Cover Image

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

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Wang, Haoming, Zou, Haoyang, Song, Huatong, Feng, Jiazhan, Fang, Junjie, Lu, Junting, Liu, Longxiang, Luo, Qinyu, Liang, Shihao, Huang, Shijue, Zhong, Wanjun, Ye, Yining, Qin, Yujia, Xiong, Yuwen, Song, Yuxin, Wu, Zhiyong, Li, Aoyan, Li, Bo, Dun, Chen, Liu, Chong, Zan, Daoguang, Leng, Fuxing, Wang, Hanbin, Yu, Hao, Chen, Haobin, Guo, Hongyi, Su, Jing, Huang, Jingjia, Shen, Kai, Shi, Kaiyu, Yan, Lin, Zhao, Peiyao, Liu, Pengfei, Ye, Qinghao, Zheng, Renjie, Xin, Shulin, Zhao, Wayne Xin, Heng, Wen, Huang, Wenhao, Wang, Wenqian, Qin, Xiaobo, Lin, Yi, Wu, Youbin, Chen, Zehui, Wang, Zihao, Zhong, Baoquan, Zhang, Xinchun, Li, Xujing, Li, Yuanfan, Zhao, Zhongkai, Jiang, Chengquan, Wu, Faming, Zhou, Haotian, Pang, Jinlin, Han, Li, Liu, Qi, Ma, Qianli, Liu, Siyao, Cai, Songhua, Fu, Wenqi, Liu, Xin, Wang, Yaohui, Zhang, Zhi, Zhou, Bo, Li, Guoliang, Shi, Jiajun, Yang, Jiale, Tang, Jie, Li, Li, Han, Qihua, Lu, Taoran, Lin, Woyu, Tong, Xiaokang, Li, Xinyao, Zhang, Yichi, Miao, Yu, Jiang, Zhengxuan, Li, Zili, Zhao, Ziyuan, Li, Chenxin, Ma, Dehua, Lin, Feng, Zhang, Ge, Yang, Haihua, Guo, Hangyu, Zhu, Hongda, Liu, Jiaheng, Du, Junda, Cai, Kai, Li, Kuanye, Yuan, Lichen, Han, Meilan, Wang, Minchao, Guo, Shuyue, Cheng, Tianhao, Ma, Xiaobo, Xiao, Xiaojun, Huang, Xiaolong, Chen, Xinjie, Du, Yidi, Chen, Yilin, Wang, Yiwen, Li, Zhaojian, Yang, Zhenzhu, Zeng, Zhiyuan, Jin, Chaolin, Li, Chen, Chen, Hao, Chen, Haoli, Chen, Jian, Zhao, Qinghao, Shi, Guang
Format:	Preprint
Published:	2025
Subjects:	Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition Human-Computer Interaction
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866912572658155520
author	Wang, Haoming Zou, Haoyang Song, Huatong Feng, Jiazhan Fang, Junjie Lu, Junting Liu, Longxiang Luo, Qinyu Liang, Shihao Huang, Shijue Zhong, Wanjun Ye, Yining Qin, Yujia Xiong, Yuwen Song, Yuxin Wu, Zhiyong Li, Aoyan Li, Bo Dun, Chen Liu, Chong Zan, Daoguang Leng, Fuxing Wang, Hanbin Yu, Hao Chen, Haobin Guo, Hongyi Su, Jing Huang, Jingjia Shen, Kai Shi, Kaiyu Yan, Lin Zhao, Peiyao Liu, Pengfei Ye, Qinghao Zheng, Renjie Xin, Shulin Zhao, Wayne Xin Heng, Wen Huang, Wenhao Wang, Wenqian Qin, Xiaobo Lin, Yi Wu, Youbin Chen, Zehui Wang, Zihao Zhong, Baoquan Zhang, Xinchun Li, Xujing Li, Yuanfan Zhao, Zhongkai Jiang, Chengquan Wu, Faming Zhou, Haotian Pang, Jinlin Han, Li Liu, Qi Ma, Qianli Liu, Siyao Cai, Songhua Fu, Wenqi Liu, Xin Wang, Yaohui Zhang, Zhi Zhou, Bo Li, Guoliang Shi, Jiajun Yang, Jiale Tang, Jie Li, Li Han, Qihua Lu, Taoran Lin, Woyu Tong, Xiaokang Li, Xinyao Zhang, Yichi Miao, Yu Jiang, Zhengxuan Li, Zili Zhao, Ziyuan Li, Chenxin Ma, Dehua Lin, Feng Zhang, Ge Yang, Haihua Guo, Hangyu Zhu, Hongda Liu, Jiaheng Du, Junda Cai, Kai Li, Kuanye Yuan, Lichen Han, Meilan Wang, Minchao Guo, Shuyue Cheng, Tianhao Ma, Xiaobo Xiao, Xiaojun Huang, Xiaolong Chen, Xinjie Du, Yidi Chen, Yilin Wang, Yiwen Li, Zhaojian Yang, Zhenzhu Zeng, Zhiyuan Jin, Chaolin Li, Chen Chen, Hao Chen, Haoli Chen, Jian Zhao, Qinghao Shi, Guang
author_facet	Wang, Haoming Zou, Haoyang Song, Huatong Feng, Jiazhan Fang, Junjie Lu, Junting Liu, Longxiang Luo, Qinyu Liang, Shihao Huang, Shijue Zhong, Wanjun Ye, Yining Qin, Yujia Xiong, Yuwen Song, Yuxin Wu, Zhiyong Li, Aoyan Li, Bo Dun, Chen Liu, Chong Zan, Daoguang Leng, Fuxing Wang, Hanbin Yu, Hao Chen, Haobin Guo, Hongyi Su, Jing Huang, Jingjia Shen, Kai Shi, Kaiyu Yan, Lin Zhao, Peiyao Liu, Pengfei Ye, Qinghao Zheng, Renjie Xin, Shulin Zhao, Wayne Xin Heng, Wen Huang, Wenhao Wang, Wenqian Qin, Xiaobo Lin, Yi Wu, Youbin Chen, Zehui Wang, Zihao Zhong, Baoquan Zhang, Xinchun Li, Xujing Li, Yuanfan Zhao, Zhongkai Jiang, Chengquan Wu, Faming Zhou, Haotian Pang, Jinlin Han, Li Liu, Qi Ma, Qianli Liu, Siyao Cai, Songhua Fu, Wenqi Liu, Xin Wang, Yaohui Zhang, Zhi Zhou, Bo Li, Guoliang Shi, Jiajun Yang, Jiale Tang, Jie Li, Li Han, Qihua Lu, Taoran Lin, Woyu Tong, Xiaokang Li, Xinyao Zhang, Yichi Miao, Yu Jiang, Zhengxuan Li, Zili Zhao, Ziyuan Li, Chenxin Ma, Dehua Lin, Feng Zhang, Ge Yang, Haihua Guo, Hangyu Zhu, Hongda Liu, Jiaheng Du, Junda Cai, Kai Li, Kuanye Yuan, Lichen Han, Meilan Wang, Minchao Guo, Shuyue Cheng, Tianhao Ma, Xiaobo Xiao, Xiaojun Huang, Xiaolong Chen, Xinjie Du, Yidi Chen, Yilin Wang, Yiwen Li, Zhaojian Yang, Zhenzhu Zeng, Zhiyuan Jin, Chaolin Li, Chen Chen, Hao Chen, Haoli Chen, Jian Zhao, Qinghao Shi, Guang
contents	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.
format	Preprint
id	arxiv_https___arxiv_org_abs_2509_02544
institution	arXiv
publishDate	2025
record_format	arxiv
spellingShingle	UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning Wang, Haoming Zou, Haoyang Song, Huatong Feng, Jiazhan Fang, Junjie Lu, Junting Liu, Longxiang Luo, Qinyu Liang, Shihao Huang, Shijue Zhong, Wanjun Ye, Yining Qin, Yujia Xiong, Yuwen Song, Yuxin Wu, Zhiyong Li, Aoyan Li, Bo Dun, Chen Liu, Chong Zan, Daoguang Leng, Fuxing Wang, Hanbin Yu, Hao Chen, Haobin Guo, Hongyi Su, Jing Huang, Jingjia Shen, Kai Shi, Kaiyu Yan, Lin Zhao, Peiyao Liu, Pengfei Ye, Qinghao Zheng, Renjie Xin, Shulin Zhao, Wayne Xin Heng, Wen Huang, Wenhao Wang, Wenqian Qin, Xiaobo Lin, Yi Wu, Youbin Chen, Zehui Wang, Zihao Zhong, Baoquan Zhang, Xinchun Li, Xujing Li, Yuanfan Zhao, Zhongkai Jiang, Chengquan Wu, Faming Zhou, Haotian Pang, Jinlin Han, Li Liu, Qi Ma, Qianli Liu, Siyao Cai, Songhua Fu, Wenqi Liu, Xin Wang, Yaohui Zhang, Zhi Zhou, Bo Li, Guoliang Shi, Jiajun Yang, Jiale Tang, Jie Li, Li Han, Qihua Lu, Taoran Lin, Woyu Tong, Xiaokang Li, Xinyao Zhang, Yichi Miao, Yu Jiang, Zhengxuan Li, Zili Zhao, Ziyuan Li, Chenxin Ma, Dehua Lin, Feng Zhang, Ge Yang, Haihua Guo, Hangyu Zhu, Hongda Liu, Jiaheng Du, Junda Cai, Kai Li, Kuanye Yuan, Lichen Han, Meilan Wang, Minchao Guo, Shuyue Cheng, Tianhao Ma, Xiaobo Xiao, Xiaojun Huang, Xiaolong Chen, Xinjie Du, Yidi Chen, Yilin Wang, Yiwen Li, Zhaojian Yang, Zhenzhu Zeng, Zhiyuan Jin, Chaolin Li, Chen Chen, Hao Chen, Haoli Chen, Jian Zhao, Qinghao Shi, Guang Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition Human-Computer Interaction 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.
title	UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning
topic	Artificial Intelligence Computation and Language Computer Vision and Pattern Recognition Human-Computer Interaction
url	https://arxiv.org/abs/2509.02544