Grounding on ScreenSpot-Pro macOS
41.39Grounding AccuracyGUIrilla-See (7B)
Evaluation Results
| Method | Links | |
|---|---|---|
| GUIrilla-See (7B)Platform=macOS, Data=Synthetic, # Images=4.2K2025.10 | 41.39 | |
| UI-TARS (72B)Platform=Multi-OS, Data=Real + Synthetic, # Images=∼20M (est.)2025.10 | 40 | |
| Qwen 2.5 VL (7B)Platform=Multi-OS, Data=Real + Synthetic, # Images=-2025.10 | 34.27 | |
| GUIrilla-See (3B)Platform=macOS, Data=Synthetic, # Images=4.2K2025.10 | 32.62 | |
| UI-TARS (7B)Platform=Multi-OS, Data=Real + Synthetic, # Images=∼20M (est.)2025.10 | 27.7 | |
| OS-Atlas (7B)Platform=Multi-OS, Data=Synthetic, # Images=2.2M2025.10 | 20 | |
| Qwen 2.5 VL (3B)Platform=Multi-OS, Data=Real + Synthetic, # Images=-2025.10 | 18.37 | |
| UI-TARS (2B)Platform=Multi-OS, Data=Real + Synthetic, # Images=∼20M (est.)2025.10 | 15.4 | |
| UGround (7B)Platform=Web + Android, Data=Real, # Images=1.3M2025.10 | 12.3 | |
| ShowUI (2B)Platform=Multi-OS, Data=Real + Synthetic, # Images=256K2025.10 | 10.8 | |
| GUIrilla-See (0.7B)Platform=macOS, Data=Synthetic, # Images=4.2K2025.10 | 7.95 | |
| CogAgent (18B)Platform=Multi-OS, Data=Real + Synthetic, # Images=40M2025.10 | 4.6 | |
| Florence-2 (0.7B)Platform=Multi-OS, Data=Real + Synthetic, # Images=-2025.10 | 0.16 |