REVS: Unlearning Sensitive Information in Language Models via Rank Editing in the Vocabulary Space.

Tomer Ashuach, Martin Tutek, Yonatan Belinkov

Browse the full ACL paper archive.