Towards Automated Circuit Discovery for Mechanistic Interpretability

Open in new window