Why do universal adversarial attacks work on large language models?: Geometry might be the answer

Add code
Sep 01, 2023

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: