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

Source: https://huggingface.co/papers/2509.02544 Published on Sep 2, 2025

#2 Paper of the day Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

UI-TARS-2, a native GUI-centered agent model, addresses challenges in data scalability, multi-turn reinforcement learning, and environment stability, achieving significant improvements over its predecessor and strong baselines across various benchmarks.

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-turnreinforcement learning(RL), the limitations ofGUI-only operation, and environment stability. In this technical report, we present UI-TARS-2, a nativeGUI-centered agent model that addresses these challenges through a systematic training methodology: adata flywheelfor scalable data generation, a stabilizedmulti-turn RLframework, a hybridGUIenvironment 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. OnGUIbenchmarks, it reaches 88.2 onOnline-Mind2Web, 47.5 onOSWorld, 50.6 onWindowsAgentArena, and 73.3 onAndroidWorld, 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) onLMGame-Bench. Additionally, the model can generalize tolong-horizon information-seeking tasksandsoftware 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 ofGUIagents and exhibit strong generalization to real-world interactive scenarios.

View arXiv pageView PDFProject pageGitHub10.3kautoAdd to collection

Get this paper in your agent:

hf papers read 2509\.02544

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper2

#### meituan/EvoCUA-32B-20260105 33B• UpdatedMar 31 • 884 • 25 #### meituan/EvoCUA-8B-20260105 9B• UpdatedMar 31 • 2.92k • 14

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2509.02544 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2509.02544 in a Space README.md to link it from this page.

Collections including this paper22

Browse 22 collections that include this paper