GUI Agent Navigation and Action on AITZ
69.15Success Rate (SR)MobileIPL
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| MobileIPLMode=IPL2025.05 | 69.15 | 88.2 | 81.9 | — | 51.08 | 91.73 | 71.45 | 83.4 | 51.69 | 78.17 | |
| Falcon-UI-7BMode=PF2025.05 | 69.1 | — | 84.7 | — | — | — | — | — | — | — | |
| OS-Atlas-7BMode=PF2025.05 | 65.11 | 87.8 | 77.83 | — | 76.12 | 75.82 | 54.83 | 81.6 | 68.67 | 81.75 | |
| UI-Tars-7BMode=PF2025.05 | 65.11 | 89.97 | 77.59 | — | 52.5 | 83.03 | 64.27 | 82.76 | 61.87 | 74.35 | |
| AriaUI-MoEMode=FT2025.05 | 63.56 | 84.2 | 78.53 | — | 53.73 | 85.51 | 60.2 | 80.8 | 63.7 | 76.38 | |
| Qwen2-VL-7B (CoaT)Mode=FT2025.05 | 60.36 | 81.96 | 74.26 | — | 47.5 | 81.53 | 59.72 | 73.85 | 58.22 | 67.39 | |
| UGround-7BMode=PF2025.05 | 60.19 | 82.56 | 74.54 | — | 58.22 | 80.94 | 58.48 | 73.85 | 58.22 | 68.78 | |
| Learn from AITZBase Model=Qwen3-VL-8B2026.01 | 58.23 | 74.72 | 534 | -0.03 | — | — | — | — | — | — | |
| AITZ-SeedMode=FT2025.05 | 55.4 | 82.56 | 73.14 | — | 42.83 | 82.48 | 53.16 | 75.29 | 56.65 | 61.82 | |
| CogAgent (CoaT)Mode=ZS2025.05 | 53.28 | 45.8 | 72.59 | — | 70.22 | 88.23 | 66.15 | 21.8 | 45.95 | 24.6 | |
| Seeclick-7BMode=PF2025.05 | 49.11 | 53.8 | 62.93 | — | 11.14 | 69.92 | 52.96 | 53 | 67.88 | 55.36 | |
| AUTO-GUI (CoaT)Mode=FT2025.05 | 47.69 | 87.2 | 82.98 | — | 61.4 | 74.56 | 32.2 | 81.4 | 57.7 | 74.4 | |
| Agent-DiceBase Model=Qwen3-VL-8B2026.01 | 47.49 | 68.28 | 40 | 0.29 | — | — | — | — | — | — | |
| Zero-ShotBase Model=Qwen3-VL-8B2026.01 | 41.14 | 56.81 | 99 | -0.02 | — | — | — | — | — | — | |
| CL from all threeBase Model=Qwen3-VL-8B2026.01 | 35.97 | 60.03 | 59 | 0.01 | — | — | — | — | — | — | |
| Learn from GUI-OdysseyBase Model=Qwen3-VL-8B2026.01 | 32.01 | 58.06 | 20 | 0.1 | — | — | — | — | — | — | |
| CL from AITZ and AndroidControlBase Model=Qwen3-VL-8B2026.01 | 31.44 | 59.45 | 0 | -0.06 | — | — | — | — | — | — | |
| Learn from AndroidControlBase Model=Qwen3-VL-8B2026.01 | 20.56 | 49.12 | 0 | -0.3 | — | — | — | — | — | — |