Reward Modeling on OSWorld Verified Class-Balanced Scripts 1.0 (test)
79.2PrecisionVAGEN
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| VAGENVerified Agent Backbone=Claude-Sonnet-4.52026.01 | 79.2 | 94.4 | 86.2 | 83.1 | |
| AndroidGenVerified Agent Backbone=Claude-Sonnet-4.52026.01 | 78.8 | 76.6 | 77.7 | 74.4 | |
| WebRLVerified Agent Backbone=Claude-Sonnet-4.52026.01 | 77.8 | 76.7 | 77.3 | 73.9 | |
| FullTrajEvalVerified Agent Backbone=Claude-Sonnet-4.52026.01 | 75.6 | 90.1 | 82.2 | 78.1 | |
| ZeroGUIVerified Agent Backbone=Claude-Sonnet-4.52026.01 | 75.1 | 89.2 | 81.5 | 77.2 | |
| DistRLVerified Agent Backbone=Claude-Sonnet-4.52026.01 | 74.9 | 74.1 | 74.5 | 70.9 | |
| DigiRLVerified Agent Backbone=Claude-Sonnet-4.52026.01 | 73.9 | 73.5 | 73.7 | 70.2 |