chain of thought

Chain of Thought (CoT) is a clever way to make large language models better at complex tasks. Instead of just asking for a direct answer, you prompt the model to 'think step by step' or 'explain your reasoning.' This simple addition can significantly improve the quality and accuracy of the model's output, especially for problems that require multiple steps of logic, arithmetic, or planning.

Think of it like asking a math student to show their work. When a language model breaks down a problem into smaller, manageable steps, it's more likely to arrive at the correct solution. This technique was a big deal when it was introduced because it unlocked new capabilities in models without needing to retrain them, just by changing how we ask questions. It essentially gives the model a chance to 'self-correct' along its internal thought process.