A Systematic Survey of Automatic Prompt Optimization Techniques (2502.16923v2)

Abstract: Since the advent of LLMs, prompt engineering has been a crucial step for eliciting desired responses for various NLP tasks. However, prompt engineering remains an impediment for end users due to rapid advances in models, tasks, and associated best practices. To mitigate this, Automatic Prompt Optimization (APO) techniques have recently emerged that use various automated techniques to help improve the performance of LLMs on various tasks. In this paper, we present a comprehensive survey summarizing the current progress and remaining challenges in this field. We provide a formal definition of APO, a 5-part unifying framework, and then proceed to rigorously categorize all relevant works based on their salient features therein. We hope to spur further research guided by our framework.

• The paper formalizes the concept of Automatic Prompt Optimization (APO) and introduces a unifying five-part framework for categorizing existing research.
• The survey addresses the challenges of manual prompt engineering for Large Language Models (LLMs) given the rapid advancements in models and tasks.
• By systematically surveying and categorizing existing techniques, the paper aims to guide and structure future research directions in the field of APO.

The paper "A Systematic Survey of Automatic Prompt Optimization Techniques" (2502.16923), published on February 24, 2025, presents a comprehensive survey of Automatic Prompt Optimization (APO) techniques, which have emerged as a way to improve the performance of LLMs on various NLP tasks. The paper addresses the challenges of prompt engineering for end-users due to rapid advancements in models and tasks. It offers a formal definition of APO and a unifying five-part framework for categorizing existing research in the field. The paper aims to guide future research using this framework.

In summary, this survey paper formalizes the concept of APO, provides a framework for understanding different techniques, and categorizes existing works to guide future research in optimizing prompts for LLMs.