Yujun Zhou

AI Research · University of Notre Dame

Yujun Zhou 周钰钧

Ph.D. student in Computer Science Graduating Spring 2027

Yujun Zhou
South Bend, Indiana

My research includes LLM reasoning and reinforcement learning, long-horizon coding agents, and reliable agent behavior and evaluation.

I am advised by Xiangliang Zhang at the University of Notre Dame. Previously, I completed research internships at Microsoft and Tencent AI Lab.

CV EmailGoogle ScholarGitHubLinkedIn

Selected Research

More publications

Learning to reason. Sustaining complex work. Making agent behavior reliable.

Alignment RiskICML 2026

Alignment Risks from Capability-Seeking RL Training

Investigates how capability-seeking RL can teach models to exploit environmental vulnerabilities, even as standard performance improves.

Yujun Zhou*, Yue Huang*, Han Bao*, Kehan Guo, Zhenwen Liang, Pin-Yu Chen, Tian Gao, Werner Geyer, Nuno Moniz, Nitesh V Chawla, Xiangliang Zhang

Paper Code
Experimental framework for studying capability-seeking reinforcement learning and alignment risks, with reward tampering and self-grading examples.
Auditing the incentives behind capability gains.View figure
LabSafety BenchNature Machine Intelligence

LabSafety Bench: Benchmarking LLMs on Safety Issues in Scientific Labs

A benchmark for evaluating how reliably language models recognize and reason about safety issues in scientific laboratories.

Yujun Zhou, Jingdong Yang, Yue Huang, Kehan Guo, Zoe Emory, Bikram Ghosh, Amita Bedar, Sujay Shekar, Pin-Yu Chen, Tian Gao, Werner Geyer, Nuno Moniz, Nitesh V Chawla, Xiangliang Zhang

Paper Code Dataset
LabSafety Bench overview of laboratory safety scenarios and benchmark evaluation tasks.
Evaluating model behavior in safety-critical settings.View figure

* Equal contribution.

Research Experience

Microsoft

Research Intern · May–Aug 2026

Trustworthy LLM evaluation and realistic benchmarks for agents in complex enterprise scenarios.

Tencent AI Lab

Research Intern · May–Dec 2025

LLM reasoning and reinforcement learning, including label-free post-training and reasoning diversity.

News

  1. Two papers have been accepted by ICML 2026, including one first-author paper!
  2. Four papers have been accepted by ACL 2026, including two main conference papers!
  3. Two papers have been accepted by ICLR 2026!
  4. One first-author paper has been accepted by Nature Machine Intelligence!
  5. Two papers have been accepted by Neurips 2025!
Earlier updates (13)
  1. One first-author paper has been accepted by EMNLP Findings 2025!
  2. I’m excited to share that I have extended my internship at Tencent AI Lab！
  3. I joined Tencent AI Lab as a research intern this summer! See you in Seattle！
  4. One Paper is accepted by ACL Findings 2025!
  5. One Paper is accepted by IJCAI 2025 Survey Track!
  6. Thrilled to be awarded OpenAI’s Researcher Access Program.
  7. One first-author paper has been accepted by EMNLP 2024!
  8. One paper has been accepted by NeurIPS 2024 Dataset and Benchmark Track as a spotlight!
  9. One Paper is accepted by ACL 2024!
  10. One first-author paper has been accepted by ICML 2024!
  11. One Paper is accepted by AAAI 2023!
  12. One Paper is accepted by BigData 2022!
  13. One paper is accepted by ICLR 2022!

More Publications

Google Scholar

* Equal contribution. Full author order is retained for every publication.

Academic Background

Education

University of Notre Dame
Ph.D. in Computer Science 2023–Spring 2027 (expected)
King Abdullah University of Science and Technology
M.S. in Computer Science 2021–2022
University of Electronic Science and Technology of China
B.Eng. in Computer Science and Engineering 2017–2021

Research community

At Notre Dame, I am part of MINE Lab and the Foundation Models and Applications Lab at the Lucy Family Institute for Data & Society.

I also work on the ND–IBM Technology Ethics Lab collaborative project, studying trustworthy LLMs in practical, safety-critical settings.

Research figure