Jiayi Zhou (周嘉懿)

Email: gaiejj@outlook.com

 

Hello! I’m a third-year PhD student at the Institute of Artificial Intelligence, Peking University, advised by Prof. Yaodong Yang. I also work with the team at Physis (逆矩阵科技), where I help lead our evaluation efforts.

My research focuses on scalable oversight and world model evaluation. I am interested in how we can evaluate increasingly capable systems when direct human assessment is difficult, and how we can rigorously assess models of the physical world.

  • Scalable oversight. I study how to extend reliable supervision beyond what humans can easily assess, with an emphasis on structured feedback, verifiable evaluation, and oversight of complex agent behavior.
  • World model evaluation. I focus on physical consistency and long-horizon consistency: whether predicted dynamics respect physical constraints, and whether states, interactions, and action consequences remain coherent over extended rollouts. I am interested in evaluation protocols that make these capabilities and failure modes measurable.

Previously, I worked on safe reinforcement learning, including OmniSafe and Safety-Gymnasium.

I welcome research discussions and potential collaborations on these topics. Feel free to get in touch!

 

Google Scholar | GitHub

News

Awards

Publications

Services