LEACE: Perfect linear concept erasure in closed form
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2023
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910902033317888 |
|---|---|
| author | Belrose, Nora Schneider-Joseph, David Ravfogel, Shauli Cotterell, Ryan Raff, Edward Biderman, Stella |
| author_facet | Belrose, Nora Schneider-Joseph, David Ravfogel, Shauli Cotterell, Ryan Raff, Edward Biderman, Stella |
| contents | Concept erasure aims to remove specified features from an embedding. 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 embedding 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. Code is available at https://github.com/EleutherAI/concept-erasure. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2306_03819 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | LEACE: Perfect linear concept erasure in closed form Belrose, Nora Schneider-Joseph, David Ravfogel, Shauli Cotterell, Ryan Raff, Edward Biderman, Stella Machine Learning Computation and Language Computers and Society Concept erasure aims to remove specified features from an embedding. 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 embedding 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. Code is available at https://github.com/EleutherAI/concept-erasure. |
| title | LEACE: Perfect linear concept erasure in closed form |
| topic | Machine Learning Computation and Language Computers and Society |
| url | https://arxiv.org/abs/2306.03819 |