---
license: apache-2.0
---

These are the runs from the paper [Towards Automated Circuit Discovery for Mechanistic Interpretability](https://arxiv.org/abs/2304.14997) (Arthur Conmy et al., 2023).

This repository contains the actual hypotheses that the various algorithms tested found. It's intended to help with reproducibility.