BAYESIAN FLOW NETWORKS

Summary from the TLDR AI newsletter

Unifying the modeling of discrete or continuous data is a challenge. More so when attempting to maintain a probabilistic likelihood interpretation of the output. This new paper showcases a new class of generative models with a pile of interesting properties such as no forward pass, gradient-based sample guidance for discrete data, and competitive language modeling likelihood.