List your company

SAE Feature Interpretability on GPT-2 Small SAE Features Key Input

4.02Interpretability ScoreQuery Lens
2.99043.25773.5253.7923May 30, 2026

Evaluation results

MethodLinks
Query Lensvariant=QL_KEY2026.054.02
Logit Lensvariant=LL_KEY2026.053.03