I guess it doesn't actually do more passes but pretending that it did might still give more precise results.

There was an article recently that said something like adding urgency to a prompt gave better results. I hope it doesn't stress the model out :D

https://arxiv.org/abs/2307.11760