Poster

LEACE: Perfect linear concept erasure in closed form

Nora Belrose · David Schneider-Joseph · Shauli Ravfogel · Ryan Cotterell · Edward Raff · Stella Biderman

Great Hall & Hall B1+B2 (level 1) #1524
[ ] [ Project Page ]
Wed 13 Dec 3 p.m. PST — 5 p.m. PST

Abstract:

Concept erasure aims to remove specified features from a representation. It can improve fairness (e.g. preventing a classifier from using gender or race) and interpretability (e.g. removing a concept to observe changes in model behavior). We introduce LEAst-squares Concept Erasure (LEACE), a closed-form method which provably prevents all linear classifiers from detecting a concept while changing the representation as little as possible, as measured by a broad class of norms. We apply LEACE to large language models with a novel procedure called concept scrubbing, which erases target concept information from every layer in the network. We demonstrate our method on two tasks: measuring the reliance of language models on part-of-speech information, and reducing gender bias in BERT embeddings. Our code is available at https://github.com/EleutherAI/concept-erasure.

Chat is not available.