Large Language Models (LLMs) have shown prominent performance in various downstream tasks in which prompt engineering plays a pivotal role in optimizing LLMs' performance. This paper, not as an overview of current prompt engineering methods, aims to highlight the limitation of designing prompts while holding an anthropomorphic assumption that expects LLMs to think like humans. From our review of 35 representative studies, we demonstrate that a goal-oriented prompt formulation, which guides LLMs to follow established human logical thinking, significantly improves the performance of LLMs. Furthermore, We introduce a novel taxonomy that categorizes goal-oriented prompting methods into five interconnected stages and we demonstrate the broad applicability of our framework by summarizing ten applicable tasks. With four future directions proposed, we hope to further emphasize and promote goal-oriented prompt engineering.

翻译：大语言模型（LLMs）在各类下游任务中展现出卓越性能，其中提示工程在优化LLMs表现中发挥关键作用。本文并非现有提示工程方法的全面综述，而是旨在揭示设计提示时存在的局限性——即持有拟人化假设，期望LLMs像人类一样思考。通过回顾35项代表性研究，我们证明：引导LLMs遵循既定人类逻辑思维的目标导向提示公式，能够显著提升LLMs的性能。此外，我们提出一种新颖的分类体系，将目标导向提示方法划分为五个相互关联的阶段，并通过总结十项适用任务验证了该框架的广泛适用性。基于提出的四个未来研究方向，我们希望进一步强调并推动目标导向提示工程的发展。