I’m surprised this is downvoted. I liked this paper (I’m mildly biased since I’ve been doing generative modeling work and am very excited about flows) but was disappointed to see MNIST and CIFAR-10.