How does Naïve Bayes relate to variational inference frameworks?

Naïve Bayes can be seen as a fully factorized variational approximation of the joint distribution.

Variational inference approximates complex distributions by simpler ones. Naïve Bayes imposes a fully factorized structure over features conditioned on class labels, effectively acting as a restrictive variational family. It minimizes divergence between true distribution and approximated factorized form.