GUI Agent Interaction on MobileWorld GUI-Only
55.6SRGemini-3-Pro + UI-Ins-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-3-Pro + UI-Ins-7BFramework Type=Agentic Framework2026.04 | 55.6 | |
| GPT-5 + UI-Ins-7BFramework Type=Agentic Framework2026.04 | 54 | |
| Claude-4.5-Sonnet + UI-Ins-7BFramework Type=Agentic Framework2026.04 | 47.8 | |
| Doubao-1.5-UI-TARSFramework Type=End-to-End Model2026.04 | 26.3 | |
| MAI-UI-8BFramework Type=End-to-End Model2026.04 | 19.7 | |
| ClawGUI-2BFramework Type=Ours2026.04 | 17.1 | |
| UI-Venus-72BFramework Type=End-to-End Model2026.04 | 16.4 | |
| Qwen3-VL-235B-A22BFramework Type=End-to-End Model2026.04 | 12.8 | |
| Qwen3-VL-32BFramework Type=End-to-End Model2026.04 | 11.9 | |
| MAI-UI-2BFramework Type=End-to-End Model2026.04 | 11.1 | |
| Qwen3-VL-8BFramework Type=End-to-End Model2026.04 | 9.4 | |
| GUI-Owl-32BFramework Type=End-to-End Model2026.04 | 8.5 | |
| UI-Venus-7BFramework Type=End-to-End Model2026.04 | 8.5 | |
| GUI-Owl-7BFramework Type=End-to-End Model2026.04 | 7.7 |