Juntong Shi

prismatic.jpg

photo: Prismatic^2

Hi, I’m Juntong “Steven” Shi. I’m a master student at Stanford Computer Science, working with Stefano Ermon. Previously, I did undergraduate at University of Southern California (USC), studying computer science and applied mathematics. I develop algorithms that make generative models, particularly language models, more efficient and capable through improved inference and post-training.

news

Apr 30, 2026 Diffusion Language Model Parallel Decoding via Product-of-Experts Bridge (PoE-Bridge) was accepted at ICML 2026!
Jan 22, 2025 TabDiff: a Mixed-typed Diffusion Model for Tabular Data Generation was accepted at ICLR 2025. See you in Singapore!
Nov 02, 2024 :rocket: Launched my personal website!

selected works

  1. ICML’26
    poe-bridge_demo.png
    Diffusion Language Model Parallel Decoding via Product-of-Experts Bridge
    Juntong Shi, Brian L. Trippe, Jure Leskovec, Stefano Ermon, and Minkai Xu
    In Forty-third International Conference on Machine Learning, 2026
    PDF Code
  2. CVPR’26
    spectrum_illustration.png
    Adaptive Spectral Feature Forecasting for Diffusion Sampling Acceleration
    Jiaqi Han*, Juntong Shi*, Puheng Li, Haotian Ye, Qiushan Guo, and Stefano Ermon
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026
    PDF Code Website
  3. ICLR’25
    tabdiff_demo.gif
    TabDiff: a Mixed-typed Diffusion Model for Tabular Data Generation
    Juntong Shi*, Minkai Xu*, Harper Hua*, Hengrui Zhang*, Stefano Ermon, and Jure Leskovec
    In The Thirteenth International Conference on Learning Representations, 2025
    PDF Video Code
  4. ICML’25
    graphbfn_mainfig.png
    Smooth Interpolation for Improved Discrete Graph Generative Models
    Yuxuan Song*, Juntong Shi*, Jingjing Gong*, Minkai Xu, Stefano Ermon, Hao Zhou, and Wei-ying Ma
    In Forty-second International Conference on Machine Learning, 2025
    PDF Code