[Articles] OPT-OUT: Investigating Entity-Level Unlearning for Large Language Models via Optimal Transport
Developed a new unlearning technique (OPT-OUT), a technique that removes a certain entity’s data without seriously harming model’s performance.