM agents and tool use → Medical and scientific domain agents
Mendel G"odel Machine: Recursive Self-Improving Coding Agents via Comparative EvolutionLLM agents and tool use → Software engineering coding agents
Contextual Value Alignment via Multilayer Combinatorial FusionLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
SkillConsist: Detecting Inconsistencies in Agent Skills via Bidirectional Graph AlignmentLLM agents and tool use → Agent skill learning and evolution
Agent-MD: Selective LLM Intervention with Event-Driven Escalation for Stateful GCMC--MD CampaignsLLM agents and tool use → Medical and scientific domain agents
From Single Chatbots to Governed Agent Ecosystems: An Agentic AI Pattern Catalogue and Orchestration Framework for Mission-Critical Hospital Information Management SystemsLLM agents and tool use → Medical and scientific domain agents
Controlled Memory Interference in Continual LLM AgentsLLM agents and tool use → Agent memory systems
CMU-Drive and V2V-VLA: Cooperative Multi-agent Unified Driving with Reasoning Benchmark and Vehicle-to-Vehicle Vision-Language-Action ModelsVision-language-action models → VLA for autonomous driving +1
FlowErase-OPD: Multi-Concept Erasure via Anchored On-Policy Distillation in Flow Matching ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
GWM-VLA: Geometry-Aware Latent World Modeling for Vision-Language-Action LearningWorld models → Embodied and robot world models +1
SimWAM: A Simple World Action Model for End-to-End Autonomous DrivingWorld models → World models for autonomous driving +1
MirrorWorld: Taming Video Diffusion Models for Mirror Reflection GenerationDiffusion models → 3D, point cloud and scene generation
CoinRAG: Contextualized Information Nugget KV Cache Reuse for Long-Context RAGRetrieval-augmented generation → Generation and context integration +1
SkillProx: Self-Evolving Agent Skills via Proximal Textual Gradient DescentLLM agents and tool use → Agent skill learning and evolution +1
Blast RadiusLLM agents and tool use → Agent memory systems
PsychoAgent: An Affect-Sensitive Cognitive Architecture for Conflict-Aware Memory in LLM AgentsLLM agents and tool use → Social simulation and multi-agent debate
Fisher-R1: Training LLM Agents for Reliable Hypothesis TestingLLM agents and tool use → Web, search, and research agents
SABRE: Scalable and Automated Benchmarking of VLMs under StressMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Diffusion LLMs as Targets and Adversaries: Mechanistic Safety ExploitsLLM safety and security → Jailbreaking
A Picture is Worth a Thousand Tokens: How Vision Language Models Cut AI Energy Costs While Improving AccuracyMultimodal LLMs / vision-language models → Visual token pruning and efficiency
CoBa: Cost-Effective Test-Time Scaling via Compute-Balanced RoutingLLM reasoning → Test-time compute and efficient thinking
Beyond Myopic World Models: Long-Horizon End-to-End Training for Direct Future PredictionWorld models → Latent predictive world models
ResidencyRL: Reinforcement Learning in Simulated Clinical EnvironmentsDeep reinforcement learning methods → RL for clinical treatment and surgery
I Seek You in Videos: Identity-Conditioned Queries for Person-Centric Video ReasoningLLM reasoning → Multimodal reasoning
UniJEPA: A Unified Joint-Embedding Predictive Architecture for Task-Agnostic Visual World ModelingWorld models → Latent predictive world models
Addressable Memory for Video World ModelsWorld models → Video-generation world models and simulators
Trajectory-Relative Hindsight Distillation for Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents
LitTraceQA: A Benchmark for Multi-Stage Grounding and Verification in Scientific Question AnsweringRetrieval-augmented generation → Agentic RAG, agents and memory
Depth-Wise Probing and Pruning of the Planning Token in a Driving Vision-Language-Action ModelVision-language-action models → VLA for autonomous driving +1
A2E : An End-to-End Agent Auditing EngineLLM agents and tool use → General LLM agent frameworks and benchmarks
From Forensics to Ecosystems: Rethinking Watermarks for Generative AI OversightLLM safety and security → Watermarking generated content and models
A Progressive Design Study of Visual Encoders and Value Estimation for Replay-Free Parallelized Q-LearningDeep reinforcement learning methods → Deep RL algorithms and representations
Learning Fault-Tolerant Locomotion with Adaptive Gait TimingDeep reinforcement learning methods → RL for robot locomotion and manipulation
TEMPO: Semantic-Action Decoupled RL Post-Training for Vision-Language-Action ModelsVision-language-action models → Reinforcement learning and post-training +1
Same Attention, Different Truths: Put Logit-Lens over Visual Attention to Detect and Mitigate LVLM Object HallucinationMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
From Optimal Actions to World Models: Identifiability of Transition Kernels in Discounted MDPsWorld models → Latent dynamics models and model-based RL
FUSE: Feature-Wise Unified Specialization with Cross-Column Exchange for Mixed-Type Tabular Flow MatchingDiffusion models → Flow matching, time series and science
Why Study Emergent Behavior When You Can Regulate It? Aligning Multi-Agent Systems with Reward PredictionLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
WNM-3D: A World Navigation Model with 3D Scene Conditioning for Closed-Loop VLNMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Finite-Sample Metric Non-Collapse for Geometrically Supervised Latent World Models in ControlWorld models → Latent predictive world models
Inverse reinforcement learning for indefinite mean-field social optimization with multiplicative noiseDeep reinforcement learning methods → RL for robot locomotion and manipulation
Stoicheia: Character-Level Masked Diffusion for Ancient Greek Textual Restoration, Parsing, and Metrical ScansionDiffusion models → Discrete and text diffusion
Learning Suffers More Than the Policy Class Under Partial Observability: A Closed-Form AnalysisDeep reinforcement learning methods → RL theory and sample complexity
Toward a Causal Data Management Ecosystem for Decision Making and Agentic AILLM agents and tool use → Medical and scientific domain agents
From Test-Time Scaling to Reusable Memory: Measuring Crystallization in Text-to-SQLLLM reasoning → Knowledge- and tool-augmented reasoning
Measuring Concept Content in Text from LLM Activations: ESG Evidence from Concept Vectors and Linear ProbesMechanistic interpretability → SAE features and their applications
Flow-Corrected Shape Optimization: Taming Manifold Drift in High-Dimensional 3D ModelsDiffusion models → 3D, point cloud and scene generation
EMAS: Stabilizing Multi-Agent System Evolution through Evidence-Guided RevisionLLM agents and tool use → General LLM agent frameworks and benchmarks
An AI4AI Framework for Visual Token PruningMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Momba: Network Modernization Improves Multi-Objective Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
Representation-driven Endoscopic Visual Embedding Alignment for Latent GenerationDiffusion models → Medical imaging and segmentation
Agent Memory Distillation: Empowering Small LLM Agents with Hierarchical Teacher MemoryLLM agents and tool use → Agent memory systems
NiyamAI - An Intent-Bound AI Agent with Cryptographically Verifiable Guardrails using Zero-Knowledge ProofsLLM agents and tool use → Agent threats, guardrails and governance +1
Fluid-DiT: Graph-Free Diffusion Transformers for Fluid Flow Simulations LearningDiffusion models → Flow matching, time series and science
Exact Adaptive Hybrid Retrieval Without Fixed Top-L CutoffsRetrieval-augmented generation → Generation and context integration
Interpretable reinforcement learning with decision-tree pruningDeep reinforcement learning methods → Interpretable and logic-guided RL
DiDPO: Diff-in-Diff Policy Optimization for Coding Agent TrainingLLM agents and tool use → Software engineering coding agents
Human-AI Perceptual Alignment by Playing Hues and CuesMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
PHOENIX: Fine-Tuned SLM-Powered Autonomous Satellite Lifetime Extension via Predictive Self-Healing and Multi-Agent AI RecoveryLLM agents and tool use → Multimodal and embodied agents
How Much, Then Where: Credit-Conserving Action-to-Token Allocation for Multi-Turn Agent Reinforcement LearningDeep reinforcement learning methods → RL training of language and agent models
Beyond Fluency: A Clinical Benchmark and Anomaly-Enhanced Baseline for Spine MRI Report GenerationMultimodal LLMs / vision-language models → Medical vision-language models
MemWM: Memory-Augmented Text-Based World ModelWorld models → World models for language agents
RoRA: Role-Oriented Regional Allocation for Visual Token Pruning in MLLMsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Beyond Isolation: Unlocking Reinforcement Learning Component Synergy for Sample-Efficient Continuous ControlDeep reinforcement learning methods → Deep RL algorithms and representations
Transformers Struggle to Use Their Emergent World Models: Revisiting the Tower of Hanoi, and the Illusion of ThinkingWorld models → World models for language agents +1
MemOPD: On-Policy Distillation through Memory State Alignment for Long-Horizon AgentsKnowledge distillation → On-policy and self-distillation for LLMs
AutoIntervene: Calibrated Intervention for Action-Chunking Imitation Learning PoliciesOffline RL and imitation learning from data → Robot manipulation imitation learning
Limit Points of Reflow with Minibatch Optimal TransportDiffusion models → Flow matching, time series and science
An Agentic Hybrid Top-Down and Bottom-Up Approach to Knowledge Graph GenerationLLM agents and tool use → Web, search, and research agents
HiSparse: Scaling Sparse-Attention Decoding with Hierarchical KV Cache ManagementEfficient LLM inference and serving → KV cache optimization
Does More Retrieved Evidence Help Visual Retrieval-Augmented Generation with Diffusion Language Models?Retrieval-augmented generation → RAG robustness and adaptive retrieval +1
Benchmarking and Reasoning Distillation of Large Language Models for Feedback Controller Design in Complex Dynamical SystemsLLM reasoning → Test-time compute and efficient thinking
HRDiT: Training-Free High-Resolution Image Generation with Off-the-Shelf Diffusion Transformer ModelsDiffusion models → Diffusion foundations and sampling
Every Cache Entry Earns Its Place: Global Allocation of Resolution and Coverage for KV Cache CompressionEfficient LLM inference and serving → KV cache optimization
Decoupling Intention from Trajectory: A Representational Deduction Framework for World Action ModelsWorld models → Embodied and robot world models
Local Epistemic Uncertainty Guided Active Sampling for Plug-and-play Diffusive Image RestorationDiffusion models → Inverse problems and image restoration
Finding Usable Weight Mechanisms with Tiled SVDMechanistic interpretability → Circuits and mechanistic analysis
How a shared state is described determines whether AI agents synchronizeLLM agents and tool use → Social simulation and multi-agent debate
Cross-View Action Consistency for Camera-Robust Vision-Language-Action PoliciesVision-language-action models → 3D and spatial-aware VLA +1

All 418