Multitask Learning with Stochastic Interpolants

TL;DR

We propose a framework for learning maps between probability distributions that broadly generalizes the time dynamics of flow and diffusion models…

Opening excerpt from the authors’ abstract. source