Chain of Thought (CoT)

Chain-of-thought prompting asks the model to show its working — laying out intermediate reasoning steps — before stating an answer. Wei et al. (2022) found that this simple change substantially improves performance on arithmetic, commonsense and symbolic-reasoning benchmarks, and that the effect strengthens with model scale.

The trade-off is speed and token cost: more reasoning means longer outputs. It helps most on problems that genuinely require several steps.