Scalable and Robust LLM Unlearning by Correcting Responses with Retrieved Exclusions

Open in new window