NeuroMAS: Multi-Agent Systems as Neural Networks with Joint Reinforcement Learning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lu, Haoran, Fang, Luyang, Zhong, Wenxuan, Ma, Ping
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914571763056640
author Lu, Haoran
Fang, Luyang
Zhong, Wenxuan
Ma, Ping
author_facet Lu, Haoran
Fang, Luyang
Zhong, Wenxuan
Ma, Ping
contents Multi-agent language systems are often built as hand-designed workflows, where agents are assigned semantic roles and communication protocols are specified in advance. We propose NeuroMAS, a method that first treats a multi-agent language system as a trainable and scalable neural-network-like architecture with LLM agents as nodes and intermediate textual signals as edges. In NeuroMAS, agent nodes are role-free but structure-aware: the topology only determines how information can flow in general, while reinforcement learning training determines how nodes communicate, specialize, and coordinate. This formulation shifts multi-agent design from workflow engineering toward architecture design, where depth, width, connectivity, and growth protocol become scalable sources of capability. Further, we provide a theoretical perspective showing why such modular textual computation is more parameter-efficient when tasks admit hierarchical decompositions. Experiments show that NeuroMAS improves significantly over both inference-time and trained multi-agent baselines. We further find that organizational scaling is path-dependent: larger systems can be challenging to train from scratch, but become feasible when grown progressively from smaller trained systems. These results suggest that learned neural multi-agent systems are a promising scaling axis for LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2605_16757
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle NeuroMAS: Multi-Agent Systems as Neural Networks with Joint Reinforcement Learning
Lu, Haoran
Fang, Luyang
Zhong, Wenxuan
Ma, Ping
Artificial Intelligence
Multiagent Systems
Methodology
Machine Learning
Multi-agent language systems are often built as hand-designed workflows, where agents are assigned semantic roles and communication protocols are specified in advance. We propose NeuroMAS, a method that first treats a multi-agent language system as a trainable and scalable neural-network-like architecture with LLM agents as nodes and intermediate textual signals as edges. In NeuroMAS, agent nodes are role-free but structure-aware: the topology only determines how information can flow in general, while reinforcement learning training determines how nodes communicate, specialize, and coordinate. This formulation shifts multi-agent design from workflow engineering toward architecture design, where depth, width, connectivity, and growth protocol become scalable sources of capability. Further, we provide a theoretical perspective showing why such modular textual computation is more parameter-efficient when tasks admit hierarchical decompositions. Experiments show that NeuroMAS improves significantly over both inference-time and trained multi-agent baselines. We further find that organizational scaling is path-dependent: larger systems can be challenging to train from scratch, but become feasible when grown progressively from smaller trained systems. These results suggest that learned neural multi-agent systems are a promising scaling axis for LLMs.
title NeuroMAS: Multi-Agent Systems as Neural Networks with Joint Reinforcement Learning
topic Artificial Intelligence
Multiagent Systems
Methodology
Machine Learning
url https://arxiv.org/abs/2605.16757