Selective Differential Privacy for Language Modeling

Weiyan Shi, Aiqi Cui, Evan Li +2

With the increasing applications of language models, it has become crucial to protect these models from leaking private information. Previous work has attempted to tackle this chal…