Autonomous GUI Interaction on GUIAct-Mobile
34.6Step Success RateGoClick-L w/ Intent Gnd.
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GoClick-L w/ Intent Gnd.Planner=GPT-4o, Grounding Model=GoClick-L w/ Intent Gnd. (ours)2026.04 | 34.6 | 29.6 | |
| GoClick-L w/ Func. Gnd.Planner=GPT-4o, Grounding Model=GoClick-L w/ Func. Gnd. (ours)2026.04 | 34.2 | 28.8 | |
| GoClick-L w/ Intent Gnd.Planner=Gemini-2-Flash-Exp, Grounding Model=GoClick-L w/ Intent Gnd. (ours)2026.04 | 28.7 | 28.6 | |
| GPT-4oPlanner=GPT-4o, Grounding Model=-2026.04 | 28.1 | 28.8 | |
| GoClick-L w/ Func. Gnd.Planner=Gemini-2-Flash-Exp, Grounding Model=GoClick-L w/ Func. Gnd. (ours)2026.04 | 27.2 | 26.1 | |
| GPT-4o + SoMPlanner=GPT-4o + SoM, Grounding Model=-2026.04 | 27.2 | 28.6 | |
| Gemini-2-Flash-Exp + SoMPlanner=Gemini-2-Flash-Exp + SoM, Grounding Model=-2026.04 | 23.3 | 25.2 | |
| Gemini-2-Flash-ExpPlanner=Gemini-2-Flash-Exp, Grounding Model=-2026.04 | 19.6 | 17.6 |