Chain-of-Thought Reasoning is a Policy Improvement Operator

Hugh Zhang, David C. Parkes

Large language models have astounded the world with fascinating new capabilities. However, they currently lack the ability to teach themselves new skills, relying instead on large…