Learn how forcing the model to reason step by step dramatically improves performance on complex tasks.