AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems

📄 arXiv: 2609.08572v1 📥 PDF

作者: Jaewon Chu, Jinwoo Seo, Jaewon Cho, Jeehye Na, Yunyang Xiong, Youngdae Kim, Hyunwoo J. Kim

分类: cs.AI

发布日期: 2026-09-08

备注: 13 pages

💡 一句话要点

提出AgentGrad以解决多智能体系统中的提示优化问题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 多智能体系统 提示优化 文本梯度 顺序干预 语义聚类 性能提升 人工智能

📋 核心要点

  1. 现有文本梯度方法在梯度提取和聚合阶段存在不足，导致无法有效优化多智能体系统的提示设计。
  2. AgentGrad框架通过顺序干预逐一修改智能体行为，并利用语义文本梯度抽象来优化提示，解决了现有方法的局限性。
  3. 实验结果显示，AgentGrad在五个基准测试中表现优异，优化时间平均减少2.5倍，显著提升了效率。

📝 摘要（中文）

基于大型语言模型的多智能体系统在使用多个专用智能体时表现出色，但其性能依赖于每个智能体的提示设计。本文识别了现有文本梯度方法在梯度提取和梯度聚合两个阶段的局限性，并提出了AgentGrad框架。该框架通过顺序干预和语义文本梯度抽象来优化提示，能够有效识别并解决失败模式。实验结果表明，AgentGrad在五个多智能体系统基准测试中实现了最先进的性能，并将优化时间平均减少了2.5倍。

🔬 方法详解

问题定义：本文旨在解决多智能体系统中提示优化的挑战，现有方法在梯度提取时缺乏代理级监督，且在梯度聚合时混合了不相关的失败模式，导致优化效果不佳。

核心思路：AgentGrad通过顺序干预逐一修改智能体的行为，以识别并解决特定的失败模式，同时利用语义文本梯度抽象来聚合相似的梯度，避免混淆不同的失败模式。

技术框架：AgentGrad的整体架构包括两个主要模块：顺序干预模块和语义文本梯度抽象模块。顺序干预模块负责逐一修改智能体的输出，语义文本梯度抽象模块则对提取的梯度进行聚类和抽象。

关键创新：最重要的创新在于引入了顺序干预和语义文本梯度抽象的结合，能够有效识别目标智能体并提取细粒度梯度，从而提升了提示优化的准确性和效率。

关键设计：在设计中，顺序干预确保每次只修改一个智能体的输出，避免了混淆；语义文本梯度抽象则通过聚类相似梯度，形成通用梯度，捕捉共享的纠正模式。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明，AgentGrad在五个多智能体系统基准测试中达到了最先进的性能，相比于下一个最快的基线，优化时间平均减少了2.5倍，显示出显著的效率提升。

🎯 应用场景

该研究的潜在应用领域包括智能客服、自动化决策系统和多智能体协作任务等。通过优化多智能体系统的提示设计，AgentGrad能够提升系统的响应能力和准确性，具有重要的实际价值和广泛的应用前景。

📄 摘要（原文）

Large language model (LLM)-based multi-agent systems (MAS) achieve strong performance by employing specialized multiple agents, yet their performance depends on the prompt design of each agent. For MAS prompt optimization, textual gradient methods that guide prompt updates using natural-language feedback have emerged as a leading paradigm. In this paper, we identify limitations in two stages of existing textual gradient approaches: gradient extraction and gradient aggregation. In gradient extraction, previous works select a target prompt without verifying whether modifying it resolves the failure, and derive gradients without agent-level supervision over the corresponding agent's intermediate output. In gradient aggregation, individual gradients are randomly grouped and concatenated, often mixing unrelated failure modes and producing prompts that fail to generalize. To address these limitations, we propose \textbf{AgentGrad}, a prompt optimization framework for multi-agent systems based on sequential intervention and semantic textual gradient abstraction. For each failure, sequential intervention modifies the behavior of one agent at a time to identify the target agent whose modification resolves the failure. The modified output of the target agent then serves as agent-level supervision for extracting a fine-grained gradient. Semantic textual gradient abstraction clusters semantically similar gradients to prevent mixing unrelated failure modes, and abstracts each cluster into a generalized gradient that captures the shared corrective pattern. Experimental results show that AgentGrad achieves state-of-the-art performance across five MAS benchmarks and reduces wall-clock optimization time by $2.5\times$ on average compared to the next-fastest baseline.