Attributing and Exploiting Safety Vectors through Global Optimization in Large Language Models

Add code
Jan 22, 2026

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: