Skip to content
FDE 101

Chain-of-thought prompting

Chain-of-thought prompting, often shortened to CoT, uses examples with intermediate reasoning steps to improve performance on some multi-step tasks.1

Tiny example

Section titled “Tiny example”

For a word problem, the prompt demonstrates how to identify quantities, apply operations, and then give a final answer instead of showing only the answer.

FDE note

Section titled “FDE note”

Do not treat hidden model reasoning as a reliable audit trail. Ask for concise, verifiable outputs—calculations, citations, tool results, and decision fields—and evaluate the final behavior.

Footnotes

Section titled “Footnotes”

  1. Wei et al., “Chain-of-Thought Prompting Elicits Reasoning in Large Language Models”, NeurIPS 2022. ↩