Agent task performance on Coding 20-task (held-out)
81.2G ScoreOffline AW Controller
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Offline AW ControllerHarness configuration=AW, Protocol=Calibrated structural verifier2026.07 | 81.2 | 17.8 | |
| Base HarnessHarness configuration=Base, Protocol=Calibrated structural verifier2026.07 | 71.2 | — |