Abstract:

Despite efforts to align large language models (LLMs) with human intentions, popular LLMs such as chatGPT, Llama, Claude, and Gemini are susceptible to jailbreaking attacks, wherein an adversary fools a targeted LLM into generating objectionable content. For this reason, interest has grown in improving the robustness of LLMs against such attacks. In this talk, we review the current state of the jailbreaking literature, including new questions about robust generalization, discussions of new black-box attacks on LLMs, defenses against jailbreaking attacks, and a new leaderboard to evaluate the robust generalization of production LLMs.

Zoom Link (if unable to attend in-person): https://upenn.zoom.us/j/93335180566