I am a first-year CS Ph.D. student at the School of Data Science, The Chinese University of Hong Kong, Shenzhen (CUHK-Shenzhen), advised by Prof. Guiliang Liu. I received my B.Eng. in Automation from Harbin Institute of Technology, Shenzhen in 2025, ranking in the top 2% of my cohort.
My research focuses on building embodied agents that can learn from interaction, adapt to new situations, and transfer robust behaviors across real-world settings. I am particularly interested in reinforcement learning for robotic control, robot representation learning, and test-time steering/adaptation.
My work has appeared in top-tier AI and robotics venues, including ICML, ICLR, and ICRA.
I have also been fortunate to work as a research intern at Shanghai AI Laboratory, mentored by Ming Zhou.
At a high level, I enjoy working on algorithms and systems that make physical intelligence more general, reliable, and deployable.
News
Selected Publications
HWC-Loco: A Hierarchical Whole-Body Control Approach to Robust Humanoid Locomotion
A hierarchical whole-body control approach for robust humanoid locomotion across diverse terrains, robot structures, and disturbance settings.
RoboFlow4D: A Lightweight Flow World Model Toward Real-Time Flow-Guided Robotic Manipulation
Sixu Lin, Huaiyuan Xu, Junliang Chen, Zhuohao Li, Guangming Wang, Yixiong Jing, Sheng Xu, Runyi Zhao, Brian Sheil, Lap-Pui Chau, Guiliang Liu</p> A 4D flow representation and policy learning framework for robotic manipulation. Sixu Lin, Yunpeng Qing, Litao Liu, Ming Zhou, Ruixing Jin, Xiaoyi Fan, Guiliang Liu</p> A dynamic grouping residual reinforcement learning framework for adaptive embodied policies. A bidirectional diffusion framework for offline reinforcement learning that models both future and history trajectories from intermediate states. SignBot: Learning Human-to-Humanoid Sign Language Interaction A human-to-humanoid sign language interaction framework spanning motion retargeting, motion control, and generative interaction. A Vision-Language-Action-Critic Model for Robotic Real-World Reinforcement Learning A vision-language-action-critic model that provides dense progress rewards and action generation for real-world robot reinforcement learning. SMAP: Self-supervised Motion Adaptation for Physically Plausible Humanoid Whole-body Control A self-supervised motion adaptation framework for physically plausible humanoid whole-body control. MASQ: Multi-Agent Reinforcement Learning for Single Quadruped Robot Locomotion A multi-agent reinforcement learning formulation that treats the legs of a single quadruped robot as cooperating agents. Research internship on embodied intelligence and robotics. Worked on vision-language-action models and reinforcement fine-tuning. Undergraduate research on robotics and reinforcement learning. Remote research collaboration on embodied AI and robotics.
