Machine Unlearning

LLMs tend to memorize information including PI information leading to privacy issues. Machine Unlearning is a new setup wherein a pretrained model is made to “forget” certain information regarding some users who request their data to be removed.