Minghuan Liu with Thomas Tuchel
Tuchel and Me!

Minghuan Liu

Postdoc

RPL Lab, UT Austin

minghuan97 [at] gmail [dot] com

GScholar GitHub LinkedIn X CV

I am currently a Postdoc at the RPL Lab, UT Austin, working with Prof. Yuke Zhu. Before that I worked as a research scientist at ByteDance Seed-Robotics. I received my Ph.D. in 2024 from Apex Lab at Shanghai Jiao Tong University, advised by Prof. Weinan Zhang and Prof. Yong Yu. During my Ph.D., I also spent time in UCSD (24-25, San Diego, US), ByteDance (23-24, Beijing, China) and Sea AI Lab (22-23, Singapore). I received my bachelor's degree from Southwest Jiaotong University in 2019.

My research centers on general decision modeling, especially on robotics. My expertise lies in RL, imitation learning, robot foundation models and whole-body control. Lately I have been thinking about how to align action with the external 3D world, instead of only replaying trajectories from generalized observation representations.

I also write paper notes in Chinese on Zhihu and actively build agent-based tooling such as AgentMonitor.

  1. Apr. 30, 2026

    In-depth analysis on sim-real co-training, continual VLA learning, and zero-shot cross-humanoid control XHugWBC were accepted by ICML 2026.

  2. Jan. 31, 2026

    Cross-humanoid locomotion H-zero and real2sim2real Re3Sim were accepted by ICRA 2026.

  3. Sept. 4, 2025

    ManipAsInSim was released.

  4. Apr. 11, 2025

    HugWBC was accepted by RSS 2025.

  5. Sept. 26, 2024

    MADiff was accepted by NeurIPS 2024.

  6. Sept. 6, 2024

    Three papers, VBC, ACE, and GenSim2, were accepted by CoRL 2024.