Xiang Zhu

I am a Ph.D. candidate (successive postgraduate and doctoral programs) at the Institute for Interdisciplinary Information Sciences, Tsinghua University, advised by Prof. Jianyu Chen. I expect to graduate in June 2027. Prior to that, I received my Bachelor's degree from the School of Automation, Beijing University of Posts and Telecommunications.

Email  /  GitHub  /  Google Scholar  

Research

My research focuses on embodied AI and robot learning, with a particular interest in leveraging human data to improve robot manipulation. Recently, I have been exploring Vision-Language-Action (VLA) models, World-Action Models (WAMs), and cross-embodiment learning to enable more generalizable robot policies through human-robot knowledge transfer. I am also interested in dexterous manipulation, humanoid robotics, reinforcement learning, and sim-to-real transfer.

(*: Equal contribution. Co-first Author.)

GALA: Geometry-Aware Latent Action Modeling for Vision-Language-Action Model Pretraining across Embodiments

HARP-VLA: Human-Robot Aligned Representation Learning for Vision-Language-Action Model

Learning Generalizable Robot Policy with Human Demonstration Video as a Prompt

UP-VLA: A Unified Understanding and Prediction Model for Embodied Agent

Advancing humanoid locomotion: Mastering challenging terrains with denoising world model learning

Best Paper Award Finalists.

A Contact-Safe Reinforcement Learning Framework for Contact-Rich Robot Manipulation

Stylized table tennis robots skill learning with incomplete human demonstrations

Design and source code from Jon Barron's website