In this paper, M²LInES postdoc Arthur Guillaumin and lead PI Laure Zanna developed a stochastic deep learning parametrization that is trained on high-resolution data from CM2.6, a coupled climate model. Read their paper here