Misalignment Detection on All datasets 9 5 in-dist + 4 OOD (pooled)
93AUROC18-indicator ensemble
Evaluation Results
| Method | Links | |
|---|---|---|
| 18-indicator ensembleLayer=272026.06 | 93 | |
| 18-indicator ensembleLayer=292026.06 | 92.3 | |
| RepE baselineLayer=272026.06 | 87.4 | |
| RepE baselineLayer=292026.06 | 82.5 |