Qi Yan

prof_pic.jpg

My name is Qi Yan (颜齐). I am a Senior Researcher at Waabi, where I work on world models and controllable data generation to advance physical intelligence for autonomous driving.

Previously, I was a Ph.D. student in ECE at UBC, co-supervised by Prof. Renjie Liao and Prof. Lele Wang. During my Ph.D., I explored generative modeling and motion forecasting, particularly diffusion and flow matching methods for trajectories, graphs, and visual scenes. I earned a master’s in Mechanical Engineering from EPFL (2022) and a bachelor’s in engineering (honors) from Shanghai Jiao Tong University (2019).

News

Sep 2026 I passed my final PhD oral defence! Thank you to all my committee members and collaborators for your support. (Photo)
Feb 2026 Our paper BQ-SRC was accepted to CVPR 2026!
Jan 2026 Our paper StreamSplat was accepted to ICLR 2026!
Sep 2025 Our paper Neural MJD and RETRO SYNFLOW were accepted to NeurIPS 2025!
Aug 2025 Updated two blog posts on permutation invariant graph generation and conditional generative models for motion prediction.

Recent Publications [full list]

  1. arXiv
    Illustration for Consistent Distribution Matching for Data-Free Diffusion Distillation
    Consistent Distribution Matching for Data-Free Diffusion Distillation
    Yuxiang Fu, Qi Yan, Zike Wu, Yongxing Zhang, Purang Abolmaesumi, Lele Wang, and Renjie Liao
    arXiv preprint arXiv:2610.09221, 2026.
    arXiv Code Website
  2. CVPR
    Illustration for Spectral Conformal Risk Control: Distribution-Free Tail Guarantees via Bayesian Quadrature
    Spectral Conformal Risk Control: Distribution-Free Tail Guarantees via Bayesian Quadrature
    Mohammad Mahdi Kazemi Esfeh, Qi Yan, Yongxing Zhang, Zahra Gholami, Renjie Liao, and Purang Abolmaesumi
    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026.
    PDF
  3. ICLR
    Illustration for StreamSplat: Towards Online Dynamic 3D Reconstruction from Uncalibrated Video Streams
    StreamSplat: Towards Online Dynamic 3D Reconstruction from Uncalibrated Video Streams
    Zike Wu, Qi Yan, Xuanyu Yi, Lele Wang, and Renjie Liao
    International Conference on Learning Representations, 2026.
    arXiv Code Website
  4. NeurIPS
    Illustration for RETRO SYNFLOW: Discrete Flow Matching for Accurate and Diverse Single-Step Retrosynthesis
    RETRO SYNFLOW: Discrete Flow Matching for Accurate and Diverse Single-Step Retrosynthesis
    Robin Yadav, Qi Yan, Guy Wolf, Avishek Joey Bose, and Renjie Liao
    NeurIPS, 2025.
    arXiv Code
  5. NeurIPS
    Illustration for Neural MJD: Neural Non-Stationary Merton Jump Diffusion for Time Series Prediction
    Neural MJD: Neural Non-Stationary Merton Jump Diffusion for Time Series Prediction
    Yuanpei Gao, Qi Yan, Yan Leng, and Renjie Liao
    NeurIPS, 2025.
    arXiv Code

Visitors