Chain-of-Thought

A prompting technique where a model is encouraged to work through a problem step by step before giving a final answer, rather than jumping straight to a conclusion. Chain-of-thought prompting significantly improves performance on complex reasoning tasks, as showing the model how to think out loud leads to more accurate and reliable outputs.
See also: prompt engineering, large language model.