Action Prediction on ScreenSpot (Web)
70.4Success Rate (Text Actions)CogAgent
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CogAgentModel Size=18B2026.05 | 70.4 | 28.6 | |
| Skill-CMIB2026.05 | 60 | 49.5 | |
| SeeClickModel Size=9.6B2026.05 | 55.7 | 32.5 | |
| Qwen2.5-VL-7B-Inst.Model Size=7B2026.05 | 51.7 | 46.6 | |
| FuyuModel Size=8B2026.05 | 33.9 | 4.4 | |
| GPT-4V2026.05 | 9.2 | 8.8 | |
| MiniGPT-v2Model Size=7B2026.05 | 6.5 | 3.4 | |
| Qwen-VLModel Size=9.6B2026.05 | 3.5 | 2.4 |