GUI Navigation on Online-Mind2Web (Medium)
66.91Success RateClaude 3.7 Sonnet CU
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude 3.7 Sonnet CUModel Type=Proprietary2026.02 | 66.91 | |
| Ovis2.5S-RF++Training Method=REINFORCE++ (Step-wise)2026.02 | 65.73 | |
| Claude 4 Sonnet CUModel Type=Proprietary2026.02 | 65.47 | |
| Ovis2.5S-GRPOTraining Method=GRPO (Step-wise)2026.02 | 62.94 | |
| Ovis2.5S-RLOOTraining Method=RLOO (Step-wise)2026.02 | 60.84 | |
| Ovis2.5SFTTraining Method=SFT2026.02 | 51.75 | |
| Qwen3-VLModel Scale=32B2026.02 | 27.97 | |
| UI-TARSModel Scale=7B, Version=1.52026.02 | 27.27 | |
| Qwen3-VLModel Scale=8B2026.02 | 25.87 |