Chain of thought
Having a model write out intermediate reasoning steps before its answer, which improves accuracy on hard problems.
Also: CoT, step-by-step reasoning
In 2022 Jason Wei and colleagues at Google showed that giving models examples with worked reasoning, rather than bare answers, greatly improved their maths and logic. The same year, Takeshi Kojima and colleagues found that simply adding "Let's think step by step" helped. Writing steps lets the model spend more computation on a problem and break it into pieces. Reasoning models are trained to do this at length on their own. Whether the written steps faithfully reflect how the model reached its answer is an open research question.