x
Sparse Coding, for Mechanistic Interpretability and Activation Engineering — AI Alignment Forum