Zheyuan Hu

Robotics InstituteCarnegie Mellon UniversityPittsburgh, PA

Portrait of Zheyuan Hu

I am a second-year Ph.D. student at the CMU Robotics Institute, advised by Prof. Aviral Kumar and Prof. Zackory Erickson. I received my Bachelor's degrees in Computer Science and Applied Mathematics from UC Berkeley, where I was advised by Prof. Sergey Levine. Previously, I had the privilege of working with Prof. Abhishek Gupta and Dr. Kelvin Xu.

My research focuses on developing the intelligence needed to solve general robotic manipulation problems in the real world. This includes scalable systems and algorithms that learn from diverse data, make use of test-time compute, and autonomously acquire new skills through online interaction.

Publications

6

* equal contribution

  1. IEEE T-RO · 2026

    RaC: Robot Learning for Long-Horizon Tasks by Scaling Recovery and Correction

    Zheyuan Hu, Robyn Wu, Naveen Enock, Jasmine Li, Riya Kadakia, Zackory Erickson, Aviral Kumar

    RaC scales up recovery and correction data via a simple yet effective human-in-the-loop data collection protocol, enabling robots to spend more rollout budget mitigating compounding errors in complex long-horizon tasks.

  2. IEEE/RSJ IROS · 2025 · Oral

    SimLauncher: Launching Sample-Efficient Real-World Robotic Reinforcement Learning via Simulation Pre-Training

    Mingdong Wu*, Lehong Wu*, Yizhuo Wu*, Weiyao Huang, Hongwei Fan, Zheyuan Hu, Haoran Geng, Jinzhou Li, Jiahe Ying, Long Yang, Yuanpei Chen, Hao Dong

    SimLauncher pre-trains a visuomotor policy in a digital-twin simulation and uses it to launch real-world RL: simulated and real demonstrations bootstrap value targets, and the pre-trained policy proposes actions for better exploration, substantially improving sample efficiency and reaching near-perfect success rates.

  3. Robotics: Science and Systems · 2024

    Yell At Your Robot: Improving On‑the‑Fly from Language Corrections

    Lucy Xiaoyang Shi, Zheyuan Hu, Tony Z. Zhao, Archit Sharma, Karl Pertsch, Jianlan Luo, Sergey Levine, Chelsea Finn

    YAY Robot leverages verbal corrections to enable on-the-fly adaptation and continuous policy improvement on complex long-horizon tasks.

  4. IEEE ICRA · 2024

    SERL: A Software Suite for Sample-Efficient Robotic Reinforcement Learning

    Jianlan Luo*, Zheyuan Hu*, Charles Xu, Siri Gadipudi, Archit Sharma, Rehaan Ahmad, Stefan Schaal, Chelsea Finn, Abhishek Gupta, Sergey Levine

    SERL (Sample-Efficient Robotic reinforcement Learning) is an open-source software framework that aims to facilitate wider adoption of RL in real-world robotics.

  5. Conference on Robot Learning · 2023

    REBOOT: Reuse Data for Bootstrapping Efficient Real-World Dexterous Manipulation

    Zheyuan Hu*, Aaron Rovinsky*, Jianlan Luo, Vikash Kumar, Abhishek Gupta, Sergey Levine

    REBOOT learns dexterous manipulation skills autonomously and entirely in the real world in under 8 hours by bootstrapping from prior data, a 2× speed-up over learning from scratch. Tested on a multi-fingered robot hand learning in-hand rotation.

  6. IEEE ICRA · 2023

    Dexterous Manipulation from Images: Autonomous Real-World RL via Substep Guidance

    Kelvin Xu*, Zheyuan Hu*, Ria Doshi, Aaron Rovinsky, Vikash Kumar, Abhishek Gupta, Sergey Levine

    AVAIL lets robots learn long-horizon manipulation skills through autonomous real-world interaction without manual engineering: users break a long task into sub-tasks with image examples, and deep RL does the rest. Tested on a four-finger robot hand on complex dexterous tasks in the real world.

    Teaser for Dexterous Manipulation from Images: Autonomous Real-World RL via Substep Guidance