While you can apply zero shot learning and get the answer to a new math problem, you are only apply the learning to significant depth after a fine-tuning session - sleep.