PeopleofSuperintelligence

Researcher

Yilun Du

Yilun Du is an assistant professor of computer science at Harvard and an investigator at the Kempner Institute. He leads the Embodied Minds Lab, which studies generative models of the world. Techniques for scaling energy-based model training appeared in 2019, in a paper by Du and Igor Mordatch. In 2020, an energy-based model of protein conformations that Du co-wrote reached performance close to the Rosetta energy function. The paper Learning Interactive Real-World Simulators, which Du co-wrote, received an ICLR 2024 outstanding paper award.

Works
184
Citations
3,977
h-index
16

Publication numbers from OpenAlex, as of 15 October 2026.

Timeline

  1. March 2019

    Scaling energy-based models

    Implicit Generation and Generalization in Energy-Based Models, a paper by Du and Igor Mordatch, presents techniques to scale energy-based model training. It appears on arXiv in March 2019.

  2. April 2020

    An energy model for proteins

    Energy-based models for atomic-resolution protein conformations, co-written by Du with Joshua Meier and Rob Fergus, is posted on arXiv in April 2020. The model achieves performance close to that of the Rosetta energy function.

  3. May 2022

    Planning by denoising trajectories

    Planning with Diffusion for Flexible Behavior Synthesis, written by Du with three co-authors, appears on arXiv in May 2022. The core of the paper is a diffusion probabilistic model that plans by iteratively denoising trajectories.

  4. January 2023

    Video generation as a policy

    Learning Universal Policies via Text-Guided Video Generation, written by Du and seven co-authors, is posted on arXiv in January 2023. The paper casts sequential decision making as text-conditioned video generation.

  5. 2024

    An ICLR outstanding paper

    In 2024, ICLR names Learning Interactive Real-World Simulators an outstanding paper, and Du is one of its seven authors. The paper introduces UniSim, a universal simulator of real-world interaction learned through generative modeling.

  6. April 2024

    A final-year PhD at MIT

    In April 2024, a final-year PhD student at MIT CSAIL, Du is supported by an NSF Graduate Research Fellowship. Earlier roles include research fellow at OpenAI, visiting researcher at FAIR, and student researcher at Google DeepMind.

  7. July 2026

    Diffusion-MPC for legged robots

    In July 2026, Diffusion-MPC is a robotic control system from Du and colleagues that aims to combine the best aspects of existing methods.

  8. As of October 2026

    Assistant professor at Harvard

    He is an assistant professor of computer science at Harvard and an investigator at the Kempner Institute, and he leads the Embodied Minds Lab, which studies generative models of the world. He is also a part-time research scientist at NVIDIA Research.

Sources and credits

Photo: We have no published photo of Yilun Du yet. Send us a free-licensed photo.

  1. [1903.08689] Implicit Generation and Generalization in Energy-Based Models arxiv.org. Checked 15 October 2026.
  2. [2004.13167] Energy-based models for atomic-resolution protein conformations arxiv.org. Checked 15 October 2026.
  3. [2205.09991] Planning with Diffusion for Flexible Behavior Synthesis arxiv.org. Checked 15 October 2026.
  4. [2302.00111] Learning Universal Policies via Text-Guided Video Generation arxiv.org. Checked 15 October 2026.
  5. ICLR 2024 Awards iclr.cc. Checked 15 October 2026.
  6. Princeton Engineering cs.princeton.edu. Checked 15 October 2026.
  7. An AI-Powered Control System For Robots With Legs seas.harvard.edu. Checked 15 October 2026.
  8. Yilun Du - Kempner Institute kempnerinstitute.harvard.edu. Checked 15 October 2026.
  9. Yilun Du | Learning & Inference for Generalizable Intelligence yilundu.github.io. Checked 15 October 2026.
  10. Embodied Minds Lab at Harvard embodied-minds-lab.github.io. Checked 15 October 2026.