Let's think Step by Step
https://arxiv.org/abs/2205.11916
Large Language Models are Zero-Shot Reasoners
Zero-shot-CoT