MASPO: Joint Prompt Optimization for LLM-based Multi-Agent Systems

2026-05-07

AI Summary

This work addresses the misalignment between locally optimized role prompts and global objectives in large language model–based multi-agent systems. To resolve this, the authors propose the MASPO framework, which employs a joint evaluation mechanism to assess how effectively a prompt facilitates downstream agents’ task success and integrates a data-driven evolutionary beam search strategy to efficiently explore the high-dimensional prompt space for globally coherent optimization. Notably, MASPO achieves system-wide goal alignment without requiring external supervision. Evaluated across six diverse collaborative tasks, the method consistently outperforms existing prompt optimization approaches, yielding an average accuracy improvement of 2.9 percentage points.

Abstract

Large language model (LLM)-based Multi-agent systems (MAS) have shown promise in tackling complex collaborative tasks, where agents are typically orchestrated via role-specific prompts. While the quality of these prompts is pivotal, jointly optimizing them across interacting agents remains a non-trivial challenge, primarily due to the misalignment between local agent objectives and holistic system goals. To address this, we introduce MASPO, a novel framework designed to automatically and iteratively refine prompts across the entire system. A core innovation of MASPO is its joint evaluation mechanism, which assesses prompts not merely by their local validity, but by their capacity to facilitate downstream success for successor agents. This effectively bridges the gap between local interactions and global outcomes without relying on ground-truth labels. Furthermore, MASPO employs a data-driven evolutionary beam search to efficiently navigate the high-dimensional prompt space. Extensive empirical evaluations across 6 diverse tasks demonstrate that MASPO consistently outperforms state-of-the-art prompt optimization methods, achieving an average accuracy improvement of 2.9. We release our code at https://github.com/wangzx1219/MASPO.

Problem

• multi-agent systems
• prompt optimization
• large language models
• joint optimization
• system alignment

Innovation

• joint prompt optimization
• multi-agent systems
• large language models
• evolutionary beam search
• prompt evaluation