Papers

New on arXiv

Fri 24 Jul 2026

307 papers

ResumeShield: Channel Separation and an Open Benchmark for Indirect Prompt Injection in AI Resume ScreeningLLM safety and security → Prompt injection and agent security
Reflective Recovery: A Self-Supervised Method for Reasoning by Learning from MistakesLLM reasoning → Test-time compute and efficient thinking
DeMTS: Denoising Trajectories as Multivariate Time Series for Hallucination Detection in Diffusion Language ModelsDiffusion models → Medical imaging and segmentation
Inference-Time Mitigation of Adversarial Political Bias in Large Language ModelsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment +1
Dynamic Coalition Formation and Communication Pricing in Skill-Based Agentic AI SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
Search-G1: Grounded Search Agents via Representation-Based Intrinsic RewardsLLM agents and tool use → Web, search, and research agents
NeuroSynth: A Biologically Inspired Continual Reinforcement Learning Architecture for Mitigating Catastrophic ForgettingDeep reinforcement learning methods → Deep RL algorithms and representations
Neuromorphic Diffusion Language Models: Addressing Compute and Memory Bottlenecks via Sparsity and Block DenoisingDiffusion models → Discrete and text diffusion
MedJudgeRAG: Option-Wise Evidence Judgment with Dynamic Knowledge Graphs for Medical MCQARetrieval-augmented generation → Medical and biomedical RAG
Building AI That Works: ESnet's Pragmatic Approach to AI-Driven Operational ExcellenceLLM agents and tool use → General LLM agent frameworks and benchmarks
Let AI Agents Translate Networks, Not Reason About ThemLLM reasoning → Reasoning for agents, planning and code +1
SAGE: Safety-First Defense-in-Depth Guardrails for Verified Lifecycle Control of High-Impact Generative AILLM safety and security → Safety alignment, refusal and guardrails
Not All LLM Reasoning is Visible in the Chain-of-ThoughtLLM reasoning → Test-time compute and efficient thinking
Agent Team Work Zone: An Automated, Persistent Workspace for Long-Lived Claude Code Agent TeamsLLM agents and tool use → Software engineering coding agents
What Can Be Enforced? A Theory of Certified Runtime Safety for Tool-Using AgentsLLM safety and security → Prompt injection and agent security
Spatial-IQ: Deconstructing Spatial Intelligence via Hierarchical Capability TestsLLM reasoning → Spatial and embodied reasoning +1
Language-Routed RAG and Direct Option Scoring for Multilingual Financial QA: DS@GT at FinMMEvalRetrieval-augmented generation → Legal and financial RAG
Disentangling Multi-View Scanning in Mamba for Network Traffic Anomaly DetectionState-space sequence models (S4, Mamba) → Time series forecasting with SSMs
Agentic Autoresearch for CT ReconstructionLLM agents and tool use → Medical and scientific domain agents
The Best Programming Language for Tokenmaxxing: An Investigation of Coding Agent Behavior Across Programming LanguagesLLM agents and tool use → Software engineering coding agents
Generative Video Compression with Adaptive Score DistillationDiffusion models → Diffusion for compression and wireless
Beyond Shapley: An Influence-Based Data Auditing Pipeline for LLM Alignment and EvaluationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Robot-Factored World Models via Robot RenderingWorld models → Embodied and robot world models
Twins: Learn to Predict Unified Representations with Focal LossDiffusion models → Image generation, text-to-image and editing
ViTacWorld: Scaling Visuo-Tactile World Models for Contact-Rich Robot ManipulationWorld models → Embodied and robot world models
Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving SkillsSynthetic data and self-improvement for LLMs → Self-evolving LLM agents
Explainable Reinforcement Learning for assisting Air Traffic ControllersDeep reinforcement learning methods → Interpretable and logic-guided RL
The Regression Tax: Decomposing Why Skills Help and Hurt LLM AgentsLLM agents and tool use → Agent skill learning and evolution
CausalSmith: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal InferenceLLM agents and tool use → Medical and scientific domain agents +1
TRACE-ROUTER: Task-Consistent and Adaptive Online Routing for Agentic AILLM agents and tool use → General LLM agent frameworks and benchmarks
Where Facts Go Missing: A Layerwise Taxonomy and Per-Layer Attribution of Information Omission in Air-Gapped LLMAgent PipelinesLLM agents and tool use → Web, search, and research agents
Dynamic Capability Scoping for Enterprise AI Agents: A Synthetic Dataset and Three-Source Permission ArchitectureLLM agents and tool use → Agent threats, guardrails and governance
On the Identifiability of Controlled World ModelsWorld models → Latent predictive world models
Unboxing Diffusion Models for the Arts: Interactive Model Bending and Practice-Based ExplainabilityDiffusion models → Image generation, text-to-image and editing
Vibe Coding: An Experiment with Test-Driven DevelopmentLLM agents and tool use → Software engineering coding agents
A Self-Calibrating Agentic AI Framework for Autonomous Edge Resource AllocationLLM agents and tool use → General LLM agent frameworks and benchmarks
SceneActBench: Can Agents Act on the 3D Scenes They See?Multimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
HiKV: Hierarchical Importance-Aware KV Cache with Hardware Acceleration for LLM DecodingEfficient LLM inference and serving → KV cache optimization
Correlation-Aware and Gaussianity-Preserving Robust Latent Angular Watermarking for Diffusion ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
Agentic Root Cause Analysis through Evidence-Grounded ReasoningLLM reasoning → Reasoning for agents, planning and code +1
A Factorial Study of Synthetic Data Generation for Low-Resource Machine Translation using Grammar BooksSynthetic data and self-improvement for LLMs → LLM synthetic data generation
IDEAgent: Agentic Quality-Diversity Search for Research Idea GenerationLLM agents and tool use → Web, search, and research agents
Active few-shot segmentation by reinforcing data selectionDeep reinforcement learning methods → RL for clinical treatment and surgery
Do Agent Benchmarks Measure Capability? Protocol Validity in the Age of Agentic AILLM agents and tool use → Software engineering coding agents
TRACTA: Benchmarking Temporal Reasoning over Semantic TrajectoriesLLM reasoning → Knowledge- and tool-augmented reasoning
Integrated Order Dispatching and Routing for Last-Mile Pickup via Deep Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
Cross-Tokenizer On-Policy Distillation via Byte-Prefix MarginalizationKnowledge distillation → On-policy and self-distillation for LLMs
Towards Trustworthy and Cost-Efficient Data Integration: From Na"ive RAG to Agentic RAGLLM agents and tool use → Web, search, and research agents
Evolution-Aware MSA Reasoning for Subsampling via Factor GraphsLLM reasoning → Clinical and biomedical applications
A Roadmap to Impactful Pluralistic Alignment ResearchLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
RadSight: Towards Perceptually Reliable Multimodal Radiology Image UnderstandingMultimodal LLMs / vision-language models → Medical vision-language models
AgentHOI: Multi-Agent Reasoning for Human-Object-Interaction Video Generation via Implicit Representation AlignmentDiffusion models → Video generation
TRaM-VSR: Importance-Aware Token Routing and Merging for One-Step Diffusion Video Super-ResolutionDiffusion models → Inverse problems and image restoration
Filling Before Advancing: Capability-Gap-Driven Post-Training for Scenario-Specialized Remote Sensing MLLMsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models
LayoutLite: Token-Level Implicit Layout Analysis for Efficient Document OCRMultimodal LLMs / vision-language models → Document understanding and multimodal RAG +1
From Score Approximation to Distribution Approximation in Score-Based Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Draining the Energy Commons: Self-Defeating Over-Appropriation as a Coordination Failure in Agentic LLM CollectivesLLM agents and tool use → Social simulation and multi-agent debate
Deconstructing Off-Policy Ratios: Entropy-Normalized Trust Regions for Asynchronous Reinforcement LearningDeep reinforcement learning methods → RL theory and sample complexity
DBA-Bench: A Production-Fidelity Benchmark for LLM-Based Database Operations AgentsLLM agents and tool use → Web, search, and research agents
InnoText: A Unified Model for Visual Text Generation and EditingDiffusion models → Image generation, text-to-image and editing
Reasoning Denoiser: Denoising Reasoning Traces for Hallucination Detection in Large Reasoning ModelsLLM reasoning → Hallucination detection and mitigation
Spectral Prior for Reducing Exposure Bias in Diffusion ModelsDiffusion models → Diffusion foundations and sampling
Nanbeige4.2-3B: Unlocking Agentic Capabilities in a Compact ModelLLM agents and tool use → General LLM agent frameworks and benchmarks
When Language Models Meet NeuroGraphs: Exploring Enhanced Agentic LLM Framework Towards Brain Network AnalysisLLM agents and tool use → Medical and scientific domain agents
Multimodal Language Models Benchmarked Against the NRC Reactor Operator Licensing Examination: Fine-Tuning and Retrieval StrategiesMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
Scaling Native Multimodal Pre-Training From ScratchLLM reasoning → Multimodal reasoning
LAMAR: An Open Language-Aware Multilingual Alignment RerankerRetrieval-augmented generation → Generation and context integration
On the Runtime Analysis of Reinforcement Learning Hyper-HeuristicsDeep reinforcement learning methods → Deep RL algorithms and representations
Small Vision-Language Models Know When They Are Wrong But Cannot Say So: A Two-Model Study of Stated versus Internal Confidence Under Realistic Image DegradationMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Agent Security Needs Redefinition through a Holistic FrameworkLLM safety and security → Prompt injection and agent security
Zero-Shot Mission-Level Evaluation for Aerial MLLM AgentsLLM reasoning → Spatial and embodied reasoning +2
Visual Saliency Steering Distillation for Multimodal Chain-of-Thought ReasoningLLM reasoning → Multimodal reasoning
Learning as Reasoning Unfolds: Progressive Rollout Allocation for Efficient Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning
Music-JEPA: Learning a World Model of Sound from ActionWorld models → Latent predictive world models
"Go Home Copilot, You're Drunk": Understanding Developer Responses to Agent-Generated Code Review CommentsLLM agents and tool use → Software engineering coding agents
Unified Static-Dynamic Pruning for Efficient LLM InferenceEfficient LLM inference and serving → Serving systems and request scheduling
J-CoT: Chain-of-Thought in J-SpaceLLM reasoning → Prompting and chain-of-thought reasoning
MoE2-LoRA: When MoE Models Meet MoE-style Low-Rank AdaptationMixture-of-experts → Mixture of LoRA experts
Teaching LLMs to Self-Evolve: Cultivating Core Meta-Skills with Reinforcement LearningLLM reasoning → Reinforcement learning for reasoning
TextSLIP: Text Self-Supervised CLIP for Medical Report GenerationMultimodal LLMs / vision-language models → Medical vision-language models
Ground Truth First: A Longitudinal Evaluation Instrument for Agent Memory, and the Tenure Crossover in Memory-Architecture RankingsLLM agents and tool use → Agent memory systems
Adaptive Undulatory Locomotion of Snake-like Robots in Dynamic Viscous Environments via Deep Reinforcement LearningDeep reinforcement learning methods → RL for robot navigation and UAV control
Efficient Online LLM Watermark Detection via Rao-Blackwellized E-ProcessesLLM safety and security → Watermarking generated content and models
Multi-Agent Debate and Visual Information Extraction for SeePhys Pro: A 1st-Place Technical Report from ICML 2026 AI4Math Track 3 ChallengeLLM agents and tool use → Multimodal and embodied agents
Leveraging External Knowledge for Historical Document Restoration via Retrieval-Augmented Large Language ModelsRetrieval-augmented generation → RAG pipeline design and evaluation
Semiotic logical hexagon theory for LLM logical reasoningLLM reasoning → Reasoning benchmarks and evaluation
Action-Conditioned World Model for Goal Plane Probe Guidance in Robotic UltrasoundWorld models → Medical and surgical world models
Reliability-Contagion Feasibility in LLM Multi-Agent NetworksLLM agents and tool use → General LLM agent frameworks and benchmarks
TRW: TRACE-RealWorld---An Auditable Consistency Contract for World Models as Materialized ViewsWorld models → World models for language agents
Claim Plane: Enforceable Change Intents and Dynamic Scope for Parallel Coding AgentsLLM agents and tool use → Software engineering coding agents
Diffusion Models in Medical Image Inpainting: Challenges, Solution Taxonomy, and Future DirectionsDiffusion models → Medical imaging and segmentation
Towards Reducing Foreign Language Anxiety Using Level-Appropriate Embodied Conversational AgentsLLM agents and tool use → General LLM agent frameworks and benchmarks
Variance-Reduced Q-Learning over Static and Time-Varying NetworksDeep reinforcement learning methods → RL theory and sample complexity
Memory That Looks Forward: A Zero-Inference Prospective Term for Personal Memory Retrieval
Transsion's Speaker-Attributed Multilingual ASR System for the MLC-SLM 2026 Challenge
Music Hallucination in Audio-Language Models: A Hierarchical Formulation and Empirical Study
MoSSGate: Memory-Modulated State-Space Gating for Skin Lesion Segmentation
Sequential Contextual Fit Predicts Human Behavioural and Neural Dynamics Across Domains
MTF-Net: Multi-Modal Temporal Feature Fusion Network for Pedestrian Intention Prediction
PaGNet: A Panel-Aware GBDT--Neural Network for Multi-Target Corporate Tax Avoidance Proxy Forecasting

All 307