Zeyu Gan's CV

Basic Information

• Name: Zeyu Gan (甘泽宇)
• Gender: Male
• Date of Birth: June 12, 1999

Education

• B.S. Computer Science, School of Information, Renmin University of China, Sep 2017 - June 2021
• M.S. Computer Application Technology, School of Information, Renmin University of China, Sep 2021 - June 2024
• PhD Artificial Intelligence, Gaoling School of Artificial Intelligence, Renmin University of China, Sep 2024 - June 2028 (Expected)

Research Experience

• Gaoling School of Artificial Intelligence
  • Renmin University of China, Sep 2024 - now
  • As a PhD student, my research interests mainly include LLM alignment and machine learning theory. During this period, I participate in plenty of research activities and focus on realistic problems toward better AI modeling.
• DW&BI ML Group in School of Information
  • Renmin University of China, Sep 2021 - June 2024
  • As a member of the laboratory, my research interests mainly include machine learning, computer vision, and weakly supervised learning. During this period, I completed numbers of scientific research tasks.

Working Experience

• Machine Learning Platform Department
  • TEG, Tencent, Aug 2023 - Nov 2023
  • As an intern, I participated in the development of the WeChat Moments advertising recommendation algorithms. I researched the deployment in big batch scenarios and helped design suitable optimizers.

Honors & Awards

• Outstanding Master Theis (10%), June, 2024
• Outstanding Graduates (10%), April, 2024
• Outstanding Postgraduate Scholarship (First Class, 10%), Nov, 2023
• College Student Innovation Experiment Program (National Project), May 2019

Publications

Superclass Learning with Representation Enhancement

Zeyu Gan, Suyun Zhao, Jinlong Kang, Liyuan Shang, Hong Chen, Cuiping Li. Superclass Learning With Representation Enhancement. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 24060-24069, 2023.

Rethinking External Slow-Thinking: From Snowball Errors to Probability of Correct Reasoning

Zeyu Gan, Yun Liao, Yong Liu. Rethinking External Slow-Thinking: From Snowball Errors to Probability of Correct Reasoning. arXiv preprint arXiv:2501.15602, 2025

Towards a Theoretical Understanding of Synthetic Data in LLM Post-Training: A Reverse-Bottleneck Perspective

Zeyu Gan, Yong Liu. Towards a Theoretical Understanding of Synthetic Data in LLM Post-Training: A Reverse-Bottleneck Perspective. In The Thirteenth International Conference on Learning Representations, 2025.

Service and Leadership

• Student Union of Renmin University of China
  • As an active member of the student union’s publicity department, I took part in and led multiple campus initiatives, which included promoting campus photography themes and choosing topics for student union campaigns.
  • I contributed significantly to the organization, coordination, and documentation of various events, such as the May 4th celebration and Freshman Orientation, among others.
• Cyber News Service of Renmin University of China
  • Cyber News Service is affiliated with the school news center of the student organization, responsible for related activities of design and publicity.
  • As a news agency member, I participated in or organized many topics and activities, including National Day 70th anniversary-related activities, graduation photography exhibition, etc.