[论文] Recursive Multi-Agent Systems

论文概要

研究领域: NLP 作者: Xiyuan Yang, Jiaru Zou, Rui Pan, etc. 发布时间: 2026-04-29 arXiv: 2504.21178

中文摘要

递归或循环语言模型最近作为一种新的扩展维度出现，通过在潜在状态上迭代优化同一模型计算来深化推理。本文将这一扩展原则从单一模型延伸到多智能体系统，并提出问题：智能体协作本身能否通过递归进行扩展？为此，我们引入了 RecursiveMAS，一种递归多智能体框架，将整个系统视为统一的潜在空间递归计算。RecursiveMAS 通过轻量级的 RecursiveLink 模块将异构智能体连接为协作循环，实现分布内潜在思维生成和跨智能体潜在状态转移。为优化框架，我们开发了内外循环学习算法，通过跨递归轮次的共享梯度信用分配进行迭代式全系统协同优化。对运行时复杂度和学习动态的理论分析表明，RecursiveMAS 比标准文本多智能体系统更高效，且在递归训练期间保持稳定的梯度。在数学、科学、医学、搜索和代码生成等9个基准测试中，RecursiveMAS 相比先进单/多智能体和递归计算基线，平均准确率提升8.3%，端到端推理加速1.2-2.4倍，token使用量减少34.6%-75.6%。

原文摘要

Recursive or looped language models have recently emerged as a new scaling axis by iteratively refining the same model computation over latent states to deepen reasoning. We extend such scaling principle from a single model to multi-agent systems, and ask: Can agent collaboration itself be scaled through recursion? To this end, we introduce RecursiveMAS, a recursive multi-agent framework that casts the entire system as a unified latent-space recursive computation. RecursiveMAS connects heterogeneous agents as a collaboration loop through the lightweight RecursiveLink module, enabling in-distribution latent thoughts generation and cross-agent latent state transfer. To optimize our framework, we develop an inner-outer loop learning algorithm for iterative whole-system co-optimization through...

自动采集于 2026-04-30

#论文 #arXiv #NLP #小凯