ReSo: A Reward-driven Self-organizing LLM-based Multi-Agent System for Reasoning Tasks

Heng Zhou, Hejia Geng, Xiangyuan Xue, Li Kang, Yiran Qin +3 more
3/4/2025
cs.MA

No official code found. Best implementation: hengzzzhou/ReSo (25 stars, last commit 2026-01-29).

Abstract

Multi-agent systems (MAS) have emerged as a promising approach for enhancing the reasoning capabilities of large language models in complex problem-solving; however, current MAS frameworks suffer from poor flexibility and scalability with underdeveloped optimization strategies. To address these challenges, we propose ReSo, which integrates task graph generation with a reward-driven two-stage agent selection process centered on our Collaborative Reward Model that provides fine-grained reward signals to optimize MAS cooperation. We also introduce an automated data synthesis framework for generating MAS benchmarks without any human annotations. Experimental results show that ReSo matches or outperforms existing methods, achieving 33.7 percent accuracy on Math-MAS and 32.3 percent accuracy on SciBench-MAS, where other approaches completely fail.

View on arXivView PDF

Code Implementations(1)

hengzzzhou/ReSoCommunity100%
252Shell, PythonFeb 17, 20258 months ago

Similar Papers

MasHost Builds It All: Autonomous Multi-Agent System Directed by Reinforcement Learning

86% match2025

A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaboration

86% match1 repo2132023

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

85% match1 repo262026

HieraMAS: Optimizing Intra-Node LLM Mixtures and Inter-Node Topology for Multi-Agent Systems

85% match2026

RewardFlow: Topology-Aware Reward Propagation on State Graphs for Agentic RL with Large Language Models

85% match1 repo132026

Active Learning for Communication Structure Optimization in LLM-Based Multi-Agent Systems

84% match2026

Cite this paper

@article{zhou2025reso,
  title  = {ReSo: A Reward-driven Self-organizing LLM-based Multi-Agent System for Reasoning Tasks},
  author = {Heng Zhou and Hejia Geng and Xiangyuan Xue and Li Kang and Yiran Qin and Zhiyong Wang and Zhenfei Yin and Lei Bai},
  year   = {2025},
  eprint = {2503.02390},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2503.02390v3}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2503.02390.svg)](https://codeofpaper.com/paper/2503.02390)

Discussion