Scoring Explanations Without Reading Them: Structural Shortcuts in Automated-Interpretability Evaluation Data
Automated interpretability scores an explanation of a sparse-autoencoder (SAE) feature by asking a model to predict, token by token, where the feature fires on evaluation examples or to judge whether marked tokens fit it.
By Xupeng ChenarXiv - cs.CL (NLP)
Read the full story
Originally published by arXiv - cs.CL (NLP) on Oct 30, 2026.
Comments
Sign in to join the discussion.