Yongjin Han

I’m an independent researcher looking for PhD positions starting in Fall 2027. I currently work with Dr. Mincheol Park. My researches focus on safe and controllable AI systems:

  • Formal methods for multi-agent systems: enforcing safety specifications on agents so that they behave as intended.
  • Code world models: representing how the world works as executable code.

I received my M.S. from UC Davis and my B.S. from Dongguk University in South Korea. In my free time, I usually cook! or watch movies.

Email  /  CV  /  LinkedIn

profile photo of Yongjin Han

Research

We are delegating more authority to AI in safety-critical systems. Yet what counts as “safe” depends on context, is contested across communities, and can never be fully specified in advance. We will sometimes get it wrong. We must therefore at least preserve our ability to check what AI systems actually do and to correct them when they fail. My research is a step in that direction.

  1. arXiv
    Lipschitz-aware Linearity Grafting for Certified Robustness
    Yongjin Han, Suhyun Kim
    arXiv, 2025
    arXiv