GUI Navigation on Online-Mind2Web (Easy)
78.31Success RateOvis2.5S-RF++
Evaluation Results
| Method | Links | |
|---|---|---|
| Ovis2.5S-RF++Training Method=REINFORCE++ (Step-wise)2026.02 | 78.31 | |
| Ovis2.5S-GRPOTraining Method=GRPO (Step-wise)2026.02 | 77.11 | |
| Ovis2.5S-RLOOTraining Method=RLOO (Step-wise)2026.02 | 75.9 | |
| Claude 3.7 Sonnet CUModel Type=Proprietary2026.02 | 72.84 | |
| Claude 4 Sonnet CUModel Type=Proprietary2026.02 | 71.6 | |
| Ovis2.5SFTTraining Method=SFT2026.02 | 66.27 | |
| PageGuideBackbone=google/gemini-3-flash-preview2026.04 | 60.76 | |
| UI-TARSModel Scale=7B, Version=1.52026.02 | 57.83 | |
| PageGuideBackbone=google/gemini-2.5-flash-preview2026.04 | 52.5 | |
| SeeActBackbone=GPT-42026.04 | 51.8 | |
| Qwen3-VLModel Scale=32B2026.02 | 38.55 | |
| Qwen3-VLModel Scale=8B2026.02 | 34.94 |