Element Grounding on ScreenSpot v2 (test)
88Grounding AccuracyUGround-v1-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| UGround-v1-7BSize=7B, Type=UI Experts2025.02 | 88 | |
| OS-ATLASSize=7B, Type=UI Experts2025.02 | 84.1 | |
| Qwen2-VL-AutoGUI702kSize=7B, Type=Finetuned, Training Data=AutoGUI 702k2025.02 | 83.2 | |
| Qwen2-VLSize=7B, Type=General2025.02 | 66.9 | |
| SeeClick w/ AutoGUI702kSize=10B, Type=Finetuned, Training Data=AutoGUI 702k2025.02 | 54.7 | |
| SeeClickSize=10B, Type=UI Experts2025.02 | 54 | |
| CogAgentSize=18B, Type=UI Experts2025.02 | 52.1 | |
| SliME-AutoGUI702kSize=8B, Type=Finetuned, Training Data=AutoGUI 702k2025.02 | 42.5 | |
| Qwen-VL-AutoGUI702kSize=10B, Type=Finetuned, Training Data=AutoGUI 702k2025.02 | 40.2 | |
| GPT-4oSize=N/A, Type=General2025.02 | 20.4 | |
| SliMESize=8B, Type=General2025.02 | 13.4 | |
| Llama-3.2-Vision-InstructSize=11B, Type=General2025.02 | 11.6 | |
| Qwen-VLSize=10B, Type=General2025.02 | 5.6 |