Mobile GUI Agent Execution on AndroidControl Curated-Hard
80.8Type RateMobileGen
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MobileGenBackbone=Qwen3-VL-8B-Inst., Training Strategy=MobileGen2026.01 | 80.8 | 45.5 | 38.4 | |
| MobileGenBackbone=InternVL3-14B, Training Strategy=MobileGen2026.01 | 78.6 | 46.7 | 36.6 | |
| MobileGenBackbone=Qwen3-VL-4B-Inst., Training Strategy=MobileGen2026.01 | 77.6 | 37.9 | 35.6 | |
| OS-GenesisBackbone=Qwen3-VL-8B-Inst., Training Strategy=OS-Genesis2026.01 | 75.2 | 31.2 | 29.6 | |
| GUI-RewalkBackbone=InternVL3-14B, Training Strategy=GUI-Rewalk2026.01 | 73.8 | 36.4 | 32.8 | |
| GUI-RewalkBackbone=Qwen3-VL-8B-Inst., Training Strategy=GUI-Rewalk2026.01 | 73 | 33.3 | 31.6 | |
| OS-GenesisBackbone=Qwen3-VL-4B-Inst., Training Strategy=OS-Genesis2026.01 | 72.8 | 29.7 | 27.8 | |
| GUI-RewalkBackbone=Qwen3-VL-4B-Inst., Training Strategy=GUI-Rewalk2026.01 | 71.8 | 28.8 | 29.2 | |
| OS-GenesisBackbone=InternVL3-14B, Training Strategy=OS-Genesis2026.01 | 71.6 | 37.9 | 31.2 | |
| Qwen3-VL-8B-Inst.Backbone=Qwen3-VL-8B-Inst., Training Strategy=Zero-shot2026.01 | 71.2 | 27.9 | 24 | |
| Qwen3-VL-4B-Inst.Backbone=Qwen3-VL-4B-Inst., Training Strategy=Zero-shot2026.01 | 70.8 | 26.4 | 25.6 | |
| InternVL3-14BBackbone=InternVL3-14B, Training Strategy=Zero-shot2026.01 | 69.4 | 32.7 | 27.6 | |
| GPT-5Backbone=GPT-52026.01 | 66.2 | 14.5 | 16.6 | |
| UI-TARS-1.5-7BBackbone=UI-TARS-1.5-7B2026.01 | 62.6 | 37.6 | 28.2 | |
| GUI-Owl-7BBackbone=GUI-Owl-7B2026.01 | 62.2 | 64.5 | 41.6 |