Laboratory for Deep Structured Learning

The University of British Columbia, Department of Electrical and Computer Engineering

Welcome to the Renjie Liao’s lab for Deep Structure Learning (DSL) at UBC!

Our long-term research goal is to build superintelligent AI agents. We pursue this from both theoretical and practical perspectives, aiming to develop foundations (e.g., long-term memory mechanisms) that could lead to superintelligence.

Our recent research covers:

  • Deep Generative Models
  • Geometric Deep Learning
  • Math/Algorithmic Reasoning with Transformers (e.g., LLMs)
  • Probabilistic Inference
  • Visual Understanding and Reasoning
  • Motion Prediction and Planning in Self-driving
  • AI for Healthcare

news

Sep 18, 2025 Our works Neural MJD: Neural Non-Stationary Merton Jump Diffusion for Time Series Prediction and RETRO SYNFLOW: Discrete Flow Matching for Accurate and Diverse Single-Step Retrosynthesis were accepted in Annual Conference on Neural Information Processing Systems (NeurIPS) 2025. Congratulations to Yuanpei, Qi, and Robin!
Aug 4, 2025 Our work An Information-Theoretic Framework for Out-of-Distribution Generalization with Applications to Stochastic Gradient Langevin Dynamics was accepted for publication in the IEEE Transactions on Information Theory (TIT). Congratulations to Wenliang!
Jul 21, 2025 Our work Learning Latent Part-Whole Hierarchies for Point Clouds was accepted for publication in the International Journal of Computer Vision (IJCV). Congratulations to Xiang!
Jun 15, 2025 Our work TrajFlow: Multi-modal Motion Prediction via Flow Matching was accepted in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) as an Oral Presentation. Congratulations to Qi, Brian, Yutong, Hsueh-Han, and Joshua!
May 1, 2025 Our work Leveraging Online Olympiad-Level Math Problems for LLMs Training and Contamination-Resistant Evaluation was accepted for publication in International Conference on Machine Learning (ICML) 2025. Congratulations to Sadegh, Muchen, and Kevin!
Feb 26, 2025 Our works MoFlow: One-Step Flow Matching for Human Trajectory Forecasting via Implicit Maximum Likelihood Estimation based Distillation and LatentHOI: On the Generalizable Hand Object Motion Generation with Latent Hand Diffusion were accepted in IEEE/CVF Conference on Computer Vision and Pattern Recognition Conference (CVPR) 2025. Congratulations to Felix, Qi, and Muchen!
Feb 11, 2025 Our work SymmetricDiffusers: Learning Discrete Diffusion on Finite Symmetric Groups was accepted in International Conference on Learning Representations (ICLR) 2025 as an Oral Presentation (top 1.8%). Congratulations to Nick and Donglin!

latest posts

selected publications

2025

  1. NeurIPS
    Neural MJD: Neural Non-Stationary Merton Jump Diffusion for Time Series Prediction
    Yuanpei Gao, Qi Yan, Yan Leng, and Renjie Liao
    In Annual Conference on Neural Information Processing Systems (NeurIPS), 2025
  2. NeurIPS
    RETRO SYNFLOW: Discrete Flow Matching for Accurate and Diverse Single-Step Retrosynthesis
    Robin Yadav, Qi Yan, Guy Wolf, Avishek Joey Bose, and Renjie Liao
    In Annual Conference on Neural Information Processing Systems (NeurIPS), 2025
  3. IEEE TIT
    An information-theoretic framework for out-of-distribution generalization with applications to stochastic gradient Langevin dynamics
    Wenliang Liu, Guanding Yu, Lele Wang, and Renjie Liao
    IEEE Transactions on Information Theory (TIT), 2025
  4. IJCV
    Learning latent part-whole hierarchies for point clouds
    Xiang Gao, Wei Hu, and Renjie Liao
    International Journal of Computer Vision (IJCV), 2025
  5. CVPR
    Moflow: One-step flow matching for human trajectory forecasting via implicit maximum likelihood estimation based distillation
    Yuxiang Fu, Qi Yan, Lele Wang, Ke Li, and Renjie Liao
    In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025
  6. ICML
    Leveraging Online Olympiad-Level Math Problems for LLMs Training and Contamination-Resistant Evaluation
    Sadegh Mahdavi, Muchen Li, Kaiwen Liu, Christos Thrampoulidis, Leonid Sigal, and Renjie Liao
    In International Conference on Machine Learning (ICML), 2025
  7. IROS
    TrajFlow: Multi-modal Motion Prediction via Flow Matching
    Qi Yan, Brian Zhang, Yutong Zhang, Daniel Yang, Joshua White, Di Chen, Jiachao Liu, Langechuan Liu, Binnan Zhuang, Shaoshuai Shi, and  others
    In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025
  8. ICLR
    SymmetricDiffusers: Learning Discrete Diffusion on Finite Symmetric Groups
    Yongxing Zhang, Donglin Yang, and Renjie Liao
    In International Conference on Learning Representations (ICLR), Oral (1.8%), 2025

2022

  1. arXiv
    Gaussian-Bernoulli RBMs without Tears
    Renjie Liao, Simon Kornblith, Mengye Ren, David J Fleet, and Geoffrey Hinton
    arXiv preprint arXiv:2210.10318, 2022