Papers

New on arXiv

Fri 13 Mar 2026

401 papers

Separating Intelligence from Execution: A Workflow Engine for the Model Context ProtocolLLM agents and tool use → Tool learning and function calling
Compile to Compress: Boosting Formal Theorem Provers by Compiler OutputsLLM reasoning → Mathematical reasoning
Semantic Consensus: Process-Aware Conflict Detection and Resolution for Enterprise Multi-Agent LLM SystemsLLM agents and tool use → General LLM agent frameworks and benchmarks
Governing the Agentic Enterprise: A Governance Maturity Model for Managing AI Agent Sprawl in Business OperationsLLM agents and tool use → Agent threats, guardrails and governance
HR-Agents: Using Multiple LLM-based Agents to Improve Q&A about Brazilian Labor LegislationRetrieval-augmented generation → Legal and financial RAG +1
Beyond Verifiable Rewards: Rubric-Based GRM for Reinforced Fine-Tuning SWE AgentsLLM agents and tool use → Software engineering coding agents +1
Token-Budget-Aware Pool Routing for Cost-Efficient LLM InferenceEfficient LLM inference and serving → Serving systems and request scheduling
TDA-RC: Task-Driven Alignment for Knowledge-Based Reasoning Chains in Large Language ModelsLLM reasoning → Prompting and chain-of-thought reasoning
Deciphering Scientific Reasoning Steps from Outcome Data for Molecule OptimizationLLM reasoning → Clinical and biomedical applications
LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from PixelsWorld models → Latent predictive world models
PrefPO: Pairwise Preference Prompt OptimizationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
MemReward: Graph-Based Experience Memory for LLM Reward Prediction with Limited LabelsLLM reasoning → Reinforcement learning for reasoning
DRCY: Agentic Hardware Design ReviewsLLM agents and tool use → Software engineering coding agents
I Know What I Don't Know: Latent Posterior Factor Models for Multi-Evidence Probabilistic ReasoningLLM reasoning → Test-time compute and efficient thinking
Locatability-Guided Adaptive Reasoning for Image Geo-Localization with Vision-Language ModelsLLM reasoning → Spatial and embodied reasoning +2
Design and evaluation of an agentic workflow for crisis-related synthetic tweet datasetsLLM agents and tool use → Social simulation and multi-agent debate
Egocentric World Model for Photorealistic Hand-Object Interaction SynthesisWorld models → Embodied and robot world models
LLM Routing as Reasoning: A MaxSAT ViewLLM reasoning → Test-time compute and efficient thinking
NCCL EP: Towards a Unified Expert Parallel Communication API for NCCLMixture-of-experts → Efficient MoE training and inference systems
Orla: A Library for Serving LLM-Based Multi-Agent SystemsLLM agents and tool use → Software engineering coding agents
EnterpriseOps-Gym: Environments and Evaluations for Stateful Agentic Planning and Tool Use in Enterprise SettingsLLM agents and tool use → General LLM agent frameworks and benchmarks
EmDT: Embedding Diffusion Transformer for Tabular Data Generation in Fraud DetectionDiffusion models → Flow matching, time series and science
NumColor: Precise Numeric Color Control in Text-to-Image GenerationDiffusion models → Image generation, text-to-image and editing
CTI-REALM: Benchmark to Evaluate Agent Performance on Security Detection Rule Generation CapabilitiesLLM agents and tool use → Agent threats, guardrails and governance
LibraGen: Playing a Balance Game in Subject-Driven Video GenerationLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Purifying Generative LLMs from Backdoors without Prior Knowledge or Clean ReferenceLLM safety and security → Backdoor and data poisoning attacks
LADR: Locality-Aware Dynamic Rescue for Efficient Text-to-Image Generation with Diffusion Large Language ModelsDiffusion models → Image generation, text-to-image and editing
NormCode Canvas: Making LLM Agentic Workflows Development Sustainable via Case-Based ReasoningLLM reasoning → Reasoning for agents, planning and code +1
Improving Channel Estimation via Multimodal Diffusion Models with Flow MatchingDiffusion models → Diffusion for compression and wireless
CtrlAttack: A Unified Attack on World-Model Control in Diffusion ModelsWorld models → Video-generation world models and simulators +1
Spatially Grounded Long-Horizon Task Planning in the WildLLM reasoning → Spatial and embodied reasoning +1
Self-Flow-Matching assisted Full Waveform InversionDiffusion models → Flow matching, time series and science
Agent Privilege Separation in OpenClaw: A Structural Defense Against Prompt InjectionLLM agents and tool use → Agent threats, guardrails and governance +1
From Gradients to Riccati Geometry: Kalman World Models for Single-Pass LearningWorld models → Latent dynamics models and model-based RL
PhysMoDPO: Physically-Plausible Humanoid Motion with Preference OptimizationDiffusion models → Diffusion and flow policies for robotics +1
Visual-ERM: Reward Modeling for Visual EquivalenceLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
Out of Sight, Out of Mind? Evaluating State Evolution in Video World ModelsWorld models → Video-generation world models and simulators
LLM Constitutional Multi-Agent GovernanceLLM agents and tool use → Social simulation and multi-agent debate
Semantic Invariance in Agentic AILLM reasoning → Reasoning for agents, planning and code +1
DiT-IC: Aligned Diffusion Transformer for Efficient Image CompressionDiffusion models → Diffusion foundations and sampling
When Right Meets Wrong: Bilateral Context Conditioning with Reward-Confidence Correction for GRPOLLM reasoning → Reinforcement learning for reasoning
Geometry-Guided Camera Motion Understanding in VideoLLMsMultimodal LLMs / vision-language models → Video understanding with MLLMs
Evaluating VLMs' Spatial Reasoning Over Robot Motion: A Step Towards Robot Planning with Motion PreferencesLLM reasoning → Spatial and embodied reasoning
Beyond Final Answers: CRYSTAL Benchmark for Transparent Multimodal Reasoning EvaluationLLM reasoning → Multimodal reasoning
Reasoning over Video: Evaluating How MLLMs Extract, Integrate, and Reconstruct Spatiotemporal EvidenceLLM reasoning → Spatial and embodied reasoning
Mitigating Memorization in Text-to-Image Diffusion via Region-Aware Prompt Augmentation and Multimodal Copy DetectionDiffusion models → Diffusion foundations and sampling
Fractals made Practical: Denoising Diffusion as Partitioned Iterated Function SystemsDiffusion models → Diffusion foundations and sampling
Topo-R1: Detecting Topological Anomalies via Vision-Language ModelsMultimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
Mending the Holes: Mitigating Reward Hacking in Reinforcement Learning for Multilingual TranslationLLM alignment / RLHF / preference optimisation → LLM preference optimization methods
ESPIRE: A Diagnostic Benchmark for Embodied Spatial Reasoning of Vision-Language ModelsLLM reasoning → Spatial and embodied reasoning +1
Purify Once, Edit Freely: Breaking Image Protections under Model MismatchDiffusion models → Robustness, privacy and adversarial aspects +1
PISmith: Reinforcement Learning-based Red Teaming for Prompt Injection DefensesLLM safety and security → Prompt injection and agent security
SAW: Toward a Surgical Action World Model via Controllable and Scalable Video GenerationWorld models → Medical and surgical world models +1
ARL-Tangram: Unleash the Resource Efficiency in Agentic Reinforcement LearningLLM agents and tool use → Reinforcement learning for LLM agents
Structured Distillation for Personalized Agent Memory: 11x Token Reduction with Retrieval PreservationLLM agents and tool use → Agent memory systems +1
Accelerating Stroke MRI with Diffusion Probabilistic Models through Large-Scale Pre-training and Target-Specific Fine-TuningDiffusion models → Medical imaging and segmentation
A Closed-Form Solution for Debiasing Vision-Language Models with Utility Guarantees Across Modalities and TasksMultimodal LLMs / vision-language models → Social bias and fairness in VLMs
DAPD: Dependency-Aware Parallel Decoding via Attention for Diffusion LLMsEfficient LLM inference and serving → Diffusion LLM and MoE inference
Test-Time Attention Purification for Backdoored Large Vision Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness +1
Exact Federated Continual Unlearning for Ridge Heads on Frozen Foundation ModelsLLM safety and security → Federated unlearning and poisoning
Long-form RewardBench: Evaluating Reward Models for Long-form GenerationLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Delta1 with LLM: symbolic and neural integration for credible and explainable reasoningLLM reasoning → Reasoning benchmarks and evaluation
Editing Away the Evidence: Diffusion-Based Image Manipulation and the Failure Modes of Robust WatermarkingDiffusion models → Robustness, privacy and adversarial aspects +1
ReMem-VLA: Empowering Vision-Language-Action Model with Memory via Dual-Level Recurrent QueriesVision-language-action models → Reasoning and chain-of-thought in VLA
RoboStream: Weaving Spatio-Temporal Reasoning with Memory in Vision-Language Models for RoboticsLLM reasoning → Spatial and embodied reasoning +1
Efficient and Interpretable Multi-Agent LLM Routing via Ant Colony OptimizationLLM agents and tool use → General LLM agent frameworks and benchmarks
DS2-Instruct: Domain-Specific Data Synthesis for Large Language Models Instruction TuningSynthetic data and self-improvement for LLMs → Code and instruction data synthesis
Rethinking VLMs for Image Forgery Detection and LocalizationLLM reasoning → Multimodal reasoning
Stake the Points: Structure-Faithful Instance UnlearningLLM safety and security → Machine unlearning
A theory of learning data statistics in diffusion models, from easy to hardDiffusion models → Diffusion foundations and sampling
Finite Difference Flow Optimization for RL Post-Training of Text-to-Image ModelsDiffusion models → Diffusion and flow policies for robotics
TRACE: Structure-Aware Character Encoding for Robust and Generalizable Document WatermarkingDiffusion models → Robustness, privacy and adversarial aspects
Beyond Imitation: Reinforcement Learning Fine-Tuning for Adaptive Diffusion Navigation PoliciesDiffusion models → Diffusion and flow policies for robotics +1
Composing Driving Worlds through Disentangled Control for Adversarial Scenario GenerationEnd-to-end autonomous driving → Generative simulation for driving
Mask2Flow-TSE: Two-Stage Target Speaker Extraction with Masking and Flow MatchingDiffusion models → Audio, speech and music
Rethinking Multiple-Choice Questions for RLVR: Unlocking Potential via Distractor DesignLLM reasoning → Reinforcement learning for reasoning
NanoVDR: Distilling a 2B Vision-Language Retriever into a 70M Text-Only Encoder for Visual Document RetrievalMultimodal LLMs / vision-language models → Document understanding and multimodal RAG
Adaptive Vision-Language Model Routing for Computer Use AgentsMultimodal LLMs / vision-language models → Embodied and 3D vision-language models +1
Context is all you need: Towards autonomous model-based process design using agentic AI in flowsheet simulationsLLM agents and tool use → Medical and scientific domain agents
OARS: Process-Aware Online Alignment for Generative Real-World Image Super-ResolutionLLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment
A Multitask Large Reasoning Model for Molecular ScienceLLM reasoning → Clinical and biomedical applications
FLUX: Accelerating Cross-Embodiment Generative Navigation Policies via Rectified Flow and Static-to-Dynamic LearningDiffusion models → Diffusion and flow policies for robotics
What Makes VLMs Robust? Towards Reconciling Robustness and Accuracy in Vision-Language ModelsMultimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness
SteerRM: Debiasing Reward Models via Sparse AutoencodersLLM alignment / RLHF / preference optimisation → Reward modeling and RLHF
Cheers: Decoupling Patch Details from Semantic Representations Enables Unified Multimodal Comprehension and GenerationEfficient LLM inference and serving → Visual token pruning in multimodal models
Think and Answer ME: Benchmarking and Exploring Multi-Entity Reasoning Grounding in Remote SensingLLM reasoning → Spatial and embodied reasoning
PVI: Plug-in Visual Injection for Vision-Language-Action ModelsVision-language-action models → Efficient VLA inference and adaptation +1
Easy-IIL: Reducing Human Operational Burden in Interactive Imitation Learning via Assistant ExpertsOffline RL and imitation learning from data → Imitation learning methods
HIFICL: High-Fidelity In-Context Learning for Multimodal TasksMultimodal LLMs / vision-language models → MLLM training and visual instruction tuning
SLICE: Semantic Latent Injection via Compartmentalized Embedding for Image WatermarkingDiffusion models → Robustness, privacy and adversarial aspects
Thinking in Dynamics: How Multimodal Large Language Models Perceive, Track, and Reason Dynamics in Physical 4D WorldLLM reasoning → Spatial and embodied reasoning +1
ToolTree: Efficient LLM Agent Tool Planning via Dual-Feedback Monte Carlo Tree Search and Bidirectional PruningLLM agents and tool use → Tool learning and function calling
AnchorVLA4D: an Anchor-Based Spatial-Temporal Vision-Language-Action Model for Robotic ManipulationLLM reasoning → Spatial and embodied reasoning +1
Altered Thoughts, Altered Actions: Probing Chain-of-Thought Vulnerabilities in VLA Robotic ManipulationLLM reasoning → Spatial and embodied reasoning +1
AI Planning Framework for LLM-Based Web AgentsLLM agents and tool use → Web, search, and research agents
Cost-Efficient Multimodal LLM Inference via Cross-Tier GPU HeterogeneityMultimodal LLMs / vision-language models → Visual token pruning and efficiency
FGTR: Fine-Grained Multi-Table Retrieval via Hierarchical LLM ReasoningLLM reasoning → Knowledge- and tool-augmented reasoning
Experimental evidence of progressive ChatGPT models self-convergenceSynthetic data and self-improvement for LLMs → LLM synthetic data generation
Colluding LoRA: A Compositional Vulnerability in LLM Safety AlignmentLLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks +1
Vision Verification Enhanced Fusion of VLMs for Efficient Visual ReasoningLLM reasoning → Multimodal reasoning +1

All 401