05 · Getting Models to Act
CoT
Chain of Thought
Prompting the model to reason step by step rather than blurt a final answer. Measurably improves math and logic in ordinary models. Modern “reasoning models” already do it internally — with those you give a clear goal and let them think, rather than telling them to “think step by step.”
Concrete example
Adding “think step by step” to a word problem often flips a wrong answer into a right one.
Why it matters
A simple prompt change that reliably lifts performance on anything requiring reasoning.