GUI Navigation on AndroidControl High (SR, GR, Type)
76.3SR (Success Rate)UILoop-7B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| UILoop-7BParadigm=UILoop Training Models2026.04 | 76.3 | 88.9 | 81.8 | |
| UILoop-3BParadigm=UILoop Training Models2026.04 | 70.5 | 85.3 | 68.9 | |
| SeeClickParadigm=Screen-to-Action Training Models2026.04 | 59.1 | 82.9 | 62.9 | |
| GUICrafter–3BTraining Stage=Full Stage 1+22026.06 | 56.5 | 73.38 | 58.96 | |
| GUICrafter–3BReward Function=Binary Reward2026.06 | 54.89 | 72.44 | 57.41 | |
| GUICrafter–3BTraining Stage=only Stage22026.06 | 54.17 | 71.73 | 58.18 | |
| GUI-R1-7BParadigm=Screen-to-Action Training Models2026.04 | 51.7 | 71.6 | 65.6 | |
| Qwen2.5-VL-7B*Supervised Fine-tuning=true (GUI-R1-3K)2026.06 | 48.11 | 69.15 | 58.69 | |
| Qwen2.5-VL-7B*Paradigm=Screen-to-Action Training Models, SFT Source=Luo et al., 20252026.04 | 48.1 | 69.2 | 58.7 | |
| Qwen2.5-VL-7BParadigm=Zero-Shot Models2026.04 | 47.1 | 68.7 | 59.7 | |
| Qwen2.5-VL-7BSupervised Fine-tuning=false2026.06 | 47.06 | 68.67 | 59.71 | |
| GUI-R1-3BParadigm=Screen-to-Action Training Models2026.04 | 46.6 | 58 | 56.2 | |
| GUI-R1-3B2026.06 | 46.55 | 58.04 | 56.24 | |
| UI-R1-3B2026.06 | 45.44 | 57.85 | 55.7 | |
| UI-R1-3BParadigm=Screen-to-Action Training Models2026.04 | 45.4 | 57.9 | 55.7 | |
| GUICrafter–3BTraining Stage=only Stage12026.06 | 44.65 | 69.64 | 57.57 | |
| Qwen2.5-VL-3B*Supervised Fine-tuning=true (GUI-R1-3K)2026.06 | 41.22 | 52.05 | 49.53 | |
| Qwen2.5-VL-3B*Paradigm=Screen-to-Action Training Models, SFT Source=Luo et al., 20252026.04 | 41.2 | 52.1 | 49.5 | |
| Qwen2.5-VL-3BParadigm=Zero-Shot Models2026.04 | 38.9 | 47.8 | 46.5 | |
| Qwen2.5-VL-3BSupervised Fine-tuning=false2026.06 | 38.9 | 47.81 | 46.51 | |
| GUI-Owl-7BParadigm=Screen-to-Action Training Models2026.04 | 37.5 | 72.9 | 53.7 | |
| OS-Atlas-7B2026.06 | 29.83 | 57.44 | 54.9 | |
| OS-Atlas-7BParadigm=Screen-to-Action Training Models2026.04 | 29.8 | 57.4 | 54.9 | |
| OS-Atlas-4BParadigm=Screen-to-Action Training Models2026.04 | 22.8 | 49 | 49.5 | |
| GPT-4oParadigm=Zero-Shot Models2026.04 | 21.2 | 63.1 | 30.9 | |
| GPT-4o2026.06 | 21.17 | 63.06 | 30.9 | |
| OS-Atlas-Pro-7BParadigm=Screen-to-Action Training Models2026.04 | 18.3 | 69.7 | 16.8 | |
| Claude-CUParadigm=Zero-Shot Models2026.04 | 12.5 | 63.7 | — | |
| Aria-UIParadigm=Screen-to-Action Training Models2026.04 | 10.2 | — | 43.2 |