Research on Generalizable Reward Learning and Large Language Model (LLM) Alignment

Yawen Duan (he/him) is a PhD student in Engineering at University of Cambridge. He currently focuses on developing detection tools, algorithms, and empirical evaluation to make reward modeling methods robust to distribution shifts and avoid overoptimization. Prior to Cambridge, Yawen worked at the Center for Human-Compatible AI, UC Berkeley, focusing on reward model robustness and adversarial policies against superhuman AI systems. He was also a research intern at Huawei Noah’s Ark Lab. Yawen received a MPhil in Machine Learning at the University of Cambridge, and a BSc at the University of Hong Kong. For more information, see his website.