EvoFlow: Evolving Diverse Agentic Workflows On The Fly

TL;DR

EvoFlow is an evolutionary framework that automatically generates diverse, high-performing, and cost-effective heterogeneous agentic workflows for multi-agent systems, surpassing existing methods in performance and efficiency.

Contribution

EvoFlow introduces a niching evolutionary algorithm to automatically evolve diverse and adaptive agentic workflows, addressing limitations of homogeneity and single-objective optimization in prior systems.

Findings

EvoFlow produces a diverse set of workflows from simple to complex tasks.

EvoFlow outperforms previous workflows by up to 29.86% in performance.

EvoFlow reduces inference costs by 87.6% compared to powerful LLMs.

Abstract

The past two years have witnessed the evolution of large language model (LLM)-based multi-agent systems from labor-intensive manual design to partial automation (\textit{e.g.}, prompt engineering, communication topology) and eventually to fully automated design. However, existing agentic automation pipelines often lack LLM heterogeneity and focus on single-objective performance optimization, limiting their potential to combine weaker models for more customized and cost-effective solutions. To address this challenge, we propose EvoFlow, a niching evolutionary algorithm-based framework to automatically search a population of heterogeneous and complexity-adaptive agentic workflows, rather than a single homogeneous, complex workflow. Technically, EvoFlow performs \textit{(1) tag-based retrieval} to extract parent workflows from an agentic population, evolves new workflows through \textit{(2) crossover} and \textit{(3) mutation}, and employs \textit{(4) niching-based selection} to maintain population diversity and quality. Extensive evaluations across seven benchmarks demonstrate that EvoFlow is: \textbf{(I) diverse}, evolving a population of workflows ranging from simple I/O tasks to complex multi-turn interactions; \textbf{(II) high-performing}, outperforming previous handcrafted and automated workflows by $1.23\%\sim29.86\%$; \textbf{(III) economical}, surpassing powerful \llmname{o1-preview} at $12.4\%$ of its inference cost using weaker open-source models.