Mining gold: Improving simulation-based inference with latent information

We summarize and discuss new inference techniques for systems that are described
by a simulator with an intractable likelihood function. The key idea is that additional information that characterizes the latent process can often be extracted from
the simulator. It can then be used to augment the training data for neural surrogates of the likelihood function. These methods have been applied to problems in
particle physics and astrophysics, and the initial results demonstrate their potential
to improve sample efficiency and quality of inference.