EMMA: Evaluation and Mitigation of Malicious Behaviors in LLMs

This project aims to systematically study security risks in foundation models, with a specific focus on backdoor threats.