Chain-of-Thought (CoT)
Level 2Short Description
A prompting and reasoning technique where the model explicitly works through intermediate steps before producing a final answer.
Friendly Description: Chain-of-Thought is when an AI shows its work, just like your math teacher always asked you to do. Instead of jumping straight to an answer, the AI walks through the problem step by step. This usually leads to better answers, because thinking out loud helps the AI catch its own mistakes along the way.
Example: Ask an AI, "If a train leaves Chicago at 3 PM going 60 mph and another leaves St. Louis at 4 PM going 70 mph, when do they meet?" With chain-of-thought, the AI first writes out the distances, then sets up the equation, then solves it, and finally states the answer. Showing its reasoning makes it both more accurate and easier for you to double-check.