2024-11-27LLMs Learn Task Heuristics from Demonstrations: A Heuristic-Driven Prompting Strategy for Document-1.概述关于基于COT的Prompt构造有很多的研究,例如:CoT(Weietal.,2022),Automate-CoT(Shumetal.,2023),Auto-CoT(Zhangetal.,2023),Iter-CoT(Sunetal.,2023),Active-CoT(Diaoetal.,2023)。本篇文章尝试给出了一种解释:LLM基于有监督的ICL(in-contextlearni