Reward Modeling on AndroidWorld
92.5PrecisionVAGEN
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| VAGENTarget Model=Claude-Sonnet-4.52026.01 | 92.5 | 95.4 | 93.9 | 93.1 | |
| FullTrajEvalTarget Model=Claude-Sonnet-4.52026.01 | 88.1 | 90.8 | 89.4 | 87.9 | |
| ZeroGUITarget Model=Claude-Sonnet-4.52026.01 | 86.8 | 90.8 | 88.7 | 87.1 | |
| VAGENTarget Model=UI-TARS-1.5-7B2026.01 | 86.1 | 88.6 | 87.3 | 92.2 | |
| AndroidGenTarget Model=Claude-Sonnet-4.52026.01 | 85.3 | 89.2 | 87.2 | 85.3 | |
| WebRLTarget Model=Claude-Sonnet-4.52026.01 | 83.8 | 87.7 | 85.7 | 83.6 | |
| DistRLTarget Model=Claude-Sonnet-4.52026.01 | 82.9 | 89.2 | 85.9 | 83.6 | |
| ZeroGUITarget Model=UI-TARS-1.5-7B2026.01 | 80.6 | 82.9 | 81.7 | 88.8 | |
| DigiRLTarget Model=Claude-Sonnet-4.52026.01 | 79.1 | 81.5 | 80.3 | 77.6 | |
| DigiRLTarget Model=UI-TARS-1.5-7B2026.01 | 77.4 | 68.6 | 72.7 | 84.5 | |
| AndroidGenTarget Model=UI-TARS-1.5-7B2026.01 | 76.9 | 85.7 | 81.1 | 87.9 | |
| DistRLTarget Model=UI-TARS-1.5-7B2026.01 | 76.5 | 74.3 | 75.4 | 85.3 | |
| FullTrajEvalTarget Model=UI-TARS-1.5-7B2026.01 | 76.3 | 82.9 | 79.5 | 87.1 | |
| WebRLTarget Model=UI-TARS-1.5-7B2026.01 | 75.7 | 80 | 77.8 | 86.2 |