Safe Navigation on AndroidWorld core20 safe general tasks
11Success Count (out of 20)GPT-5
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5Setting=Base2026.04 | 11 | 55 | 100 | |
| CORASetting=Ours2026.04 | 8 | 40 | 72.7 | |
| AutoGLMSetting=Prev setting2026.04 | 6 | 30 | 54.5 | |
| UI-TARS-1.5Setting=Base, avg of 3 runs2026.04 | 4 | 20 | 36.4 |