GUI reasoning on AndroidControl High
65.91TypeCAPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CAPOPrompting=zero-shot2025.12 | 65.91 | 61.47 | 47.71 | |
| GRPOPrompting=zero-shot2025.12 | 60.1 | 58.25 | 46.81 | |
| Os-Atlas-4BPrompting=zero-shot2025.12 | 49.01 | 49.51 | 22.77 | |
| QwenVL2.5-3BPrompting=zero-shot2025.12 | 47.81 | 46.51 | 38.9 |