Zhouliang Yu (郁昼亮)
PhD Student, Chinese University of Hong Kong
Large Language Models & Agentic Reasoning
Reinforcement Learning & Formal Mathematics
Embodied AI & World Models
Email: zhouliangyu at link dot cuhk dot edu dot cn
[Email]
[CV]
[Google Scholar]
[GitHub]
[Twitter]
I'm Zhouliang Yu (郁昼亮), a PhD student at the Scalable Principles for Learning and Reasoning Lab (SphereLab) in the Chinese University of Hong Kong, Department of Computer Science & Engineering, advised by Prof. Weiyang Liu. My research focuses on large language models, deep learning, reinforcement learning, and formal reasoning.
My primary research (2024–2027) centers on exploration-based reinforcement learning for formal mathematics reasoning using agentic large language models. I am also actively learning RL infrastructure to support large model training.
Beyond my core focus, I am interested in applications of reinforcement learning in model-based embodied AI and scientific discovery through formal verification (I have not yet published in these areas), such as projects like Scientist AI and PhysLean.
Previously, I pursued doctoral studies at HKUST under Academician Yike Guo. I have also conducted research at Alibaba's Tongyi Lab. Earlier, I received my bachelor's degree in computer science from CUHK-SZ.
郁昼亮，香港中文大学计算机科学与工程学系博士生，在 SphereLab 师从 刘威杨 教授。读博期间，我主要研究以可扩展的学习与推理原理为驱动的大模型训练算法，面向形式化推理及更广义的科学发现场景，提升探索式求解与发现能力。现阶段，我的工作主要聚焦于大模型在形式化数学推理上的训练算法。此前，我曾在香港科技大学攻读博士学位，师从 郭毅可 院士。我亦曾在阿里巴巴通义实验室从事研究工作；更早以前，我于 香港中文大学（深圳） 取得计算机专业学士学位。