Hugging Face's logo

pauljmorris
/
UI-TARS-7B-DPO

Image-Text-to-Text
Transformers
English
multimodal
gui
Inference Endpoints
Model card Files Community
pauljmorris commited on
Commit
4357791
·
verified ·
1 Parent(s): 3aa12e6

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +188 -4
README.md CHANGED
@@ -3,10 +3,194 @@ license: apache-2.0
3
  language:
4
  - en
5
  pipeline_tag: image-text-to-text
6
- library_name: transformers
7
  tags:
8
  - multimodal
9
  - gui
10
- base_model:
11
- - bytedance-research/UI-TARS-7B-DPO
12
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  language:
4
  - en
5
  pipeline_tag: image-text-to-text
 
6
  tags:
7
  - multimodal
8
  - gui
9
+ library_name: transformers
10
+ ---
11
+
12
+
13
+ # UI-TARS-7B-DPO
14
+ [UI-TARS-2B-SFT](https://huggingface.co/bytedance-research/UI-TARS-2B-SFT)  | 
15
+ [UI-TARS-7B-SFT](https://huggingface.co/bytedance-research/UI-TARS-7B-SFT)  | 
16
+ [**UI-TARS-7B-DPO**](https://huggingface.co/bytedance-research/UI-TARS-7B-DPO)(Recommended)  | 
17
+ [UI-TARS-72B-SFT](https://huggingface.co/bytedance-research/UI-TARS-72B-SFT)  | 
18
+ [**UI-TARS-72B-DPO**](https://huggingface.co/bytedance-research/UI-TARS-72B-DPO)(Recommended)
19
+ ## Introduction
20
+
21
+ UI-TARS is a next-generation native GUI agent model designed to interact seamlessly with graphical user interfaces (GUIs) using human-like perception, reasoning, and action capabilities. Unlike traditional modular frameworks, UI-TARS integrates all key components—perception, reasoning, grounding, and memory—within a single vision-language model (VLM), enabling end-to-end task automation without predefined workflows or manual rules.
22
+ <!-- ![Local Image](figures/UI-TARS.png) -->
23
+ <p align="center">
24
+ <img src="https://github.com/bytedance/UI-TARS/blob/main/figures/UI-TARS-vs-Previous-SOTA.png?raw=true" width="90%"/>
25
+ <p>
26
+ <p align="center">
27
+ <img src="https://github.com/bytedance/UI-TARS/blob/main/figures/UI-TARS.png?raw=true" width="90%"/>
28
+ <p>
29
+
30
+ <!-- ![Local Image](figures/UI-TARS-vs-Previous-SOTA.png) -->
31
+
32
+ This repository contains the model for the paper [UI-TARS: Pioneering Automated GUI Interaction with Native Agents](https://huggingface.co/papers/2501.12326).
33
+
34
+ Code: https://github.com/bytedance/UI-TARS
35
+
36
+ ## Performance
37
+ **Perception Capabilty Evaluation**
38
+ | Model | VisualWebBench | WebSRC | SQAshort |
39
+ |---------------------------|---------------|---------|----------|
40
+ | Qwen2-VL-7B | 73.3 | 81.8 | 84.9 |
41
+ | Qwen-VL-Max | 74.1 | 91.1 | 78.6 |
42
+ | Gemini-1.5-Pro | 75.4 | 88.9 | 82.2 |
43
+ | UIX-Qwen2-7B | 75.9 | 82.9 | 78.8 |
44
+ | Claude-3.5-Sonnet | 78.2 | 90.4 | 83.1 |
45
+ | GPT-4o | 78.5 | 87.7 | 82.3 |
46
+ | **UI-TARS-2B** | 72.9 | 89.2 | 86.4 |
47
+ | **UI-TARS-7B** | 79.7 | **93.6** | 87.7 |
48
+ | **UI-TARS-72B** | **82.8** | 89.3 | **88.6** |
49
+
50
+ **Grounding Capability Evaluation**
51
+ - **ScreenSpot Pro**
52
+
53
+ | Agent Model | Dev-Text | Dev-Icon | Dev-Avg | Creative-Text | Creative-Icon | Creative-Avg | CAD-Text | CAD-Icon | CAD-Avg | Scientific-Text | Scientific-Icon | Scientific-Avg | Office-Text | Office-Icon | Office-Avg | OS-Text | OS-Icon | OS-Avg | Avg-Text | Avg-Icon | Avg |
54
+ |--------------------------|----------|----------|----------|--------------|--------------|--------------|---------|---------|---------|---------------|---------------|---------------|------------|------------|------------|--------|--------|--------|---------|---------|------|
55
+ | QwenVL-7B | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.7 | 0.0 | 0.4 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.1 | 0.0 | **0.1** |
56
+ | GPT-4o | 1.3 | 0.0 | 0.7 | 1.0 | 0.0 | 0.6 | 2.0 | 0.0 | 1.5 | 2.1 | 0.0 | 1.2 | 1.1 | 0.0 | 0.9 | 0.0 | 0.0 | 0.0 | 1.3 | 0.0 | **0.8** |
57
+ | SeeClick | 0.6 | 0.0 | 0.3 | 1.0 | 0.0 | 0.6 | 2.5 | 0.0 | 1.9 | 3.5 | 0.0 | 2.0 | 1.1 | 0.0 | 0.9 | 2.8 | 0.0 | 1.5 | 1.8 | 0.0 | **1.1** |
58
+ | Qwen2-VL-7B | 2.6 | 0.0 | 1.3 | 1.5 | 0.0 | 0.9 | 0.5 | 0.0 | 0.4 | 6.3 | 0.0 | 3.5 | 3.4 | 1.9 | 3.0 | 0.9 | 0.0 | 0.5 | 2.5 | 0.2 | **1.6** |
59
+ | OS-Atlas-4B | 7.1 | 0.0 | 3.7 | 3.0 | 1.4 | 2.3 | 2.0 | 0.0 | 1.5 | 9.0 | 5.5 | 7.5 | 5.1 | 3.8 | 4.8 | 5.6 | 0.0 | 3.1 | 5.0 | 1.7 | **3.7** |
60
+ | ShowUI-2B | 16.9 | 1.4 | 9.4 | 9.1 | 0.0 | 5.3 | 2.5 | 0.0 | 1.9 | 13.2 | 7.3 | 10.6 | 15.3 | 7.5 | 13.5 | 10.3 | 2.2 | 6.6 | 10.8 | 2.6 | **7.7** |
61
+ | CogAgent-18B | 14.9 | 0.7 | 8.0 | 9.6 | 0.0 | 5.6 | 7.1 | 3.1 | 6.1 | 22.2 | 1.8 | 13.4 | 13.0 | 0.0 | 10.0 | 5.6 | 0.0 | 3.1 | 12.0 | 0.8 | **7.7** |
62
+ | Aria-UI | 16.2 | 0.0 | 8.4 | 23.7 | 2.1 | 14.7 | 7.6 | 1.6 | 6.1 | 27.1 | 6.4 | 18.1 | 20.3 | 1.9 | 16.1 | 4.7 | 0.0 | 2.6 | 17.1 | 2.0 | **11.3** |
63
+ | UGround-7B | 26.6 | 2.1 | 14.7 | 27.3 | 2.8 | 17.0 | 14.2 | 1.6 | 11.1 | 31.9 | 2.7 | 19.3 | 31.6 | 11.3 | 27.0 | 17.8 | 0.0 | 9.7 | 25.0 | 2.8 | **16.5** |
64
+ | Claude Computer Use | 22.0 | 3.9 | 12.6 | 25.9 | 3.4 | 16.8 | 14.5 | 3.7 | 11.9 | 33.9 | 15.8 | 25.8 | 30.1 | 16.3 | 26.9 | 11.0 | 4.5 | 8.1 | 23.4 | 7.1 | **17.1** |
65
+ | OS-Atlas-7B | 33.1 | 1.4 | 17.7 | 28.8 | 2.8 | 17.9 | 12.2 | 4.7 | 10.3 | 37.5 | 7.3 | 24.4 | 33.9 | 5.7 | 27.4 | 27.1 | 4.5 | 16.8 | 28.1 | 4.0 | **18.9** |
66
+ | UGround-V1-7B | - | - | 35.5 | - | - | 27.8 | - | - | 13.5 | - | - | 38.8 | - | - | 48.8 | - | - | 26.1 | - | - | **31.1** |
67
+ | **UI-TARS-2B** | 47.4 | 4.1 | 26.4 | 42.9 | 6.3 | 27.6 | 17.8 | 4.7 | 14.6 | 56.9 | 17.3 | 39.8 | 50.3 | 17.0 | 42.6 | 21.5 | 5.6 | 14.3 | 39.6 | 8.4 | **27.7** |
68
+ | **UI-TARS-7B** | 58.4 | 12.4 | 36.1 | 50.0 | 9.1 | 32.8 | **20.8**| 9.4 | **18.0**| 63.9 | **31.8** | **50.0** | **63.3** | 20.8 | 53.5 | 30.8 | **16.9**| 24.5 | 47.8 | 16.2 | **35.7** |
69
+ | **UI-TARS-72B** | **63.0** | **17.3** | **40.8** | **57.1** | **15.4** | **39.6** | 18.8 | **12.5**| 17.2 | **64.6** | 20.9 | 45.7 | **63.3** | **26.4** | **54.8** | **42.1**| 15.7 | **30.1**| **50.9**| **17.5**| **38.1** |
70
+
71
+
72
+ - **ScreenSpot**
73
+
74
+ | Method | Mobile-Text | Mobile-Icon/Widget | Desktop-Text | Desktop-Icon/Widget | Web-Text | Web-Icon/Widget | Avg |
75
+ |--------|-------------|-------------|-------------|-------------|-------------|---------|---------|
76
+ | **Agent Framework** | | | | | | | |
77
+ | GPT-4 (SeeClick) | 76.6 | 55.5 | 68.0 | 28.6 | 40.9 | 23.3 | **48.8** |
78
+ | GPT-4 (OmniParser) | 93.9 | 57.0 | 91.3 | 63.6 | 81.3 | 51.0 | **73.0** |
79
+ | GPT-4 (UGround-7B) | 90.1 | 70.3 | 87.1 | 55.7 | 85.7 | 64.6 | **75.6** |
80
+ | GPT-4o (SeeClick) | 81.0 | 59.8 | 69.6 | 33.6 | 43.9 | 26.2 | **52.3** |
81
+ | GPT-4o (UGround-7B) | 93.4 | 76.9 | 92.8 | 67.9 | 88.7 | 68.9 | **81.4** |
82
+ | **Agent Model** | | | | | | | |
83
+ | GPT-4 | 22.6 | 24.5 | 20.2 | 11.8 | 9.2 | 8.8 | **16.2** |
84
+ | GPT-4o | 20.2 | 24.9 | 21.1 | 23.6 | 12.2 | 7.8 | **18.3** |
85
+ | CogAgent | 67.0 | 24.0 | 74.2 | 20.0 | 70.4 | 28.6 | **47.4** |
86
+ | SeeClick | 78.0 | 52.0 | 72.2 | 30.0 | 55.7 | 32.5 | **53.4** |
87
+ | Qwen2-VL | 75.5 | 60.7 | 76.3 | 54.3 | 35.2 | 25.7 | **55.3** |
88
+ | UGround-7B | 82.8 | 60.3 | 82.5 | 63.6 | 80.4 | 70.4 | **73.3** |
89
+ | Aguvis-G-7B | 88.3 | 78.2 | 88.1 | 70.7 | 85.7 | 74.8 | **81.8** |
90
+ | OS-Atlas-7B | 93.0 | 72.9 | 91.8 | 62.9 | 90.9 | 74.3 | **82.5** |
91
+ | Claude Computer Use | - | - | - | - | - | - | **83.0** |
92
+ | Gemini 2.0 (Project Mariner) | - | - | - | - | - | - | **84.0** |
93
+ | Aguvis-7B | **95.6** | 77.7 | 93.8 | 67.1 | 88.3 | 75.2 | **84.4** |
94
+ | Aguvis-72B | 94.5 | **85.2** | 95.4 | 77.9 | **91.3** | **85.9** | **89.2** |
95
+ | **Our Model** | | | | | | | |
96
+ | **UI-TARS-2B** | 93.0 | 75.5 | 90.7 | 68.6 | 84.3 | 74.8 | **82.3** |
97
+ | **UI-TARS-7B** | 94.5 | **85.2** | **95.9** | 85.7 | 90.0 | 83.5 | **89.5** |
98
+ | **UI-TARS-72B** | 94.9 | 82.5 | 89.7 | **88.6** | 88.7 | 85.0 | **88.4** |
99
+
100
+
101
+ - **ScreenSpot v2**
102
+
103
+ | Method | Mobile-Text | Mobile-Icon/Widget | Desktop-Text | Desktop-Icon/Widget | Web-Text | Web-Icon/Widget | Avg |
104
+ |--------|-------------|-------------|-------------|-------------|-------------|---------|---------|
105
+ | **Agent Framework** | | | | | | | |
106
+ | GPT-4o (SeeClick) | 85.2 | 58.8 | 79.9 | 37.1 | 72.7 | 30.1 | **63.6** |
107
+ | GPT-4o (OS-Atlas-4B) | 95.5 | 75.8 | 79.4 | 49.3 | 90.2 | 66.5 | **79.1** |
108
+ | GPT-4o (OS-Atlas-7B) | 96.2 | 83.4 | 89.7 | 69.3 | **94.0** | 79.8 | **87.1** |
109
+ | **Agent Model** | | | | | | | |
110
+ | SeeClick | 78.4 | 50.7 | 70.1 | 29.3 | 55.2 | 32.5 | **55.1** |
111
+ | OS-Atlas-4B | 87.2 | 59.7 | 72.7 | 46.4 | 85.9 | 63.1 | **71.9** |
112
+ | OS-Atlas-7B | 95.2 | 75.8 | 90.7 | 63.6 | 90.6 | 77.3 | **84.1** |
113
+ | **Our Model** | | | | | | | |
114
+ | **UI-TARS-2B** | 95.2 | 79.1 | 90.7 | 68.6 | 87.2 | 78.3 | **84.7** |
115
+ | **UI-TARS-7B** | **96.9** | **89.1** | **95.4** | 85.0 | 93.6 | 85.2 | **91.6** |
116
+ | **UI-TARS-72B** | 94.8 | 86.3 | 91.2 | **87.9** | 91.5 | **87.7** | **90.3** |
117
+
118
+
119
+ **Offline Agent Capability Evaluation**
120
+ - **Multimodal Mind2Web**
121
+
122
+ | Method | Cross-Task Ele.Acc | Cross-Task Op.F1 | Cross-Task Step SR | Cross-Website Ele.Acc | Cross-Website Op.F1 | Cross-Website Step SR | Cross-Domain Ele.Acc | Cross-Domain Op.F1 | Cross-Domain Step SR |
123
+ |--------|----------------------|-------------------|--------------------|----------------------|--------------------|-------------------|--------------------|-------------------|-------------------|
124
+ | **Agent Framework** | | | | | | | | | |
125
+ | GPT-4o (SeeClick) | 32.1 | - | - | 33.1 | - | - | 33.5 | - | - |
126
+ | GPT-4o (UGround) | 47.7 | - | - | 46.0 | - | - | 46.6 | - | - |
127
+ | GPT-4o (Aria-UI) | 57.6 | - | - | 57.7 | - | - | 61.4 | - | - |
128
+ | GPT-4V (OmniParser) | 42.4 | 87.6 | 39.4 | 41.0 | 84.8 | 36.5 | 45.5 | 85.7 | 42.0 |
129
+ | **Agent Model** | | | | | | | | | |
130
+ | GPT-4o | 5.7 | 77.2 | 4.3 | 5.7 | 79.0 | 3.9 | 5.5 | 86.4 | 4.5 |
131
+ | GPT-4 (SOM) | 29.6 | - | 20.3 | 20.1 | - | 13.9 | 27.0 | - | 23.7 |
132
+ | GPT-3.5 (Text-only) | 19.4 | 59.2 | 16.8 | 14.9 | 56.5 | 14.1 | 25.2 | 57.9 | 24.1 |
133
+ | GPT-4 (Text-only) | 40.8 | 63.1 | 32.3 | 30.2 | 61.0 | 27.0 | 35.4 | 61.9 | 29.7 |
134
+ | Claude | 62.7 | 84.7 | 53.5 | 59.5 | 79.6 | 47.7 | 64.5 | 85.4 | 56.4 |
135
+ | Aguvis-7B | 64.2 | 89.8 | 60.4 | 60.7 | 88.1 | 54.6 | 60.4 | 89.2 | 56.6 |
136
+ | CogAgent | - | - | 62.3 | - | - | 54.0 | - | - | 59.4 |
137
+ | Aguvis-72B | 69.5 | 90.8 | 64.0 | 62.6 | 88.6 | 56.5 | 63.5 | 88.5 | 58.2 |
138
+ | **Our Model** | | | | | | | | | |
139
+ | **UI-TARS-2B** | 62.3 | 90.0 | 56.3 | 58.5 | 87.2 | 50.8 | 58.8 | 89.6 | 52.3 |
140
+ | **UI-TARS-7B** | 73.1 | 92.2 | 67.1 | 68.2 | 90.9 | 61.7 | 66.6 | 90.9 | 60.5 |
141
+ | **UI-TARS-72B** | **74.7** | **92.5** | **68.6** | **72.4** | **91.2** | **63.5** | **68.9** | **91.8** | **62.1** |
142
+
143
+
144
+ - **Android Control and GUI Odyssey**
145
+
146
+ | Agent Models | AndroidControl-Low Type | AndroidControl-Low Grounding | AndroidControl-Low SR | AndroidControl-High Type | AndroidControl-High Grounding | AndroidControl-High SR | GUIOdyssey Type | GUIOdyssey Grounding | GUIOdyssey SR |
147
+ |---------------------|----------------------|----------------------|----------------|----------------------|----------------------|----------------|----------------|----------------|----------------|
148
+ | Claude | 74.3 | 0.0 | 19.4 | 63.7 | 0.0 | 12.5 | 60.9 | 0.0 | 3.1 |
149
+ | GPT-4o | 74.3 | 0.0 | 19.4 | 66.3 | 0.0 | 20.8 | 34.3 | 0.0 | 3.3 |
150
+ | SeeClick | 93.0 | 73.4 | 75.0 | 82.9 | 62.9 | 59.1 | 71.0 | 52.4 | 53.9 |
151
+ | InternVL-2-4B | 90.9 | 84.1 | 80.1 | 84.1 | 72.7 | 66.7 | 82.1 | 55.5 | 51.5 |
152
+ | Qwen2-VL-7B | 91.9 | 86.5 | 82.6 | 83.8 | 77.7 | 69.7 | 83.5 | 65.9 | 60.2 |
153
+ | Aria-UI | -- | 87.7 | 67.3 | -- | 43.2 | 10.2 | -- | 86.8 | 36.5 |
154
+ | OS-Atlas-4B | 91.9 | 83.8 | 80.6 | 84.7 | 73.8 | 67.5 | 83.5 | 61.4 | 56.4 |
155
+ | OS-Atlas-7B | 93.6 | 88.0 | 85.2 | 85.2 | 78.5 | 71.2 | 84.5 | 67.8 | 62.0 |
156
+ | Aguvis-7B | -- | -- | 80.5 | -- | -- | 61.5 | -- | -- | -- |
157
+ | Aguvis-72B | -- | -- | 84.4 | -- | -- | 66.4 | -- | -- | -- |
158
+ | **UI-TARS-2B** | **98.1** | 87.3 | 89.3 | 81.2 | 78.4 | 68.9 | 93.9 | 86.8 | 83.4 |
159
+ | **UI-TARS-7B** | 98.0 | 89.3 | 90.8 | 83.7 | 80.5 | 72.5 | 94.6 | 90.1 | 87.0 |
160
+ | **UI-TARS-72B** | **98.1** | **89.9** | **91.3** | **85.2** | **81.5** | **74.7** | **95.4** | **91.4** | **88.6** |
161
+
162
+ **Online Agent Capability Evaluation**
163
+
164
+ | Method | OSWorld (Online) | AndroidWorld (Online) |
165
+ |--------|-------------------|------------------|
166
+ | **Agent Framework** | | |
167
+ | GPT-4o (UGround) | - | 32.8 |
168
+ | GPT-4o (Aria-UI) | 15.2 | 44.8 |
169
+ | GPT-4o (Aguvis-7B) | 14.8 | 37.1 |
170
+ | GPT-4o (Aguvis-72B) | 17.0 | - |
171
+ | GPT-4o (OS-Atlas-7B) | 14.6 | - |
172
+ | **Agent Model** | | |
173
+ | GPT-4o | 5.0 | 34.5 (SoM) |
174
+ | Gemini-Pro-1.5 | 5.4 | 22.8 (SoM) |
175
+ | Aguvis-72B | 10.3 | 26.1 |
176
+ | Claude Computer-Use | 14.9 (15 steps) | 27.9 |
177
+ | Claude Computer-Use | 22.0 (50 steps) | - |
178
+ | **Our Model** | | |
179
+ | **UI-TARS-7B-SFT** | 17.7 (15 steps) | 33.0 |
180
+ | **UI-TARS-7B-DPO** | 18.7 (15 steps) | - |
181
+ | **UI-TARS-72B-SFT** | 18.8 (15 steps) | **46.6** |
182
+ | **UI-TARS-72B-DPO** | **22.7** (15 steps) | - |
183
+ | **UI-TARS-72B-DPO** | **24.6** (50 steps) | - |
184
+
185
+
186
+ ## Citation
187
+ If you find our paper and model useful in your research, feel free to give us a cite.
188
+
189
+ ```BibTeX
190
+ @article{qin2025ui,
191
+ title={UI-TARS: Pioneering Automated GUI Interaction with Native Agents},
192
+ author={Qin, Yujia and Ye, Yining and Fang, Junjie and Wang, Haoming and Liang, Shihao and Tian, Shizuo and Zhang, Junda and Li, Jiahao and Li, Yunxin and Huang, Shijue and others},
193
+ journal={arXiv preprint arXiv:2501.12326},
194
+ year={2025}
195
+ }
196
+ ```