Background This repository contains an implementation of Relaxed Linear Adversarial Concept Erasure (RLACE). Given a dataset X of dense representation