Chain-of-Thought

A prompting technique where the model is instructed (or trained) to write out its reasoning step by step before producing a final answer, which tends to improve accuracy on complex tasks.

Example: Instead of directly answering 'What's 17 × 24?', the model writes: '17 × 20 = 340, 17 × 4 = 68, total = 408.'

In practice: Chain-of-thought is foundational to most agentic reasoning. The 'thought' portion of ReAct is essentially chain-of-thought applied in an action loop. Visible reasoning also makes agent behavior easier to debug.

• builds on ReActThe 'Reason' step in ReAct is essentially chain-of-thought reasoning applied before each action.