Xiangxin Zhao

Xiangxin Zhao

PhD Candidate (CS) @ UCL, working on LLM coding agents. Supervised by Dr. He Ye & Prof. Federica Sarro.

About

Hi, I’m Xiangxin (Sean) Zhao 👋

I am a PhD Candidate in Computer Science at University College London,
supervised by Dr. He Ye and Prof. Federica Sarro.

My research is on the reliability of LLM coding agents. I study how and why coding agents fail, and whether a failed agent execution can be repaired rather than restarted — extending Automated Program Repair from programs to agent trajectories. I also work on evaluating whether AI-generated code reviews can be trusted.

Research interests: LLM coding agents · agent failure analysis · automated agent repair · automated program repair · code review

News

  • 2026 — My paper From Automated Program Repair to Automated Agent Repair: Why Failed Trajectories Matter was accepted to the ASE 2026 Student Research Competition (Munich, Oct 12–16).
  • Jul 2026 — Preprint Failure as a Process: An Anatomy of CLI Coding Agent Trajectories is on arXiv.

Selected Publications

  • From Automated Program Repair to Automated Agent Repair: Why Failed Trajectories Matter
    Xiangxin Zhao
    ASE 2026 Student Research Competition (SRC) · Accepted · [pdf] [doi]

  • Failure as a Process: An Anatomy of CLI Coding Agent Trajectories
    Xiangxin Zhao, Han Li, Shuaiting Li, Tianyi Zhao, Earl T. Barr, Federica Sarro, He Ye
    arXiv:2607.09510 · Under review · [paper]

  • CRJudgeBench: Can AI Detect Plausible but Invalid Code Reviews?
    Yue Pan, Jiawei Li, Ziyuan Zhang, Xiangxin Zhao, He Ye
    Under review

See all on the Publications page.

Background

  • PhD in Computer Science, University College London
  • MSc in Environmental Data Science & Machine Learning, Imperial College London
  • BSc in Mathematics, University College London

Contact