Evaluating the similarity of the functions learned by two models

Question

Say there are two machine learning models learning from two different datasets with the same features. Is there any way to concretely evaluate (or check) that they have lea