We also report on reproducing Kharlapenko et al.'s "self explaining sparse autoencoder" features experiments: https://
transformer-circuits.pub/2024/august-up
date/index.html#self-explaining-sae
…

→ View original post on X — @anthropicai