Multi-agent Architecture Search via Agentic Supernet

TL;DR

This paper introduces MaAS, a framework that optimizes a probabilistic supernet of multi-agent architectures, enabling dynamic, query-dependent resource allocation and significantly reducing inference costs while improving performance.

Contribution

It proposes the agentic supernet and MaAS framework to automate and optimize multi-agent system design for dynamic resource allocation and cost efficiency.

Findings

MaAS reduces inference costs by 6-45% compared to existing systems.

MaAS outperforms existing systems by 0.54-11.82% in accuracy.

MaAS demonstrates strong transferability across datasets and LLM backbones.

Abstract

Large Language Model (LLM)-empowered multi-agent systems extend the cognitive boundaries of individual agents through disciplined collaboration and interaction, while constructing these systems often requires labor-intensive manual designs. Despite the availability of methods to automate the design of agentic workflows, they typically seek to identify a static, complex, one-size-fits-all system, which, however, fails to dynamically allocate inference resources based on the difficulty and domain of each query. To address this challenge, we shift away from the pursuit of a monolithic agentic system, instead optimizing the \textbf{agentic supernet}, a probabilistic and continuous distribution of agentic architectures. We introduce MaAS, an automated framework that samples query-dependent agentic systems from the supernet, delivering high-quality solutions and tailored resource allocation (\textit{e.g.}, LLM calls, tool calls, token cost). Comprehensive evaluation across six benchmarks demonstrates that MaAS \textbf{(I)} requires only $6\sim45\%$ of the inference costs of existing handcrafted or automated multi-agent systems, \textbf{(II)} surpasses them by $0.54\%\sim11.82\%$, and \textbf{(III)} enjoys superior cross-dataset and cross-LLM-backbone transferability.