Sizhe Chen (陈思哲)

Hi! I am a CS Ph.D. candidate at UC Berkeley AI Research (BAIR) with David Wagner (also closely with Chuan Guo and Sewon Min). I study real-world AI security problems with general and principled solutions. I led research towards the most secure open LLMs against prompt injection, the top-1 threat to AI agents. Featured work includes StruQ, (Meta-)SecAlign, SecOPD, and SecPO, see slides for my talks and lectures. Check out our latest Qwen3.6-27B-Instruct-SecPO that is robust to the current strongest adaptive attacks. My research is supported by industry fundings, including NVIDIA Graduate Fellowship, Meta FAIR and Google DeepMind (BAIR Commons). I spent two great years at Meta FAIR. I obtained an M.Eng. and a B.Eng. from Shanghai Jiao Tong University.