Chain of Thought (CoT) prompting encourages the model to break down complex problems into intermediate reasoning steps before arriving at a final answer. Instead of jumping to conclusions, the model "thinks out loud."
This technique dramatically improves performance on math, logic, and multi-step reasoning tasks by making the reasoning process explicit and verifiable.