GUI Task Completion on RISK-Bench (online)
88.7Completion RateQwen2.5-VL-72B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-VL-72BModel Category=General Open-source Models2025.09 | 88.7 | 68.9 | |
| RISK-R1-7BModel Category=Ours2025.09 | 87.6 | 70.5 | |
| RISK-SFT-7BModel Category=Ours2025.09 | 86.1 | 67 | |
| Qwen-VL-MaxModel Category=Commercial Models2025.09 | 85.2 | 66.2 | |
| Qwen2.5-VL-7BModel Category=General Open-source Models2025.09 | 8.3 | 46.4 | |
| OS-Atlas-7BModel Category=GUI-specific Models, Training Protocol=SFT2025.09 | 4.4 | 37.2 | |
| InternVL3-8BModel Category=General Open-source Models2025.09 | 0 | 0 | |
| UI-TARS-7BModel Category=GUI-specific Models, Training Protocol=SFT2025.09 | 0 | 0 | |
| UI-TARS-72BModel Category=GUI-specific Models, Training Protocol=SFT2025.09 | 0 | 0 | |
| ShowUI-2BModel Category=GUI-specific Models, Training Protocol=SFT2025.09 | 0 | 0 | |
| GUI-R1-7BModel Category=GUI-specific Models, Training Protocol=RL2025.09 | 0 | 0 | |
| InfiGUI-R1-3BModel Category=GUI-specific Models, Training Protocol=RL2025.09 | 0 | 0 | |
| GUI-G1-3BModel Category=GUI-specific Models, Training Protocol=RL2025.09 | 0 | 0 | |
| UI-TARS-1.5-7BModel Category=GUI-specific Models, Training Protocol=RL2025.09 | 0 | 0 |