There's https://journals.sagepub.com/doi/abs/10.1177/027836491985942.... I think Aaron Ames' group is working on theoretical guarantees of neural net policies as well.