Automated Interpretability on Pythia-160M
0.857AutoInterp ScoreEnsembling (NB)
Evaluation Results
| Method | Links | |
|---|---|---|
| Ensembling (NB)Ensemble size=82025.05 | 0.857 | |
| Expanded SAE2025.05 | 0.852 | |
| Ensembling (Boosting)Ensemble size=82025.05 | 0.852 |
| Method | Links | |
|---|---|---|
| Ensembling (NB)Ensemble size=82025.05 | 0.857 | |
| Expanded SAE2025.05 | 0.852 | |
| Ensembling (Boosting)Ensemble size=82025.05 | 0.852 |