This AI Paper from UC Berkeley Introduces a Data-Efficient Approach to Long Chain-of-Thought Reasoning for Large Language Models