Large Language Models (LLMs) - Stanford CME296 Diffusion & Large Vision Models | Spring 2026 | Lecture 3 - Flow matching

Learn more details about this course: https://online.stanford.edu/courses/cme296-diffusion-and-large-vision-models To follow along with the course schedule and syllabus, visit: https://cme296.stanford.edu/syllabus/ Chapters: 00:00:00 Introduction 00:00:55 Recap of last episodes 00:06:58 Problem formulation 00:10:53 Trajectory 00:11:54 Flow 00:13:59 Probability path 00:15:05 Vector field 00:18:42 Ordinary differential equation 00:23:51 Continuity equation 00:36:34 Micro and macro perspectives 00:41:15 Flow models 00:43:01 Continuous normalizing flows 00:46:16 Flow matching 00:47:58 Conditional probability path and vector field 01:01:10 Marginal probability path and vector field 01:10:54 Conditional flow matching 01:26:02 Why not velocity matching 01:28:50 Training 01:29:55 Inference 01:30:31 Rectified flow 01:43:05 Links with diffusion and score matching For more information about Stanford’s graduate programs, visit: https://online.stanford.edu/graduate-education Afshine Amidi is an Adjunct Lecturer at Stanford University. Shervine Amidi is an Adjunct Lecturer at Stanford University. View the course playlist: https://www.youtube.com/playlist?list=PLoROMvodv4rNdy8rt2rZ4T2xM0OjADnfu