Toggle control execution on Dynamic evaluation benchmark
55Task Success RateOS-Atlas-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| OS-Atlas-7BStaR training=w/2025.09 | 55 | |
| AgentCPM-GUI-8BStaR training=w/2025.09 | 42.5 | |
| UI-TARS-7BStaR training=w/2025.09 | 40 | |
| UI-TARS-7BStaR training=w/o2025.09 | 35 | |
| AgentCPM-GUI-8BStaR training=w/o2025.09 | 20 | |
| OS-Atlas-7BStaR training=w/o2025.09 | 10 |