A Toolbox for Improving Evolutionary Prompt Search

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Grießhaber, Daniel, Kimmich, Maximilian, Maucher, Johannes, Vu, Ngoc Thang
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908635840380928
author Grießhaber, Daniel
Kimmich, Maximilian
Maucher, Johannes
Vu, Ngoc Thang
author_facet Grießhaber, Daniel
Kimmich, Maximilian
Maucher, Johannes
Vu, Ngoc Thang
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2511_05120
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Toolbox for Improving Evolutionary Prompt Search
Grießhaber, Daniel
Kimmich, Maximilian
Maucher, Johannes
Vu, Ngoc Thang
Computation and Language
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.
title A Toolbox for Improving Evolutionary Prompt Search
topic Computation and Language
url https://arxiv.org/abs/2511.05120