With standard prompting, language models are given input-output pairs before predicting the answer. With Chain of Thought prompting, they are prompted to produce intermediate reasoning steps before giving the final answer.

→ View original post on X — @saboo_shubham_