AgentSlimming: Towards Efficient and Cost-Aware Multi-Agent Systems

TL;DR

AgentSlimming is a framework that compresses multi-agent workflows to reduce token costs by pruning and replacing redundant agents, maintaining performance and improving efficiency.

Contribution

It introduces a novel, plug-and-play compression method for multi-agent systems that effectively reduces costs while preserving or enhancing task performance.

Findings

Reduced token cost by up to 78.9% with negligible performance loss.

Achieved a Pareto-optimal trade-off between cost and quality.

Code is publicly available at https://github.com/CitrusYL/AgentSlimming

Abstract

Large Language Model-based Multi-Agent Systems (MAS) have demonstrated remarkable capabilities in complex tasks. However, manually designing optimal communication topologies is labor-intensive, while automated expansion methods often result in bloated structures with redundant agents, leading to excessive token consumption. To address this problem, we introduce \textbf{AgentSlimming}, a plug-and-play compression framework for graph-structured multi-agent workflows. Motivated by pruning and quantization in neural networks, AgentSlimming compresses workflows by first estimating the importance score of each agent with a hybrid mechanism, and then removes redundant agents or replaces them with low-cost ones, where each operation is validated using a baseline-anchored acceptance rule to prevent performance collapse. Experiments show that AgentSlimming reduces average token cost by up to 78.9\% with negligible performance degradation, and sometimes even improves accuracy, achieving a strong Pareto-optimal trade-off between cost and quality. \textit{Our code is publicly available at https://github.com/CitrusYL/AgentSlimming