ChainOfThought

A way of prompting LanguageModels to improve their performance.

Or maybe the terminology that means have extra prompts to tell the language model to plan / write out how it will produce the answer or solve the problem, and feed that back into it again. Basically it's a feedback loop or iteration of multi-runs through the language model.

https://towardsdatascience.com/chain-of-thought-prompting-facilitate-llms-reasoning-abilities-313cd7714938