Mobile Agent Navigation on GUI Odyssey 1.0 (test)
81.7Step Success Rate (SR)ADPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ADPONumber of samples (N)=82026.01 | 81.7 | 94.8 | 84.7 | |
| ADPONumber of samples (N)=42026.01 | 81.6 | 94.7 | 84.5 | |
| GRPONumber of samples (N)=82026.01 | 81.5 | 94.6 | 84.4 | |
| ADPONumber of samples (N)=122026.01 | 81.4 | 94.4 | 84.5 | |
| GRPONumber of samples (N)=42026.01 | 81.3 | 94.7 | 83.9 | |
| GRPONumber of samples (N)=122026.01 | 81.1 | 94.5 | 84 | |
| GRPONumber of samples (N)=12026.01 | 79.8 | 93.1 | 83.9 | |
| ADPONumber of samples (N)=12026.01 | 79.7 | 94.2 | 82.5 | |
| SpiritSight-8BNumber of samples (N)=12026.01 | 75.8 | — | — | |
| AgentCPM-GUI-8BNumber of samples (N)=12026.01 | 75 | 90.8 | — | |
| UI-TARS-7BNumber of samples (N)=12026.01 | 67.9 | 86.1 | — | |
| Qwen2.5-VL-7B (Base)Number of samples (N)=12026.01 | 52.8 | 81.1 | 61.4 | |
| Qwen2.5-VL-7B (Base)Number of samples (N)=122026.01 | 46.9 | 76.9 | 55.5 | |
| Qwen2.5-VL-7B (Base)Number of samples (N)=82026.01 | 46.6 | 76.7 | 55.4 | |
| Qwen2.5-VL-7B (Base)Number of samples (N)=42026.01 | 46.5 | 76.9 | 55.3 |