A Toolbox for Improving Evolutionary Prompt Search

TL;DR

This paper introduces a comprehensive toolbox for enhancing evolutionary prompt search by decomposing the process, incorporating LLM-based verification, integrating human feedback, and optimizing evaluation efficiency, thereby improving prompt refinement for large language models.

Contribution

It presents novel methods for improving evolutionary prompt search, including modular steps, LLM judges, human feedback integration, and efficient evaluation strategies, applicable broadly to prompt optimization.

Findings

Enhanced prompt optimization quality

Reduced computational overhead

Generalizable improvements to evolutionary methods

Abstract

Evolutionary prompt optimization has demonstrated effectiveness in refining prompts for LLMs. However, existing approaches lack robust operators and efficient evaluation mechanisms. In this work, we propose several key improvements to evolutionary prompt optimization that can partially generalize to prompt optimization in general: 1) decomposing evolution into distinct steps to enhance the evolution and its control, 2) introducing an LLM-based judge to verify the evolutions, 3) integrating human feedback to refine the evolutionary operator, and 4) developing more efficient evaluation strategies that maintain performance while reducing computational overhead. Our approach improves both optimization quality and efficiency. We release our code, enabling prompt optimization on new tasks and facilitating further research in this area.