In this article, we consider sufficient dimension folding for the regression mean function when predictors are matrix- or array-valued. We propose a new concept named central mean dimension folding ...