Regularizing the entropy to be higher could encourage the model to give more solutions.
I do think that you could probably regularize the entropy to be higher. So you're encouraging the model to give you more and more solutions, but you don't want it to start deviating too much from the training data.