Seminar| Institute of Mathematical Sciences
Time: Friday, October 10th, 2025,10:30-11:30
Location: RS408, I
Speaker: Xuhui Fan, Macquarie University
Abstract: Pre-trained diffusion models are commonly used to generate clean data (e.g., images) from random noises, effectively forming pairs of noises and corresponding clean images. Distillation on these pre-trained models can be viewed as the process of constructing advanced trajectories within the pair to accelerate sampling. For instance, consistency model distillation develops consistent projection functions to regulate trajectories, although sampling efficiency remains a concern. Rectified flow method enforces straight trajectories to enable faster sampling, yet relies on numerical ODE solvers, which may introduce approximation errors. In this work, we bridge the gap between the consistency model and the rectified flow method by proposing a Straight Consistent Trajectory (SCoT) model. SCoT enjoys the benefits of both approaches for fast sampling, producing trajectories with consistent and straight properties simultaneously. These dual properties are strategically balanced by targeting two critical objectives: (1) regulating the gradient of SCoT’s mapping to a constant, (2) ensuring trajectory consistency. Extensive experimental results demonstrate the effectiveness and efficiency of SCoT.
报告人简介: Dr Xuhui Fan is currently a lecturer in AI (equivalent to assistant professor) in the School of Computing at the Macquarie University, Australia. Before that, he received a bachelor's degree in mathematical statistics in China, and a PhD degree in computer science from the University of Technology Sydney, Australia. He then worked as a project engineer at Data61 (previously NICTA), CSIRO, as a postdoc fellow in the School of Mathematics and Statistics at the University of New South Wales, and as a lecturer at the University of Newcastle, Australia. Dr Fan's current research focuses on generative AI. He has published related research work in NeurIPS, ICML, AISTATS, JMLR, T-PAMI, etc. He serves as senior PC members in IJCAI and AAAI, PC members in ICML, NeurIPS, AISTATS, ICLR, COLT, and reviewers for JMLR, T-PAMI.