What is Chain-of-Thought?
Also known as: CoT
Chain-of-Thought (CoT) is a prompting technique that asks an AI model to generate intermediate reasoning steps before giving a final answer, helping it solve complex problems more reliably.
Instead of jumping straight to an answer, CoT prompting guides the model to break a task into explicit logical steps, similar to how humans show their work on math or logic problems.
It can be triggered with few-shot examples that include step-by-step reasoning or with simple zero-shot instructions like 'think step by step.' This elicits more structured thinking from large language models.
The approach improves performance on arithmetic, commonsense, and symbolic reasoning tasks by making the model's intermediate logic visible and correctable.
Example
When asked 'If John has 3 apples and buys 2 more, then eats 1, how many does he have?', a CoT prompt leads the model to write: 'John starts with 3. He buys 2 more, so 3+2=5. He eats 1, so 5-1=4. Answer: 4.'
Why it matters
CoT has become a standard method to boost reasoning accuracy in large language models without additional training, making AI outputs more reliable and interpretable for complex tasks.
Frequently asked questions
No, it helps with many reasoning tasks including logic puzzles, commonsense questions, and multi-step planning.
Related terms
Prompt engineering is the practice of designing and refining text inputs (prompts) to guide AI models like large language models toward producing accurate, relevant, or creative outputs.
Few-shot learning (in prompting) is a technique where a language model is given a handful of input-output examples directly in the prompt to guide it on a new task.
A prompt is the input text, question, or instruction given to an AI model (especially a large language model) to guide what it should generate or how it should respond.
A system prompt is the initial set of instructions given to an AI model that defines its overall behavior, role, rules, and tone for the conversation.
Zero-shot learning (or zero-shot prompting) is when an AI model completes a task it has never seen examples of during training, relying only on a natural language description of what to do.