What is chain-of-thought prompting?

Answer

Chain-of-thought prompting asks a language model to work through a problem step by step before giving a final answer, rather than responding immediately. Exposing intermediate reasoning often improves accuracy on math, logic, and multi-step tasks because the model allocates more computation to the problem. The trade-off is longer, more expensive outputs, and the visible reasoning is a plausible explanation rather than a guaranteed account of the model's internal process.