Structured Reasoning for Large Language Models

Jinyi Han, Zixiang Di, Zishang Jiang +4

Large language models (LLMs) achieve strong performance by generating long chains of thought, but longer traces always introduce redundant or ineffective reasoning steps. One typic…