CoT

Chain-of-Thought

A reasoning framework that forces explicit step-by-step thinking before the answer. Improves accuracy on maths, logic, and multi-step problems.

The components

1Problem framing
A clear statement of the problem the LLM is solving.
2Reasoning instruction
An explicit directive to think step by step before answering.
3Answer instruction
An explicit directive to give the final answer at the end, clearly separated from the reasoning.

Worked example

Rough prompt

If a train leaves Mombasa at 6am at 80kph and another leaves Nairobi at 7am at 100kph, when do they meet?

Super-prompt produced by CoT

[Problem framing] Two trains travel between Mombasa and Nairobi (approximately 480 km apart by rail) toward each other. The first leaves Mombasa at 06:00 at 80 km/h. The second leaves Nairobi at 07:00 at 100 km/h. The user wants to know the time at which they meet, accounting for the staggered departure.
[Reasoning instruction] Solve this step by step. First, compute the distance the first train has covered by 07:00. Second, compute the remaining distance between the trains at 07:00. Third, compute their combined closing speed. Fourth, compute the additional time required to close that distance. Fifth, add that time to 07:00 to find the meeting time. Show each step explicitly.
[Answer instruction] At the end, on a new line beginning with the word "Answer:", state the meeting time as HH:MM in 24-hour format. Do not include the reasoning in the answer line.

Try CoT on your own prompt.

Open the builder with CoT pre-selected and watch a rough idea become a structured super-prompt.

Build with CoT →