Shuchen Xue

scxue.jpg

Shuchen Xue is a Research Scientist in the Efficient AI team at NVIDIA Research, working with Dr. Enze Xie and Prof. Song Han. His research lies at the intersection of generative model theory, efficient training and inference, and reinforcement learning for generative models.

He received his Ph.D. in Statistics from the University of Chinese Academy of Sciences (UCAS) in 2026, with research conducted at the Academy of Mathematics and Systems Science (AMSS), Chinese Academy of Sciences (CAS), advised by Prof. Zhi-Ming Ma. He received his B.S. degree in Mathematics from UCAS in 2021.

Before joining NVIDIA, he was a research intern at ByteDance Seed, working with Tao Yang and Yuxin Fang; at Adobe Research, working with Chongjian Ge; and at Huawei Noah’s Ark Lab, working with Mingyang Yi, Tianyang Hu, and Zhaoqiang Liu. He was also a visiting student at Columbia University.

news

May 25, 2026 Started as a Research Scientist at NVIDIA Research.
May 01, 2026 Three papers were accepted by ICML 2026. One paper was selected as an Oral Presentation.
Jan 26, 2026 Three papers were accepted by ICLR 2026.
Nov 10, 2025 Thrilled to begin my internship at Bytedance Seed.
Sep 29, 2025 Advantage Weighted Matching, a principled RL method for diffusion and flow models, is released. Check out our paper and code.
Jul 23, 2025 One paper was selected as Highlight by ICCV 2025.
Jul 07, 2025 One paper was selected as Oral Presentation by ICML 2025 (ES-FoMo workshop) (Top 3.42%).
Jun 25, 2025 Two papers were accepted by ICCV 2025.
Jun 16, 2025 Thrilled to begin my internship at Adobe Research.
May 01, 2025 One paper was accepted by ICML 2025.
Jan 22, 2025 One paper was accepted by ICLR 2025.
May 01, 2024 One paper was accepted by ICML 2024.
Feb 26, 2024 One paper was accepted by CVPR 2024.
Sep 21, 2023 One paper was accepted by NeurIPS 2023.

selected publications

  1. ICML
    Advantage Weighted Matching: Aligning RL with Pretraining in Diffusion Models
    Shuchen Xue, Chongjian Ge, Shilong Zhang, and 2 more authors
    International Conference on Machine Learning, 2026
    arXiv Code
  2. ICMLOral
    Any-Order GPT as Masked Diffusion Model: Decoupling Formulation and Architecture
    Shuchen Xue, Tianyu Xie, Tianyang Hu, and 5 more authors
    International Conference on Machine Learning (Oral Presentation, Top 0.70%), 2026
    Abridged in the ICML 2025 ES-FoMo workshop Oral Presentation (Top 3.42%)
    arXiv Code
  3. ICLR
    Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding
    Chengyue Wu*, Hao Zhang*, Shuchen Xue, and 6 more authors
    International Conference on Learning Representations, 2026
    arXiv Code Website
  4. ICCVHighlight
    SANA-Sprint: One-Step Diffusion with Continuous-Time Consistency Distillation
    Junsong Chen*, Shuchen Xue*, Yuyang Zhao, and 6 more authors
    IEEE/CVF International Conference on Computer Vision (Highlight), 2025
    arXiv Code Website
  5. CVPR
    Accelerating Diffusion Sampling with Optimized Time Steps
    Shuchen Xue, Zhaoqiang Liu, Fei Chen, and 4 more authors
    IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024
    arXiv Code
  6. NeurIPS
    SA-Solver: Stochastic Adams Solver for Fast Sampling of Diffusion Models
    Shuchen Xue, Mingyang Yi, Weijian Luo, and 4 more authors
    Advances in Neural Information Processing Systems, 2023
    arXiv Code
 

    Total site views:  |      Site visitors: