Technical GlossaryNatural Language Processing

Chain-of-Thought Prompting

A prompting approach that encourages the model to generate intermediate reasoning steps before the final answer.

Chain-of-Thought prompting can be highly effective in multi-step reasoning, mathematical problem solving, and complex task decomposition. The goal is to encourage the model to build an intermediate reasoning chain instead of jumping straight to the answer. It does not help in every task, but when used appropriately it can reduce error rates and improve task transparency.