Papers

New on arXiv

Wed 8 Jul 2026

362 papers

Lexical Prompt Compression for Large Language Models: A Training-Free, Deterministic Pipeline with Empirical Pareto Analysis Across Eleven Task CategoriesEfficient LLM inference and serving → Prompt and context compression
Cognitive Cells: A Compositional Framework for Populations of Small Language ModelsLLM agents and tool use → General LLM agent frameworks and benchmarks
CityReal: Human-Aligned Urban Behavior and City Dynamics Simulation with Large-Scale LLM AgentsLLM agents and tool use → Multimodal and embodied agents
When Do LLMs Apply the Wrong Law? Diagnosing LLM Failures in Temporal Legal ReasoningLLM reasoning → Reasoning for agents, planning and code
Representation Robustness Under Executable Reasoning Constraints in Large Language Models for Mathematical Problem SolvingLLM reasoning → Mathematical reasoning
Semantic Drift and the Stability of Operator Control in Reasoning-Class Decision Support SystemsLLM reasoning → Reasoning benchmarks and evaluation
Adversarially Guided Diffusion for LiDAR Range Image SynthesisWorld models → World modeling from LiDAR generation +1
Length Penalties Make Chain-of-Thought Less MonitorableLLM reasoning → Test-time compute and efficient thinking
HELP: Human-Efficient Large-Scale Robot Post-Training with Rollout SegmentationVision-language-action models → Reinforcement learning and post-training
GReFEM: Multimodal LLMs as Zero-Shot Semantic Assistants for Physics-Guided 3D Mesh RefinementMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
EgoWAM: World Action Models Beyond Pixels with In-the-Wild Egocentric Human DataWorld models → Embodied and robot world models
Hallucination Self-Play: Bootstrapping Reinforced Detector via Evolved GeneratorLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment +1
Who Broke the System? Failure Localization in LLM-Based Multi-Agent SystemsLLM agents and tool use → Software engineering coding agents
Agentic Neural Architecture SearchLLM agents and tool use → General LLM agent frameworks and benchmarks
3100 Opinions on Code Review in an AI World: Building Causal Theory from Practitioner DiscourseLLM agents and tool use → Software engineering coding agents
When Implausible Tokens Get Reinforced: Tail-Aware Credit Calibration for LLM Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning
Expressivity and Statistical Trade-offs in Diffusion Policy LearningDiffusion models → Diffusion and flow policies for robotics +1
DeepSWE: Measuring Frontier Coding Agents on Original, Long-Horizon Engineering TasksLLM agents and tool use → Software engineering coding agents
Efficient Safety Alignment of Language Models via Latent Personality TraitsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Multimodal Unlearning Across Vision, Language, Video, and Audio: Survey of Methods, Datasets, and BenchmarksLLM safety and security → Machine unlearning
Mechanistic Interpretability of LLM Jailbreaks via Internal Attribution GraphsLLM safety and security → Jailbreaking +1
Reachability-Preserving Bellman Operator for the Discounted Reach-Cost Value Function: Uniting Hamilton-Jacobi Reachability and Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
Nigeria Machinery: A Low-Resource Industrial Dataset with a Domain-Grounded Reasoning LayerLLM reasoning → Reasoning benchmarks and evaluation
Feedback Manipulation Regularization: Enabling Offline Agent Alignment for Imitation LearningLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF +1
Agentic AI and Retrieval-Augmented Models in Straight-Through UnderwritingRetrieval-augmented generation → Legal and financial RAG +1
Multi-agent Autoformalization of Tensor Network TheoryLLM agents and tool use → Educational agents and math problem solving
When Does Continual Learning Require LearningEfficient LLM inference and serving → Efficient reasoning and agent inference
Shift & Drift: A Zero-Shot Benchmark for Generalizable and Robust Autonomous Driving Motion PlanningEnd-to-end autonomous driving → End-to-end system architecture
Infinity-Parser2 Technical ReportMultimodal LLMs / vision-language models → Document understanding and multimodal RAG +1
Open Models, Open Risks: Measuring Unsafe Generation in Text-to-Image Models In the WildLLM safety and security → Vision-language and text-to-image safety
From Triggers to Emotions: A CPM-Grounded Appraisal Multi-Agent for Dynamic Emotional Evolution in Persona-Based DialogueLLM agents and tool use → Social simulation and multi-agent debate
DeepSearch-World: Self-Distillation for Deep Search Agents in a Verifiable EnvironmentLLM agents and tool use → Web, search, and research agents +2
ScopeJudge: Cost-Aware Pre-Execution Gating for Offensive Security AgentsLLM agents and tool use → Agent threats, guardrails and governance
Principled Analysis of Deep Reinforcement Learning Evaluation and Design ParadigmsDeep reinforcement learning methods → Deep RL algorithms and representations
Unlocking Temporal Generalization in Hamiltonian Video Dynamics ModelsWorld models → Latent predictive world models
Aligning Clinical Needs and AI Capabilities: A Survey on LLMs for Medical ReasoningLLM reasoning → Clinical and biomedical applications
Image classification via a quantum-inspired strategy involving a mixture of expertsMixture-of-experts → MoE theory and gating functions
Selective Left-Shift: Turning Test-Time Compute and Difficulty-based Curation into Training Data for Low-Resource Code GenerationLLM reasoning → Reasoning for agents, planning and code
PERFOPT-Bench: Evaluating Coding Agents on Software Performance OptimizationLLM agents and tool use → Software engineering coding agents
Jet-Long: Efficient Long-Context Extension with Dynamic Bifocal RoPERetrieval-augmented generation → Generation and context integration +1
From Noisy Traces to Root Causes: Structural Trajectory Analysis and Causal Extraction for Agent OptimizationLLM agents and tool use → Software engineering coding agents
Institutional Red-Teaming: Deployment Rules, Not Just Models, Causally Shape Multi-Agent AI SafetyLLM agents and tool use → Agent threats, guardrails and governance +1
Selective Timestep Weighting and Advantage-Based Replay for Sample-Efficient Diffusion RLHFLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Agent Delivery Engineering Predictive Reliability FrameworkLLM agents and tool use → Software engineering coding agents
SkillCenter: A Large-Scale Source-Grounded Skill Library for Autonomous AI AgentsLLM agents and tool use → Agent skill learning and evolution
Scaling Mixture-of-Experts Video Pretraining for Embodied IntelligenceMixture-of-experts → MoE in reinforcement learning and robotics
A hierarchical memory architecture overcomes context limits in long-horizon multi-agent computational modelingLLM reasoning → Reasoning for agents, planning and code +1
Guidance Breaks the Fitted Operator: A Terminal-Fitted Repair for Classifier-Free GuidanceDiffusion models → Diffusion foundations and sampling
Unlearning to Protect: A Distilled Reinforcement Learning Framework with Privacy-Preserving Feature Unlearning and XAI for IoT SecurityLLM safety and security → Machine unlearning
Rethinking Code Performance Benchmarks for LLMsLLM reasoning → Reasoning for agents, planning and code
Towards Agentic AI Governance: A Preliminary AssessmentLLM agents and tool use → General LLM agent frameworks and benchmarks
Dual Latent Memory in Vision-Language-Action Models for Robotic ManipulationVision-language-action models → Reasoning and chain-of-thought in VLA
CARLA-GS: Decoupling Representation, Reasoning, and Physics Simulation for Autonomous Driving Corner-Case SynthesisWorld models → World models for autonomous driving +2
What Makes a Good Bug Report for an AI Agent?LLM agents and tool use → Software engineering coding agents
Think Big, Search Small: Where Capacity Matters in Hierarchical Search Agents?LLM reasoning → Knowledge- and tool-augmented reasoning +1
A Unified Detection Framework for AI-Related Content and ArtifactsLLM safety and security → Watermarking generated content and models
Single-Rollout Asynchronous Optimization for Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents
HIVE: Understanding Post-Hallucination Reasoning in Vision Language ModelsLLM reasoning → Hallucination detection and mitigation +1
Do LLM-Generated Skills Make Better AI Data Scientists? A Component Ablation Across Data-Science WorkflowsLLM agents and tool use → Agent skill learning and evolution
Beyond Attack-Success Rate: Action-Graded Severity Scale for Tool-Using AI AgentsLLM agents and tool use → Agent threats, guardrails and governance +1
SpaCellAgent: A Self-Evolving LLM-Based Multi-Agent Framework for Trajectory AnalysisLLM agents and tool use → Medical and scientific domain agents +1
Mitigating Taint-Style Vulnerabilities in MCP Servers via Security-Aware Tool DescriptionsLLM agents and tool use → Agent threats, guardrails and governance
EmbodiedGen V2: An Agentic, Simulation-Ready 3D World Engine for Embodied AILLM agents and tool use → Multimodal and embodied agents
The Blind Curator: How a Biased Judge Silently Disables Skill Retirement in Self-Evolving AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Beware of Agentic Botnets: Scalable Untargeted Promptware Attacks via Universal and Transferable Adversarial HalluSquattingLLM agents and tool use → Agent threats, guardrails and governance +1
The Optimal Sample Complexity of Learning Autoregressive Chain-of-ThoughtLLM reasoning → Prompting and chain-of-thought reasoning
DeLS-Spec: Decoupled Long-Short Contexts for Parallel Speculative DraftingEfficient LLM inference and serving → Speculative decoding
Reason Less, Verify More: Deterministic Gates Recover a Silent Policy-Violation Failure Mode in Tool-Using LLM AgentsLLM reasoning → Reasoning for agents, planning and code +1
Multi-Agent Robotic Control with Onboard Vision-Language ModelsLLM agents and tool use → Multimodal and embodied agents
Agentic Data EnvironmentsLLM agents and tool use → Agent threats, guardrails and governance
When Prompts Ignore Structure: Graph-Based Attribute Reasoning for Calibrated VLMsLLM reasoning → Multimodal reasoning +1
MIRA-Math: A Benchmark for Minimal Information Requesting and Mathematical ReasoningLLM reasoning → Mathematical reasoning
A Large Language Model-Driven Agent-Based Modeling Framework with Multi-Round Communication for Simulating Vaccine Opinion DynamicsLLM agents and tool use → Social simulation and multi-agent debate
MMAgent-R2: Learning to Rerank and Reject for Agentic mRAGLLM agents and tool use → Multimodal and embodied agents
Interpretable Uncertainty for Adaptive Retrieval and Reasoning in Question AnsweringLLM reasoning → Knowledge- and tool-augmented reasoning
Physics-Audited Agentic Discovery in Scientific Machine LearningLLM agents and tool use → Web, search, and research agents
On Adversarial Vulnerability of Vision-Language Models through the Lens of Intermediate Spectral SubspacesMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
Improving greenhouse fruit-production control by integrating reinforcement learning into short-horizon model predictive controlDeep reinforcement learning methods → RL for energy, networks and operations
BUS: Brain-Inspired Unsupervised Self-Reflection via Backward Prediction for Multimodal ReasoningLLM reasoning → Multimodal reasoning +1
HumAIN: Human-Aware Implicit Social Robot NavigationKnowledge distillation → Multimodal distillation
From Atomic Actions to Standard Operating Procedures: Iterative Tool Optimization for Self-Evolving LLM AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks +1
Mechanistic Interpretability for Neural Networks: Circuits, Sparse Features and Symbolic ReasoningLLM reasoning → Reasoning benchmarks and evaluation +1
InfraQR: Edge-Placed QR-Inspired Structured Patch Attacks on Infrared Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
FMMVCC: Fuzzy Mamba-based Multi-View Contrastive Clustering for Univariate Time SeriesState-space sequence models (S4, Mamba) → Time series forecasting with SSMs
Safe Reinforcement Learning using Ideas from Model Predictive ControlDeep reinforcement learning methods → Safe and constrained RL
Evaluation of Multilingual Ability to Use Spatial Deictic Expressions in Vision-Language ModelsLLM reasoning → Spatial and embodied reasoning +1
ORCAID: Oblique Rule-Based Continuous-Action Interpretation for Deep RL PoliciesDeep reinforcement learning methods → Interpretable and logic-guided RL
HPG-Diff: Hierarchical physics-guided diffusion with differentiable connectivity constraints for topology optimizationDiffusion models → Flow matching, time series and science
DiPhon: Diffusion on Graphons for Scalable Graph GenerationDiffusion models → Flow matching, time series and science
Reasoning Consistency Scanning: A Framework for Auditing Chain-of-Thought Validity in AI Safety EvaluationsLLM reasoning → Prompting and chain-of-thought reasoning
Memory Scarcity, Open Models, and the Restructuring of the AI Industry, 2026-2030 -- A quantitative scenario analysis of inference economics, training-cost divergence, and infrastructure solvencyEfficient LLM inference and serving → Serving systems and request scheduling
Validate the Dream Before You Trust Its Verdict: Admissibility for World-Model SimulatorsWorld models → Embodied and robot world models
Does AI Understand Imaging? A Systematic Benchmark of Agentic AI for Computational Imaging TasksMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning +1
Predicting LLM Safety Before Release by Simulating DeploymentLLM safety and security → Safety alignment, refusal and guardrails
Entropy Pacing Policy Optimization for Multi-Task Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents
Stage-Aware Adaptation and Distribution Calibration for Subject-Driven Personalized Text-to-Image GenerationDiffusion models → Image generation, text-to-image and editing
NoDrift3R: Raymap-Guided Coupling for Drift-Robust Unposed Feed-Forward 3D ReconstructionWorld models → World models for autonomous driving
Fractal KV-Cache Archives: Lossless Symbolic Storage with In-Place Retrieval for Long-Context LLM InferenceEfficient LLM inference and serving → KV cache quantization
ColorFM: An Optimization-to-Learning Framework for Color Transfer via Flow MatchingDiffusion models → Image generation, text-to-image and editing
Tree-of-Thoughts Reasoning for Text-to-Image In-Context LearningLLM reasoning → Multimodal reasoning

All 362