Flow Matching
Flow matching learns to transform noise into data by modelling the continuous path between them, rather than the step-by-step denoising process used in traditional Diffusion Models. Originally described in paper Flow Matching for Generative Modeling.