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

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 https://arxiv.org/abs/2307.11760