Codebook Features: Sparse and Discrete Interpretability for Neural Networks

Re-assign community