Yihuai Gao

personal_photo2.png

Robotics & Embodiment AI Lab (REAL) @ Stanford University

Hi! I am Yihuai Gao (高义淮), a third-year Ph.D. student in EE Department, Stanford University. I’m fortunate to be advised by Prof. Shuran Song in the field of robot learning. I’m generally interested in robot manipulation policies and systems, while currently exploring video generation for robots.

Prior to Stanford, I obtained a dual bachelor’s degree (B.Eng in EE and B.S. in Math) from Tsinghua University in 2023, where I worked with Prof. Bowen Zhou and Prof. Heng Yang (at Harvard). I also had a great summer working with Prof. Dirk Englund at MIT in 2022.

Contact: yihuai at stanford dot edu

News

Jun 01, 2026 Cosmos 3, NVIDIA’s omnimodal world model for Physical AI, is released! Check out the report and code.
Apr 21, 2026 Our new paper Gated Memory Policy is released on arXiv!
Jan 22, 2026 Paper Cosmos Policy is accepted to ICLR 2026!
Apr 16, 2025 Paper UVA is accepted to RSS 2025!
Sep 06, 2024 Paper UMI-on-Legs is accepted to CoRL 2024!
May 13, 2024 Paper CLOSURE is accepted to RSS 2024!

Selected Publications

  1. Gated Memory Policy
    Yihuai Gao, Jinyun Liu, Shuang Li, and Shuran Song
    arXiv preprint arXiv:2604.18933, 2026
  2. Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planning
    In International Conference on Learning Representations (ICLR), 2026
  3. Unified Video Action Model
    Shuang Li, Yihuai Gao, Dorsa Sadigh, and Shuran Song
    In Robotics: Science and Systems (RSS), 2025
  4. UMI on Legs: Making Manipulation Policies Mobile with Manipulation-Centric Whole-body Controllers
    Huy Ha*, Yihuai Gao*, Zipeng Fu, Jie Tan, and Shuran Song
    In Conference on Robot Learning (CoRL), 2024
  5. closure.png
    CLOSURE: Fast Quantification of Pose Uncertainty Sets
    Yihuai Gao*, Yukai Tang*, Han Qi, and Heng Yang
    In Robotics: Science and Systems (RSS), 2024
  6. nature-cmos.png
    Heterogeneous integration of spin–photon interfaces with a CMOS platform
    Nature, 2024