Unseen UI Generalization on OSExpert Eval
25Tableau Success RateOSExpert (Explore w/ Qwen-3-VL-8B)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OSExpert (Explore w/ Qwen-3-VL-8B)Agent Type=Agentic Framework2026.03 | 25 | 37 | |
| OSExpert (Explore w/ GPT-5 & UI-TARS)Agent Type=Agentic Framework2026.03 | 25 | 37 | |
| OSExpert (Explore w/ Qwen-3-VL-8B, wo/ Fine Skills)Agent Type=Agentic Framework2026.03 | 23 | 30 | |
| CoAct-1Agent Type=Agentic Framework2026.03 | 5 | 6 | |
| Agent-S3 w/ GPT-5Agent Type=Agentic Framework2026.03 | 3 | 2 | |
| OpenCUA-7BAgent Type=Specialized Model2026.03 | 2 | 0 | |
| Computer-Use-PreviewAgent Type=Specialized Model2026.03 | 0 | 0 | |
| Qwen-3-VL-8BAgent Type=General Model2026.03 | 0 | 0 |