Papers

New on arXiv

Mon 20 Jul 2026

406 papers

Learning to Stack: Cube-Stacking Imitation Learning from Virtual Reality DemonstrationsOffline RL and imitation learning from data → Robot manipulation imitation learning
RecEvolve: A Knowledge-Driven Autonomous Agent System for Recommender SystemsLLM agents and tool use → Agentic recommender systems
PACE: Publisher-Adaptive Content Extraction via Agentic AutomationLLM agents and tool use → Web, search, and research agents
LLM Safety Alignment in Low-Resource Languages: A Systematic Literature ReviewLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
AutoWorldModel-Bench: A State-Centric Benchmark for Automated World-Model ResearchWorld models → World models for language agents +1
SIRIN: A Unified Toolkit for Detecting Contextual Hallucinations in Retrieval-Augmented and Memory-Grounded LLM SystemsRetrieval-augmented generation → Hallucination detection in RAG +1
CORVUS: Context Optimization and Reduction Via Underlying Synchronization for LLM Coding AgentsLLM agents and tool use → Software engineering coding agents
StepX-Edge: An On-Device UI Vision-Language Model via Architecture-Training-Deployment Co-DesignMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Toward User-Conditioned Evaluation of Personal LLM Agents under Temporal InterventionsLLM agents and tool use → Agent memory systems
MotifRole-Diff: Risk-Optimal Role-Aware Corruption for Masked Molecular Graph DiffusionDiffusion models → Flow matching, time series and science
The Chronos Vulnerability: A Taxonomy of Temporal Persistence and Memory-Based Deception in Agentic AILLM agents and tool use → Agent threats, guardrails and governance
ChainWatch: A Kill Chain-Aligned Sequential Detection Framework for Multi-Step Attacks in MCP-Based AI Agent SystemsLLM agents and tool use → Agent threats, guardrails and governance
ChannelGuard: Safe Models Do Not Compose into Safe Multi-Agent SystemsLLM agents and tool use → Agent threats, guardrails and governance
JailMeter: An Evidence-Based Evaluation Framework for Jailbreak Attacks on Large Language ModelsLLM safety and security → Jailbreaking
Planning as Emergent Behavior in Reinforcement Learning with Relational Hidden StatesDeep reinforcement learning methods → Deep RL algorithms and representations
STeP: Signal Temporal Logic for Precise Specifications for Action Generation with Vision Language ModelsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Attention Without Grounding: Causal Evaluation of Visual Explanations in Medical VLMsMultimodal LLMs / vision-language models → Medical vision-language models
RECEIPT: Deterministic, Reward-Hacking-Resistant Verification for White-Box Agentic XSS DiscoveryLLM agents and tool use → Agent threats, guardrails and governance
For What Reason? Interpreting Models' Encoding of Causation and AntithesisLLM reasoning → Reasoning benchmarks and evaluation
The Story Shapes the Agent: Narrative Priors in LLM BehaviorLLM agents and tool use → Social simulation and multi-agent debate
Integrity-Gated Eco-CACC: Epistemic Admissibility for Cooperative Driving at Signalized IntersectionsEnd-to-end autonomous driving → Cooperative and V2X driving
HALO: Interactive Co-abductive Reasoning in Scientific Hypothesis GenerationLLM reasoning → Clinical and biomedical applications
Engineering Trustworthy Agentic AI for Critical SystemsLLM agents and tool use → Agent threats, guardrails and governance
AniGS: Bridging Rendering and Diffusion Prior for 3D Scene AnimationDiffusion models → 3D, point cloud and scene generation
MAGE: Human-Like Macro Placement via Agentic Multimodal ReasoningLLM reasoning → Spatial and embodied reasoning +1
Reasoning Fine-Tuning Induces Persistent Latent Policy StatesLLM reasoning → Test-time compute and efficient thinking
EduPanel: A Three-Agent LLM Judge for Teaching Videos -- Reliability, Complementarity, and Human Trust CalibrationLLM agents and tool use → Educational agents and math problem solving
Signed Rectified Flow: Negativity-Controlled GenerationDiffusion models → Flow matching, time series and science
DuSPiT: Dual-Branch Sub-Patch Pixel Diffusion TransformerDiffusion models → Diffusion foundations and sampling
The Open Ant: A Robot Platform for Reinforcement Learning ResearchDeep reinforcement learning methods → RL for robot locomotion and manipulation
Trusted Credentials, Untrusted Behavior: Benchmarking LLM-Agent Security in High-Performance ComputingLLM agents and tool use → Agent threats, guardrails and governance
RRPO: Reference-Relative Policy Optimization with Stratified Conditional RolloutsLLM reasoning → Reinforcement learning for reasoning
ChainMark: Model-Free LLM Watermarking with Closed-Form CalibrationLLM safety and security → Watermarking generated content and models
Computational models of pragmatic reasoning with flexible generation of meaning and expression alternativesLLM reasoning → Reasoning benchmarks and evaluation
Relay-Bench: Evaluating LLMs on Multi-Domain Reasoning ChainsLLM reasoning → Reasoning benchmarks and evaluation
Neuro-Symbolic Meta-Policies for Temporal Knowledge-Graph Memory under Partial ObservabilityDeep reinforcement learning methods → Interpretable and logic-guided RL
AlayaWorld: Interactive Long-Horizon World Modeling -- Full Technical ReportWorld models → Video-generation world models and simulators
Operational Hallucination and Safety Drift in AI AgentsLLM agents and tool use → Agent threats, guardrails and governance
Towards Torque-Driven Reinforcement Learning for Quadruped LocomotionDeep reinforcement learning methods → RL for robot locomotion and manipulation
A Classifier That Teaches Itself: Self-Improving, Frozen-gate Training (SIFT) for Dynamic Document ClassificationSynthetic data and self-improvement for LLMs → Self-training for few-shot NLP
A Decision-Centered Reference Architecture for Trustworthy Agentic CommerceLLM agents and tool use → Agent threats, guardrails and governance
Addressing Limited Data in Auditory Attention Decoding with Diffusion Generative ModelsDiffusion models → Medical imaging and segmentation
FM-VLA: Force-based Memory for Vision-Language-Action Models in Contact-Rich ManipulationVision-language-action models → Tactile, force-aware and humanoid VLA
Simple Domain Generalization for Strong Pixel-Level Image Tampering Detection in Modern VLMsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Logical Judgments Under Pressure: Diagnosing Syllogistic Stability with Learned Soft PrefixesMixture-of-experts → MoE large language models
Vector Search As Nearest Neighbor Matching: RAG-based Policy Learning in Causal InferenceRetrieval-augmented generation → Agentic RAG, agents and memory
SWE-Pruner Pro: The Coder LLM Already Knows What to PruneLLM agents and tool use → Software engineering coding agents
PPL-Factory: Task-Aware and Budget-Aware Data Selection from Language Modeling to ReasoningLLM reasoning → Reasoning benchmarks and evaluation
FlashRT: Agent Harness for Guiding Agents to Deploy Real-Time Multimodal ApplicationsLLM agents and tool use → Multimodal and embodied agents
OR Else: A Differentiable Trust Region for Policy OptimizationLLM reasoning → Reinforcement learning for reasoning
TRIM: Reducing AI-Generated CodeSlop via Agent Trajectory MinimizationLLM agents and tool use → Software engineering coding agents
Testing Retrieval-Augmented Generation Systems with Chunk CoverageRetrieval-augmented generation → RAG pipeline design and evaluation
jina-reranker-v3.5: An Efficient Listwise Reranker with Hybrid Attention and Self-DistillationLLM agents and tool use → Web, search, and research agents
LLMs and Agentic AI Systems for Smart Grids: A Tutorial on Architectures and ApplicationsLLM agents and tool use → Financial trading and forecasting agents
Do Language Models Dream of Binding Molecules? Benchmarking LLMs under Spatial ConstraintsDiffusion models → Flow matching, time series and science
A CXL Memory Rack for Multi-Turn LLM ServingEfficient LLM inference and serving → KV cache optimization
Isaac Sim-to-Real: Reinforcement Learning based Locomotion for QuadrupedsDeep reinforcement learning methods → RL for robot locomotion and manipulation
Importance Sampling and PCA for Finding Failures in Commercial Autonomous VehiclesEnd-to-end autonomous driving → Adversarial scenarios and safety testing
FinSAgent: Corpus-Aligned Multi-Agent RAG Framework for Evidence-Grounded SEC Filing Question AnsweringRetrieval-augmented generation → Legal and financial RAG
Can We Break LLMs Out of Self-Loops? Fine-Grained Reasoning Control with Activation SteeringLLM reasoning → Test-time compute and efficient thinking +1
WorldCupArena: Fine-Grained Evaluation of Language Models and Deep-Research Agents on Football ForecastingLLM agents and tool use → Financial trading and forecasting agents
CriPO: Enhancing Rubric-based RL via Self-DistillationKnowledge distillation → On-policy and self-distillation for LLMs
Sparse Evidence Can Suffice: Agentic Evidence Seeking for Multimodal Video Misinformation DetectionLLM agents and tool use → Web, search, and research agents
Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in PancasilaLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
Autoresearch with Coding Agents: Generalizers and Metric-Maximizers on Quran Recitation DataLLM agents and tool use → Software engineering coding agents
Adaptive Adversaries: A Multi-Turn, Multi-LLM Benchmark for LLM Agent SecurityLLM agents and tool use → Agent threats, guardrails and governance +1
Test Coverage Analysis of Agentic Pull RequestsLLM agents and tool use → Software engineering coding agents
An Early Warning of Emerging Biosecurity Risks in Frontier LLMsLLM safety and security → Jailbreaking
SEE: Structure-aware Exploring & Exploiting for Long-horizon GUI Agent Trajectory SynthesisLLM agents and tool use → GUI and mobile agents
Adaptive Mamba Neural OperatorsState-space sequence models (S4, Mamba) → Time series forecasting with SSMs
Anticipate Before Acting: Future-State-Conditioned Vision-Language NavigationMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
Evidence-in-the-Loop: Trace-Driven Optimization for Customer-Service LLM AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Rethinking Heterogeneous LLM Merging: A Weighted Model Averaging PerspectiveLLM reasoning → Test-time compute and efficient thinking
Closing the Loop in Humanoid VLA: Persistent 3D Object Tokens for Verifiable Loco-ManipulationVision-language-action models → Tactile, force-aware and humanoid VLA
MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for Parameter-Efficient Reasoning in Compact ModelsLLM reasoning → Reasoning for agents, planning and code +1
PAMD: Structured Adaptive Distances for Bisimulation Representations in Visual Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
ExpertPlex: A High-Goodput Disaggregated Serving System for MoE LLMs with Adaptive Persistent KernelsMixture-of-experts → Efficient MoE training and inference systems +1
HAS: Highlight-guided Attention Steering for Multimodal LLM Video SummarizationMultimodal LLMs / vision-language models → Video understanding with MLLMs
Information-Based Exploration via Random Features for Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
SAGE: Subgoal-Conditioned Action Generation for Latent World Model PlanningWorld models → Latent predictive world models
DiFA: Inference-Time Forward-Process Alignment for Diffusion ModelsDiffusion models → Diffusion foundations and sampling
RT-SHCUA: Real-Time Self-Hosted Computer-Use Agent for UAV ControlLLM agents and tool use → Multimodal and embodied agents
Towards Agentic Agent-based Models: Feasibility, Performance, and Statistical Model CheckingLLM agents and tool use → General LLM agent frameworks and benchmarks
A Geometric Perspective on Stabilizing Value Conflict ResolutionLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
When and How Context Rot Appears in Coding Agents: A White-Box Study of Agent Skills in Code AuditingLLM agents and tool use → Software engineering coding agents
PEARL: Auditable Repair for Scientific Reasoning Graph ExtractionLLM reasoning → Knowledge- and tool-augmented reasoning
Distributional Soft Bellman Operator under the Cram'er GeometryDeep reinforcement learning methods → RL theory and sample complexity
Stress Testing Concept Erasure with Large Language Model AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
ST-Veto: Spatio-Temporal Token Veto for Diffusion MLLMs via Taylor Prediction and Visual GroundingLLM reasoning → Multimodal reasoning +1
When a Name Is Not a Name: A Benchmark Dataset and Distilled Reasoning for Culturally Entangled Bangla Homographs in Low-Resource LLMsLLM reasoning → Reasoning benchmarks and evaluation +1
Theoretical Foundations of max@k Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning +1
PRiSM: Prototype Regularization for Few-Shot VLMsMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
Reasoning as a Double-Edged Sword: Architecture and Cross-Stage Robustness in Vision-Language-Action ModelsLLM reasoning → Spatial and embodied reasoning +1
Dynamic Defense Profiling Enables Cognitive Jailbreak of Text-to-Image ModelsLLM safety and security → Vision-language and text-to-image safety
Measuring Monosemanticity in Sparse Autoencoders via Latent Activation CoherenceMechanistic interpretability → SAE features and their applications
FIFA World Cup 2026 as a Contamination-Free Benchmark for LLM Forecasting Agents: Four Models, a Bookmaker, and 104 MatchesLLM agents and tool use → Financial trading and forecasting agents
Generalize and Guide: Decomposing Rewards for Few-Shot Inverse Reinforcement LearningOffline RL and imitation learning from data → Inverse reinforcement learning +1
Semantically Similar, Logically Distinct: Diagnosing the Semantic-Answerability Gap in Table RAGRetrieval-augmented generation → RAG robustness and adaptive retrieval
MXSens: Sensitivity-Aware Mixed-Precision Quantization for Efficient LLM InferenceEfficient LLM inference and serving → KV cache quantization
SR-Agent: An Experience-Driven Agentic Framework for Post-Ranking Strategy Refinement in E-Commerce RecommendationLLM agents and tool use → Agentic recommender systems

All 406