This repository contains code for replicating results from the NeurIPS 2022 paper "Towards Understanding Grokking: An Effective Theory of Representation Learning". We are currently working on improving the organization of the repository... stay tuned!