SNAP: Unlearning Selective Knowledge in Large Language Models with Negative Instructions

Open in new window