GUI Grounding on ScreenSpot Pro (click accuracy)
61.9Click AccuracyUI-Venus-72B
Evaluation Results
| Method | Links | |
|---|---|---|
| UI-Venus-72BTraining Protocol=Fine-tuned2025.09 | 61.9 | |
| GTA1-72BTraining Protocol=Fine-tuned2025.09 | 58.4 | |
| UI-Ins-32BTraining Protocol=Fine-tuned2025.09 | 57 | |
| GTA1-32BTraining Protocol=Fine-tuned2025.09 | 53.6 | |
| UI-Ins-7BTraining Protocol=Fine-tuned2025.09 | 52.2 | |
| UI-Venus-7BTraining Protocol=Fine-tuned2025.09 | 50.8 | |
| GTA1-7BTraining Protocol=Fine-tuned2025.09 | 50.1 | |
| Mark-Grid ScaffoldTraining Protocol=Training-free2025.09 | 47.8 | |
| UI-TARS-1.5-7BTraining Protocol=Fine-tuned2025.09 | 42 | |
| Jedi-7BTraining Protocol=Fine-tuned2025.09 | 39.5 | |
| UI-TARS-72BTraining Protocol=Fine-tuned2025.09 | 38.1 | |
| Jedi-3BTraining Protocol=Fine-tuned2025.09 | 36.1 | |
| InfiGUI-R1-3BTraining Protocol=Fine-tuned2025.09 | 35.7 | |
| UI-TARS-7BTraining Protocol=Fine-tuned2025.09 | 35.7 | |
| UGround-V1-72BTraining Protocol=Fine-tuned2025.09 | 34.5 | |
| UI-R1-E-3BTraining Protocol=Fine-tuned2025.09 | 33.5 | |
| Aguvis-7BTraining Protocol=Fine-tuned2025.09 | 22.9 | |
| OS-ATLAS-7BTraining Protocol=Fine-tuned2025.09 | 18.9 | |
| UGround-7BTraining Protocol=Fine-tuned2025.09 | 16.5 | |
| Scaffold PromptingTraining Protocol=Training-free2025.09 | 11.9 | |
| CogAgent-18BTraining Protocol=Fine-tuned2025.09 | 7.7 | |
| ShowUI-2BTraining Protocol=Fine-tuned2025.09 | 7.7 | |
| OS-ATLAS-4BTraining Protocol=Fine-tuned2025.09 | 3.7 | |
| Grid-Augmented VisionTraining Protocol=Training-free2025.09 | 1.3 | |
| SeeClick-7BTraining Protocol=Fine-tuned2025.09 | 1.1 |