Papers

New on arXiv

Wed 18 Mar 2026

446 papers

CentaurTA Studio: A Self-Improving Human-Agent Collaboration System for Thematic AnalysisLLM agents and tool use → Software engineering coding agents +1
SaFeR-Steer: Evolving Multi-Turn MLLMs via Synthetic Bootstrapping and Feedback DynamicsLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Dental-TriageBench: Benchmarking Multimodal Reasoning for Hierarchical Dental TriageLLM reasoning → Clinical and biomedical applications
Attention-Based Sampler for Diffusion Language ModelsDiffusion models → Discrete and text diffusion
Temperature-Dependent Performance of Prompting Strategies in Extended Reasoning Large Language ModelsLLM reasoning → Prompting and chain-of-thought reasoning
Toward Evaluation Frameworks for Multi-Agent Scientific AI SystemsLLM agents and tool use → Web, search, and research agents
FEMBA on the Edge: Physiologically-Aware Pre-Training, Quantization, and Deployment of a Bidirectional Mamba EEG Foundation Model on an Ultra-low Power MicrocontrollerState-space sequence models (S4, Mamba) → Physiological signal modeling
Full waveform inversion method based on diffusion modelDiffusion models → Flow matching, time series and science
Rethinking Retrieval-Augmentation as Synthesis: A Query-Aware Context Merging ApproachLLM reasoning → Knowledge- and tool-augmented reasoning +1
On the Fragility of AI Agent CollusionLLM agents and tool use → Financial trading and forecasting agents
FalconBC: Flow matching for Amortized inference of Latent-CONditioned physiologic Boundary ConditionsDiffusion models → Medical imaging and segmentation
The Verifier Tax: Horizon Dependent Safety Success Tradeoffs in Tool Using LLM AgentsLLM agents and tool use → Agent threats, guardrails and governance
Synthetic Data Generation for Training Diversified Commonsense Reasoning ModelsLLM reasoning → Reasoning benchmarks and evaluation +1
LGESynthNet: Controlled Scar Synthesis for Improved Scar Segmentation in Cardiac LGE-MRI ImagingDiffusion models → Medical imaging and segmentation
Interpretability without actionability: mechanistic methods cannot correct language model errors despite near-perfect internal representationsMechanistic interpretability → Interpreting biomedical and science models
Shifting Uncertainty to Critical Moments: Towards Reliable Uncertainty Quantification for VLA ModelVision-language-action models → Failure recovery and robustness
ManiDreams: An Open-Source Library for Robust Object Manipulation via Uncertainty-aware Task-specific Intuitive PhysicsWorld models → Embodied and robot world models
Can LLMs Reason Like Automated Theorem Provers for Rust Verification? VCoT-Bench: Evaluating via Verification Chain of ThoughtLLM reasoning → Reasoning for agents, planning and code
Escaping Offline Pessimism: Vector-Field Reward Shaping for Safe Frontier ExplorationOffline RL and imitation learning from data → Offline RL algorithms +1
Learning to Reason with Curriculum I: Provable Benefits of AutocurriculumLLM reasoning → Reinforcement learning for reasoning +1
DriveVLM-RL: Neuroscience-Inspired Reinforcement Learning with Vision-Language Models for Safe and Deployable Autonomous DrivingMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection +1
Approximate Subgraph Matching with Neural Graph Representations and Reinforcement LearningDeep reinforcement learning methods → RL for games and combinatorial optimization
Path-Constrained Mixture-of-ExpertsMixture-of-experts → Expert routing strategies
CycleCap: Improving VLMs Captioning Performance via Self-Supervised Cycle Consistency Fine-TuningLLM reasoning → Multimodal reasoning +1
Detection Is Cheap, Routing Is Learned: Why Refusal-Based Alignment Evaluation FailsLLM safety and security → Safety alignment, refusal and guardrails
EDM-ARS: A Domain-Specific Multi-Agent System for Automated Educational Data Mining ResearchLLM agents and tool use → Educational agents and math problem solving
Retrieval-Augmented LLM Agents: Learning to Learn from ExperienceRetrieval-augmented generation → Agentic RAG, agents and memory +1
Enactor: From Traffic Simulators to Surrogate World ModelsWorld models → World models for autonomous driving
Sharpness-Aware Minimization in Logit Space Efficiently Enhances Direct Preference OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
Discovering What You Can Control: Interventional Boundary Discovery for Reinforcement LearningDeep reinforcement learning methods → Deep RL algorithms and representations
MolRGen: A Training and Evaluation Setting for De Novo Molecular Generation with Reasonning ModelsLLM reasoning → Clinical and biomedical applications
Who Tests the Testers? Systematic Enumeration and Coverage Audit of LLM Agent Tool Call SafetyLLM agents and tool use → Agent threats, guardrails and governance
Toward Reliable, Safe, and Secure LLMs for Scientific ApplicationsLLM agents and tool use → Agent threats, guardrails and governance
A Hybrid Conditional Diffusion-DeepONet Framework for High-Fidelity Stress Prediction in Hyperelastic MaterialsDiffusion models → Flow matching, time series and science
GoalVLM: VLM-driven Object Goal Navigation for Multi-Agent SystemLLM agents and tool use → Multimodal and embodied agents
R2-Dreamer: Redundancy-Reduced World Models without Decoders or AugmentationWorld models → Latent dynamics models and model-based RL +1
Access Controlled Website Interaction for Agentic AI with Delegated Critical TasksLLM agents and tool use → Web, search, and research agents
Retrieval-Augmented LLMs for Security Incident AnalysisRetrieval-augmented generation → Privacy and security of RAG
VLM-AutoDrive: Post-Training Vision-Language Models for Safety-Critical Autonomous Driving EventsMultimodal LLMs / vision-language models → Driving VLMs and open-vocabulary detection +1
Insight-V++: Towards Advanced Long-Chain Visual Reasoning with Multimodal Large Language ModelsLLM reasoning → Multimodal reasoning +1
VC-Soup: Value-Consistency Guided Multi-Value Alignment for Large Language ModelsLLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment
A Trace-Based Assurance Framework for Agentic AI Orchestration: Contracts, Testing, and GovernanceLLM agents and tool use → Agent threats, guardrails and governance
Q-Drift: Quantization-Aware Drift Correction for Diffusion Model SamplingDiffusion models → Diffusion foundations and sampling
Action Draft and Verify: A Self-Verifying Framework for Vision-Language-Action ModelVision-language-action models → Efficient VLA inference and adaptation
CytoSyn: a Foundation Diffusion Model for Histopathology -- Tech ReportDiffusion models → Medical imaging and segmentation
Enhancing Reinforcement Learning Fine-Tuning with an Online RefinerDeep reinforcement learning methods → RL training of language and agent models
Uncovering Latent Phase Structures and Branching Logic in Locomotion Policies: A Case Study on HalfCheetahDeep reinforcement learning methods → RL for robot locomotion and manipulation
SLEA-RL: Step-Level Experience Augmented Reinforcement Learning for Multi-Turn Agentic TrainingLLM agents and tool use → Reinforcement learning for LLM agents
Adapting Technical-Service LLM Agents with Latent Logic Augmentation, Robust Noise Reduction, and Hybrid Reward ModelingLLM agents and tool use → Reinforcement learning for LLM agents +1
Continually self-improving AISynthetic data and self-improvement for LLMs → Self-rewarding alignment and self-training
Guardrails as Infrastructure: Policy-First Control for Tool-Orchestrated WorkflowsLLM agents and tool use → Agent threats, guardrails and governance +1
Fundamental Limits of Neural Network Sparsification: Evidence from Catastrophic Interpretability CollapseMechanistic interpretability → Sparse autoencoder training and evaluation
Unified Spatio-Temporal Token Scoring for Efficient Video VLMsMultimodal LLMs / vision-language models → Visual token pruning and efficiency +1
Universal Skeleton Understanding via Differentiable Rendering and MLLMsLLM reasoning → Spatial and embodied reasoning +1
Loc3R-VLM: Language-based Localization and 3D Reasoning with Vision-Language ModelsLLM reasoning → Spatial and embodied reasoning +1
AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and ReuseLLM agents and tool use → General LLM agent frameworks and benchmarks +1
The Unreasonable Effectiveness of Text Embedding Interpolation for Continuous Image SteeringLLM reasoning → Multimodal reasoning
GMT: Goal-Conditioned Multimodal Transformer for 6-DOF Object Trajectory Synthesis in 3D ScenesLLM reasoning → Spatial and embodied reasoning
Feeling the Space: Egomotion-Aware Video Representation for Efficient and Accurate 3D Scene UnderstandingLLM reasoning → Spatial and embodied reasoning +1
AHOY! Animatable Humans under Occlusion from YouTube Videos with Gaussian Splatting and Video Diffusion PriorsDiffusion models → 3D, point cloud and scene generation
TDAD: Test-Driven Agentic Development - Reducing Code Regressions in AI Coding Agents via Graph-Based Impact AnalysisLLM agents and tool use → Software engineering coding agents
LaDe: Unified Multi-Layered Graphic Media Generation and DecompositionDiffusion models → Image generation, text-to-image and editing
Unified Policy Value Decomposition for Rapid AdaptationDeep reinforcement learning methods → Deep RL algorithms and representations
CARE: Covariance-Aware and Rank-Enhanced Decomposition for Enabling Multi-Head Latent AttentionEfficient LLM inference and serving → KV cache optimization
IndicSafe: A Benchmark for Evaluating Multilingual LLM Safety in South AsiaLLM safety and security → Safety alignment, refusal and guardrails
Differential Privacy in Generative AI Agents: Analysis and Optimal TradeoffsLLM agents and tool use → Agent threats, guardrails and governance
Operator-Theoretic Foundations and Policy Gradient Methods for General MDPs with Unbounded CostsDeep reinforcement learning methods → RL theory and sample complexity
DiscoGen: Procedural Generation of Algorithm Discovery Tasks in Machine LearningDeep reinforcement learning methods → RL training of language and agent models
ProbeFlow: Training-Free Adaptive Flow Matching for Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation +1
Generative Control as Optimization: Time Unconditional Flow Matching for Adaptive and Robust Robotic ControlDiffusion models → Diffusion and flow policies for robotics +1
Text-to-Stage: Spatial Layouts from Long-form NarrativesLLM reasoning → Spatial and embodied reasoning
RPMS: Enhancing LLM-Based Embodied Planning through Rule-Augmented Memory SynergyLLM agents and tool use → Multimodal and embodied agents
CodeScout: An Effective Recipe for Reinforcement Learning of Code Search AgentsLLM agents and tool use → Software engineering coding agents +1
TINA: Text-Free Inversion Attack for Unlearned Text-to-Image Diffusion ModelsDiffusion models → Robustness, privacy and adversarial aspects +1
Steering Video Diffusion Transformers with Massive ActivationsDiffusion models → Video generation
Multi-Source Evidence Fusion for Audio Question AnsweringLLM reasoning → Audio-language model reasoning
Federated Distributional Reinforcement Learning with Distributional Critic RegularizationDeep reinforcement learning methods → RL for energy, networks and operations
Process Supervision for Chain-of-Thought Reasoning via Monte Carlo Net Information GainLLM reasoning → Reinforcement learning for reasoning
ChopGrad: Pixel-Wise Losses for Latent Video Diffusion via Truncated BackpropagationDiffusion models → Video generation
Dropout Robustness and Cognitive Profiling of Transformer Models via Stochastic InferenceLLM reasoning → Test-time compute and efficient thinking
Fine-Grained Post-Training Quantization for Large Vision Language Models with Quantization-Aware Integrated GradientsMultimodal LLMs / vision-language models → Visual token pruning and efficiency
EVA: Aligning Video World Models with Executable Robot Actions via Inverse Dynamics RewardsWorld models → Embodied and robot world models +1
CoVerRL: Breaking the Consensus Trap in Label-Free Reasoning via Generator-Verifier Co-EvolutionLLM reasoning → Reinforcement learning for reasoning
Grounded Multimodal Retrieval-Augmented Drafting of Radiology Impressions Using Case-Based Similarity SearchRetrieval-augmented generation → Medical and biomedical RAG
TAPESTRY: From Geometry to Appearance via Consistent Turntable VideosDiffusion models → 3D, point cloud and scene generation
SARE: Sample-wise Adaptive Reasoning for Training-free Fine-grained Visual RecognitionLLM reasoning → Multimodal reasoning +1
VolumeDP: Modeling Volumetric Representation for Manipulation Policy LearningOffline RL and imitation learning from data → Robot manipulation imitation learning
From Virtual Environments to Real-World Trials: Emerging Trends in Autonomous DrivingEnd-to-end autonomous driving → Generative simulation for driving
MALLES: A Multi-agent LLMs-based Economic Sandbox with Consumer Preference AlignmentLLM agents and tool use → Social simulation and multi-agent debate +1
Learning Transferable Temporal Primitives for Video Reasoning via Synthetic VideosLLM reasoning → Multimodal reasoning +1
Flow Matching Policy Optimization with Mirror Descent and Entropy ConstraintsDiffusion models → Diffusion and flow policies for robotics +1
Sensi: Learn One Thing at a Time -- Curriculum-Based Test-Time Learning for LLM Game AgentsLLM agents and tool use → Reinforcement learning for LLM agents
WeatherReasonSeg: A Benchmark for Weather-Aware Reasoning Segmentation in Visual Language ModelsLLM reasoning → Multimodal reasoning +1
Adaptive Guidance for Retrieval-Augmented Masked Diffusion ModelsRetrieval-augmented generation → RAG robustness and adaptive retrieval +1
Towards Reliable Local Security Agents: Verifiable Post-Training for Linux Privilege EscalationLLM agents and tool use → Agent threats, guardrails and governance
Few-Step Diffusion Sampling Through Instance-Aware DiscretizationsDiffusion models → Diffusion foundations and sampling
AgentVLN: Towards Agentic Vision-and-Language NavigationLLM agents and tool use → Multimodal and embodied agents
FINER: MLLMs Hallucinate under Fine-grained Negative QueriesMultimodal LLMs / vision-language models → Hallucination mitigation in LVLMs
Interpretable Cross-Domain Few-Shot Learning with Rectified Target-Domain Local AlignmentMultimodal LLMs / vision-language models → Prompt learning and adaptation of CLIP
VectorWorld: Efficient Streaming World Model via Diffusion Flow on Vector GraphsWorld models → World models for autonomous driving +1

All 446