Opening Mechanistic Interpretability: How Researchers Try To See Inside Models from Linear Digressions. Pick your app and we’ll remember it.
Just open the episode page