One Trigger Token Is Enough: A Defense Strategy for Balancing Safety and Usability in Large Language Models

Add code
May 12, 2025

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: