Portrait of Xiaoyang Cao

Xiaoyang Cao 曹骁扬

M.S. student at MIT, in Computational Science & Engineering and Technology & Policy

Cambridge, MA

Hi! I’m a master’s student at MIT, advised by Michiel Bakker. Before MIT, I received my B.S. in Mathematics and Physics from Tsinghua University in 2025.

My research draws on ideas from mathematical logic, cognitive science, and statistical physics to study how intelligence arises. I’m currently exploring this through self-referential AI and multi-agent systems.

Feel free to reach out at xycao [at] mit [dot] edu.

Research Interests

Self-Referential AI. Hofstadter’s strange loops argue that a mind is what happens when a system models itself. Schmidhuber’s Gödel Machine gave this intuition a computational form. I see the same principle at work in self-reward, self-play, self-distillation, and recursive self-improvement. They look like separate techniques. I think they share a root: a model that reasons over its own outputs to reshape itself.

Multi-Agent Emergence. My starting point is Anderson’s “more is different”: at scale, new phenomena emerge that you can’t predict from the parts. Wolfram’s A New Kind of Science makes the same case computationally. Simple rules, iterated, produce genuinely complex behavior. I want to study this in multi-agent AI, where agent swarms, social intelligence, and game theory give rise to collective phenomena that no single agent could produce.

Publications

  1. RE-PO: Robust Enhanced Policy Optimization as a General Framework for LLM Alignment

    Xiaoyang Cao, Zelai Xu, Mo Guang, Kaiwen Long, Michiel Bakker, Yu Wang, Chao Yu

    ICLR 2026 arXiv Website

  2. Pareto Control Barrier Function for Inner Safe Set Maximization Under Input Constraints

    Xiaoyang Cao, Zhe Fu, Alexandre M. Bayen

    ACC 2025 arXiv Slides

  3. Virtual Nodes Improve Long-term Traffic Prediction

    Xiaoyang Cao, Dingyi Zhuang, Shenhao Wang, Jinhua Zhao

    TRB Annual Meeting 2025 arXiv Slides

News

RE-PO, our framework for robust LLM alignment, was accepted at ICLR 2026.
Started my M.S. at MIT.
Admitted to MIT for Fall 2025.
Our paper on Pareto Control Barrier Functions was accepted at ACC 2025.
Our paper on virtual nodes for traffic prediction was accepted at TRB 2025.

Education

  • 2025–2027
    Massachusetts Institute of Technology Dual M.S., Computational Science & Engineering and Technology & Policy · expected May 2027
  • 2024
    University of California, Berkeley Exchange student, Jan–Aug 2024 · GPA 3.93 / 4.00
  • 2021–2025
    Tsinghua University B.S. in Mathematics and Physics · GPA 3.95 / 4.00, ranked 3 / 60