GUI Navigation on OSWorld Online evaluation
57.1Accuracy (50 steps)GUI-Cursor-7B (Qwen2.5-VL-7B) w/ o3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GUI-Cursor-7B (Qwen2.5-VL-7B) w/ o3Backbone=Qwen2.5-VL-7B, Reasoning Model=o32025.09 | 57.1 | — | — | |
| GUI-Cursor-7B (UI-TARS-1.5-7B) w/ o3Backbone=UI-TARS-1.5-7B, Reasoning Model=o32025.09 | 55.2 | — | — | |
| o3Reasoning Model=o32025.09 | 17.2 | 9.1 | 23 | |
| GTA1-32B w/ o3Backbone=GTA1-32B, Reasoning Model=o32025.09 | — | — | 55.4 | |
| GTA1-7B w/ o3Backbone=GTA1-7B, Reasoning Model=o32025.09 | — | — | 53.1 |