356. Heavy-Tailed Test Errors
medium

A model's prediction errors appear normally distributed on the training set but show heavy tails on the test set. What is the most likely explanation?