Bypassing the Safety Training of Open-Source LLMs with Priming Attacks.

Jason Vega, Isha Chaudhary, Changming Xu, Gagandeep Singh

Browse the full ICLR paper archive.