They’re referring to a method of prompting the model that encourages it to think through something step by step rather than spit out the answer. I think this i

They’re referring to a method of prompting the model that encourages it to think through something step by step rather than spit out the answer. I think this is the paper that really kicked off this technique: https://arxiv.org/abs/2201.11903 https://arxiv.org/abs/2201.11903