Reasoning Step-by-Step

While standard few-shot prompting works for simple tasks, it fails at complex arithmetic, commonsense, and symbolic reasoning. Chain-of-Thought (CoT) prompting solves this by enabling intermediate reasoning steps.

Instead of just showing the AI the question and the final answer, Few-Shot CoT shows the AI the 'scratchpad'—the step-by-step logic required to reach that answer.

If you don't have time to write out examples, you can use Zero-Shot CoT. Research by Kojima et al. (2022) proved that simply appending the phrase 'Let's think step by step' to your prompt drastically increases the model's accuracy on reasoning tasks.