mlsanityData to 7 Oct 2026

UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning

arXiv 2509.02544 ↗First posted 2 Sep 2025cs.AI cs.CL cs.CVcited by 205

On the map: LLM agents and tool use → GUI and mobile agents

Abstract (arXiv)

The development of autonomous agents for graphical user interfaces (GUIs) presents major challenges in artificial intelligence. While recent advances in native agent models have shown promise by unifying perception, reasoning, action, and memory through end-to-end learning, open problems remain in data scalability, multi-turn reinforcement learning (RL), the limitations of GUI-only operation, and environment stability. In this technical report, we present UI-TARS-2, a native GUI-centered agent model that addresses these challenges through a systematic training methodology: a data flywheel for scalable data generation, a stabilized multi-turn RL framework, a hybrid GUI environment that integrates file systems and terminals, and a unified sandbox platform for large-scale rollouts. Empirical evaluation demonstrates that UI-TARS-2 achieves significant improvements over its predecessor UI-TARS-1.5. On GUI benchmarks, it reaches 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld, outperforming strong baselines such as Claude and OpenAI agents. In game environments, it attains a mean normalized score of 59.8 across a 15-game suite-roughly 60% of human-level performance-and remains competitive with frontier proprietary models (e.g., OpenAI o3) on LMGame-Bench. Additionally, the model can generalize to long-horizon information-seeking tasks and software engineering benchmarks, highlighting its robustness across diverse agent tasks. Detailed analyses of training dynamics further provide insights into achieving stability and efficiency in large-scale agent RL. These results underscore UI-TARS-2's potential to advance the state of GUI agents and exhibit strong generalization to real-world interactive scenarios.

Key facts. Cited by 205 papers in the map (data through 7 Oct 2026). Placed in GUI and mobile agents, one of 16 directions in LLM agents and tool use. Builds on 55 of its 72 references that are in the map. First posted on arXiv 2 Sep 2025.

Builds on

55 of 72 references.

Closest ancestors · 13

Same direction, newest first

  1. 2025OS Agents: A Survey on MLLM-based Agents for General Computing Devices Use6 Aug
  2. 2025MobileGUI-RL: Advancing Mobile GUI Agent through Reinforcement Learning in Online Environment8 Jul
  3. 2025ARPO:End-to-End Policy Optimization for GUI Agents with Experience Replay22 May
  4. 2025A Survey on (M)LLM-Based GUI Agents27 Mar
  5. 2025UI-TARS: Pioneering Automated GUI Interaction with Native Agents21 Jan
  6. 2024GUI Agents: A Survey18 Dec
  7. 2024Large Language Model-Brained GUI Agents: A Survey27 Nov
  8. 2024GUI Agents with Foundation Models: A Comprehensive Survey7 Nov
  9. 2024OS-ATLAS: A Foundation Action Model for Generalist GUI Agents30 Oct
  10. 2024Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale12 Sep
  11. 2024OSWorld: Benchmarking Multimodal Agents for Open-Ended Tasks in Real Computer Environments11 Apr
  12. 2024SeeClick: Harnessing GUI Grounding for Advanced Visual GUI Agents17 Jan
1 more
  1. 2023CogAgent: A Visual Language Model for GUI Agents14 Dec

Other references · 42

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  3. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  4. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  5. 2023Toolformer: Language Models Can Teach Themselves to Use Toolscited by 4,072
  6. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  7. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,655
  8. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,193
  9. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
  10. 2023WebArena: A Realistic Web Environment for Building Autonomous Agentscited by 1,908
  11. 2024SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineeringcited by 1,628
  12. 2024A Survey on LLM-as-a-Judgecited by 1,384
30 more
  1. 2023Mind2Web: Towards a Generalist Agent for the Webcited by 1,210
  2. 2022A Generalist Agentcited by 802
  3. 2025BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agentscited by 616
  4. 2022MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledgecited by 469
  5. 2025GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Modelscited by 446
  6. 2025Search-o1: Agentic Search-Enhanced Large Reasoning Modelscited by 423
  7. 2025ReTool: Reinforcement Learning for Strategic Tool Use in LLMscited by 377
  8. 2022Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videoscited by 338
  9. 2024Training Software Engineering Agents and Verifiers with SWE-Gymcited by 300
  10. 2023ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solvingcited by 276
  11. 2024Aguvis: Unified Pure Vision Agents for Autonomous GUI Interactioncited by 271
  12. 2023JARVIS-1: Open-World Multi-task Agents with Memory-Augmented Multimodal Language Modelscited by 170
  13. 2025An Illusion of Progress? Assessing the Current State of Web Agentscited by 165
  14. 2022MRKL Systems: A modular, neuro-symbolic architecture that combines large language models, external knowledge sources and discrete reasoningcited by 153
  15. 2025BrowseComp-ZH: Benchmarking Web Browsing Ability of Large Language Models in Chinesecited by 139
  16. 2025SWE-rebench: An Automated Pipeline for Task Collection and Decontaminated Evaluation of Software Engineering Agentscited by 127
  17. 2025Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systemscited by 125
  18. 2025Alita: Generalist Agent Enabling Scalable Agentic Reasoning with Minimal Predefinition and Maximal Self-Evolutioncited by 114
  19. 2023ProAgent: Building Proactive Cooperative Agents with Large Language Modelscited by 97
  20. 2025What's Behind PPO's Collapse in Long-CoT? Value Optimization Holds the Secretcited by 95
  21. 2025Acting Less is Reasoning More! Teaching Model to Act Efficientlycited by 80
  22. 2023Large Language Models Play StarCraft II: Benchmarks and A Chain of Summarization Approachcited by 79
  23. 2024RepoAgent: An LLM-Powered Open-Source Framework for Repository-level Code Documentation Generationcited by 69
  24. 2024Scaling Instructable Agents Across Many Simulated Worldscited by 67
  25. 2025SimpleDeepSearcher: Deep Information Seeking via Web-Powered Reasoning Trajectory Synthesiscited by 55
  26. 2023OWL: A Large Language Model for IT Operationscited by 43
  27. 2025R1-Searcher++: Incentivizing the Dynamic Knowledge Acquisition of LLMs via Reinforcement Learningcited by 30
  28. 2025JARVIS-VLA: Post-Training Large-Scale Vision Language Models to Play Visual Games with Keyboards and Mousecited by 27
  29. 2022Exploring Mode Connectivity for Pre-trained Language Modelscited by 26
  30. 2025ManuSearch: Democratizing Deep Search in Large Language Models with a Transparent and Open Multi-Agent Frameworkcited by 11

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 52; 2026: 153.

205 papers.

Sort:

By area:

  1. 2025A Survey of Reinforcement Learning for Large Reasoning Modelscited by 175
  2. 2026Agent Skills for Large Language Models: Architecture, Acquisition, Security, and the Path Forwardcited by 105
  3. 2025MAI-UI Technical Report: Real-World Centric Foundation GUI Agentscited by 88
  4. 2026OpenClaw-RL: Train Any Agent Simply by Talkingcited by 83
  5. 2025MobileWorld: Benchmarking Autonomous Mobile Agents in Agent-User Interactive and MCP-Augmented Environmentscited by 69
  6. 2025WebSailor-V2: Bridging the Chasm to Proprietary Agents via Synthetic Data and Scalable Reinforcement Learningcited by 59
  7. 2025Tongyi DeepResearch Technical Reportcited by 53
  8. 2026EvoCUA: Evolving Computer Use Agents via Learning from Scalable Synthetic Experiencecited by 51
  9. 2025AdvancedIF: Rubric-Based Benchmarking and Reinforcement Learning for Advancing LLM Instruction Followingcited by 46
  10. 2026A Survey of Agent Memory in the Second Half: Towards Self-Evolving and Long-Horizon Agentscited by 40
  11. 2026Code as Agent Harnesscited by 40
  12. 2026UI-Venus-1.5 Technical Reportcited by 39
38 more
  1. 2025Step-GUI Technical Reportcited by 38
  2. 2025Fara-7B: An Efficient Agentic Model for Computer Usecited by 34
  3. 2025CostBench: Evaluating Multi-Turn Cost-Optimal Planning and Adaptation in Dynamic Environments for LLM Tool-Use Agentscited by 32
  4. 2026CUA-Skill: Develop Skills for Computer Using Agentcited by 32
  5. 2025Search Self-play: Pushing the Frontier of Agent Capability without Supervisioncited by 31
  6. 2025SR-Scientist: Scientific Equation Discovery With Agentic AIcited by 30
  7. 2025UltraCUA: A Foundation Model for Computer Use Agents with Hybrid Actioncited by 30
  8. 2026OSWorld 2.0: Benchmarking Computer Use Agents on Long-Horizon Real-World Taskscited by 29
  9. 2025Turn-PPO: Turn-Level Advantage Estimation with PPO for Improved Multi-Turn RL in Agentic LLMscited by 28
  10. 2025Efficient Multi-turn RL for GUI Agents via Decoupled Training and Adaptive Data Curationcited by 24
  11. 2025Game-TARS: Pretrained Foundation Models for Scalable Generalist Multimodal Game Agentscited by 23
  12. 2026Beyond Trajectory-Level Attribution: Graph-Based Credit Assignment for Agentic Reinforcement Learningcited by 23
  13. 2025VideoAgentTrek: Computer Use Pretraining from Unlabeled Videoscited by 22
  14. 2025SALT: Step-level Advantage Assignment for Long-horizon Agents via Trajectory Graphcited by 22
  15. 2026CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agentscited by 21
  16. 2025UI-Ins: Enhancing GUI Grounding with Multi-Perspective Instruction-as-Reasoningcited by 20
  17. 2025A Survey on Agentic Multimodal Large Language Modelscited by 19
  18. 2026OS-Symphony: A Holistic Framework for Robust and Generalist Computer-Using Agentcited by 19
  19. 2026RLAnything: Forge Environment, Policy, and Reward Model in Completely Dynamic RL Systemcited by 19
  20. 2025MobileWorldBench: Towards Semantic World Modeling For Mobile Agentscited by 18
  21. 2026MolmoWeb: Open Visual Web Agent and Open Data for the Open Webcited by 17
  22. 2025CUARewardBench: A Benchmark for Evaluating Reward Models on Computer-using Agentcited by 16
  23. 2026PersonalAlign: Hierarchical Implicit Intent Alignment for Personalized GUI Agent with Long-Term User-Centric Recordscited by 16
  24. 2026GUI-GENESIS: Automated Synthesis of Efficient Environments with Verifiable Rewards for GUI Agent Post-Trainingcited by 16
  25. 2026GUI Agents for Continual Game Generationcited by 16
  26. 2025Ferret-UI Lite: Lessons from Building Small On-Device GUI Agentscited by 14
  27. 2026GUI-Libra: Training Native GUI Agents to Reason and Act with Action-aware Supervision and Partially Verifiable RLcited by 14
  28. 2025Environment Scaling for Interactive Agentic Experience Collection: A Surveycited by 13
  29. 2026OmegaUse: Building a General-Purpose GUI Agent for Autonomous Task Executioncited by 13
  30. 2026ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI Agentscited by 13
  31. 2026The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agentscited by 12
  32. 2025WebSTAR: Scalable Data Synthesis for Computer Use Agents with Step-Level Filteringcited by 11
  33. 2026Adaptive Milestone Reward for GUI Agentscited by 11
  34. 2026UI-Voyager: A Self-Evolving GUI Agent Learning via Failed Experiencecited by 11
  35. 2025ProRe: A Proactive Reward System for GUI Agents via Reasoner-Actor Collaborationcited by 10
  36. 2025R-WoM: Retrieval-augmented World Model For Computer-use Agentscited by 10
  37. 2025AgentProg: Empowering Long-Horizon GUI Agents with Program-Guided Context Managementcited by 10
  38. 2026Behavior Knowledge Merge in Reinforced Agentic Modelscited by 10

50 most cited of 205 All 205 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025A Survey on GUI Agents with Foundation Models Enhanced by Reinforcement Learning
  2. 2025GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning
  3. 2025CRAFT-GUI: Curriculum-Reinforced Agent For GUI Tasks
  4. 2025UItron: Foundational GUI Agent with Advanced Perception and Planning
  5. 2025UI-R1: Enhancing Efficient Action Prediction of GUI Agents by Reinforcement Learning
  6. 2025UI-S1: Advancing GUI Automation via Semi-online Reinforcement Learning
  7. 2025GUI-Bee: Align GUI Action Grounding to Novel Environments via Autonomous Exploration
  8. 2025Enhancing Visual Grounding for GUI Agents via Self-Evolutionary Reinforcement Learning