SAE Feature Interpretability on GPT-2 Small SAE Features Value Output
5.95Interpretability ScoreQuery Lens
Evaluation results
| Method | Links | |
|---|---|---|
| Query Lensvariant=QL_VALUE2026.05 | 5.95 | |
| Logit Lensvariant=LL_VALUE2026.05 | 4.87 |
| Method | Links | |
|---|---|---|
| Query Lensvariant=QL_VALUE2026.05 | 5.95 | |
| Logit Lensvariant=LL_VALUE2026.05 | 4.87 |