SAE Feature Interpretability on Qwen-3-1.7B SAE Features Value Output
6.96Interpretability ScoreQuery Lens
Evaluation results
| Method | Links | |
|---|---|---|
| Query Lensvariant=QL_VALUE2026.05 | 6.96 | |
| Logit Lensvariant=LL_VALUE2026.05 | 5.06 |
| Method | Links | |
|---|---|---|
| Query Lensvariant=QL_VALUE2026.05 | 6.96 | |
| Logit Lensvariant=LL_VALUE2026.05 | 5.06 |