Short-term planning on OmniAct Desktop
92.38Type Success RateSSL-7B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SSL-7Bmode=Post-Training2026.01 | 92.38 | 83.39 | 83.39 | |
| GUI-R1-7Bmode=Post-Training2026.01 | 92.2 | 83.36 | 83.33 | |
| GUI-R1-3Bmode=Post-Training2026.01 | 91.86 | 78.37 | 78.31 | |
| RL-Continous-7Bmode=Post-Training2026.01 | 91.69 | 81.35 | 82.04 | |
| SSL-3Bmode=Post-Training2026.01 | 91.53 | 80.17 | 80.17 | |
| RL-Binary-7Bmode=Post-Training2026.01 | 91.52 | 81.35 | 81.35 | |
| Os-Atlas-7Bmode=Post-Training2026.01 | 90.24 | 62.87 | 56.73 | |
| RL-Continous-3Bmode=Post-Training2026.01 | 89.83 | 78.3 | 78.3 | |
| QwenVL2.5-7B*mode=Post-Training, SFT=Mix-3K2026.01 | 86.23 | 80.17 | 79.8 | |
| QwenVL2.5-7Bmode=Zero Shot2026.01 | 84.74 | 79.89 | 79.66 | |
| GPT-4omode=Zero Shot2026.01 | 79.97 | 63.25 | 50.67 | |
| RL-Binary-3Bmode=Post-Training2026.01 | 79.97 | 72.2 | 72.2 | |
| QwenVL2.5-3B*mode=Post-Training, SFT=Mix-3K2026.01 | 77.62 | 62.54 | 63.76 | |
| UI-R1-3Bmode=Post-Training2026.01 | 73.41 | 64.12 | 63.98 | |
| Os-Atlas-4Bmode=Post-Training2026.01 | 63.3 | 42.55 | 26.94 | |
| QwenVL2.5-3Bmode=Zero Shot2026.01 | 56.95 | 47.97 | 46.89 |