UI Action Grounding on ScreenSpot Desktop
74.2Success Rate (Text)CogAgent
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CogAgentModel Size=18B2026.05 | 74.2 | — | — | 20 | |
| SeeClickModel Size=9.6B2026.05 | 72.2 | — | — | 30 | |
| Qwen2.5-VL-7B-Inst.Model Size=7B2026.05 | 57.7 | — | — | 42.1 | |
| Skill-CMIB2026.05 | 54.1 | — | — | 46.4 | |
| FuyuModel Size=8B2026.05 | 33 | — | — | 3.6 | |
| GPT-4V2026.05 | 20.2 | — | — | 11.8 | |
| MiniGPT-v2Model Size=7B2026.05 | 6.2 | — | — | 2.9 | |
| Qwen-VLModel Size=9.6B2026.05 | 5.7 | — | — | 5 | |
| FuyuSize=8B, Evaluation Protocol=Zero-shot2025.02 | — | 33 | 3.6 | — | |
| GPT-4VSize=n/a, Evaluation Protocol=Zero-shot2025.02 | — | 20.2 | 11.8 | — | |
| GPT-4V-OmniParserSize=n/a, Evaluation Protocol=Zero-shot, Detection Tool=OmniParser2025.02 | — | 92.7 | 49.4 | — | |
| Magma-8BSize=8.6B, Evaluation Protocol=Zero-shot2025.02 | — | 75.3 | 52.9 | — | |
| Qwen-VLSize=9.6B, Evaluation Protocol=Zero-shot2025.02 | — | 5.7 | 5 | — | |
| SeeClickSize=9.6B, Evaluation Protocol=Zero-shot2025.02 | — | 72.2 | 30 | — |