About
I am a PhD student in the Department of Civil Engineering at the University of Hong Kong (HKU), advised by Prof. Xintao Yan. I am also a long-term research intern at the Institute for AI Research (AIR), Tsinghua University, supervised by Prof. Xianyuan Zhan. Previously, I obtained my BSc in Computer Science with First Class Honours from the Chinese University of Hong Kong (CUHK).
My research focuses on advancing autonomous systems, with particular emphasis on safety and flexible decision-making. I am also interested in developing robust and generalizable reinforcement learning algorithms to support broader industry adoption. Recently, my attention has focused on designing horizon-adaptive reinforcement learning frameworks to improve value learning and support long-horizon planning.
I dream of a small, focused team dedicated to foundational work, with a commitment to craftsmanship over hype. Welcome to follow my LOVE TEAM, DiffusionAD!
Selected Publications
View All →Dichotomous Diffusion Policy Optimization
Ruiming Liang†, Yinan Zheng†, Kexin Zheng†, Tianyi Tan†, Jianxiong Li, Liyuan Mao, Zhihao Wang, Guang Chen, Hangjun Ye, Jingjing Liu, Jinqiao Wang, Xianyuan Zhan
The Fourteenth International Conference on Learning Representations (ICLR)
A reinforcement learning algorithm designed for stable and controllable optimization of diffusion-based policies.
Towards Robust Zero-Shot Reinforcement Learning
Kexin Zheng†, Lauriane Teyssier†, Yinan Zheng, Yu Luo, Xianyuan Zhan
The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS)
An upgraded Forward-Backward representations (FB)-based zero-shot reinforcement learning framework that simultaneously enhances learning stability, policy extraction capability, and representation learning quality.
Diffusion-Based Planning for Autonomous Driving with Flexible Guidance
Yinan Zheng†, Ruiming Liang†, Kexin Zheng†, Jinliang Zheng, Liyuan Mao, Jianxiong Li, Weihao Gu, Rui Ai, Shengbo Eben Li, Xianyuan Zhan, Jingjing Liu
The Thirteenth International Conference on Learning Representations (ICLR)
A novel transformer-based Diffusion Planner for autonomous driving, which can effectively model multi-modal driving behavior and ensure trajectory quality without any rule-based refinement.
News
I was awarded the HKU Presidential PhD Scholarship (HKU-PS) !!!
Our work DIPOLE has been accepted by ICLR 2026 🎉