Reward Modeling on HelpSteer
0.543R2@50UARM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| UARMMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.543 | 0.387 | 0.423 | |
| TorchNautMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.527 | 0.499 | 0.519 | |
| MCNFMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.527 | 0.428 | 0.507 | |
| ClearMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.521 | 0.396 | 0.478 | |
| SCCPMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.512 | 0.491 | 0.553 | |
| ACIMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.476 | 0.469 | 0.544 | |
| Packed EnsembleMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.462 | 0.476 | 0.537 | |
| WCPMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.438 | 0.545 | 0.57 | |
| DERMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.42 | 0.557 | 0.581 | |
| CQRMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.409 | 0.432 | 0.458 | |
| Deep EnsembleMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.395 | 0.433 | 0.509 | |
| SCPMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.378 | 0.544 | 0.569 | |
| MC-DropoutMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.369 | 0.437 | 0.506 | |
| NaiveMiscoverage rate (alpha)=0.1, Confidence threshold=50% most confident samples2026.06 | 0.357 | 0.611 | 0.595 |