Self-Evolving Multi-Agent Systems via Decentralized Memory

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hao, Guangya, Long, Yunbo, Zhao, Zhuokai
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913153246298112
author Hao, Guangya
Long, Yunbo
Zhao, Zhuokai
author_facet Hao, Guangya
Long, Yunbo
Zhao, Zhuokai
contents Self-evolving multi-agent systems (MAS) have emerged as a promising route to LLM agents that continually improve from experience, with persistent memory at their foundation. However, existing designs almost exclusively adopt a centralized repository shared across agents, incurring communication and coordination overhead, raising privacy concerns, and collapsing agent diversity. We propose DecentMem, a decentralized memory framework in which each agent maintains its own dual-pool memory -- an exploitation pool of consolidated past trajectories and an exploration pool of LLM-generated candidates for unseen contexts. The two pools are reweighted online based on stage-wise feedback from an LLM-as-a-judge. Theoretically, we prove that this design guarantees global reachability of the solution space and achieves $O(\log T)$ cumulative regret, matching the stochastic bandit lower bound up to constants. In practice, across three MAS frameworks (AutoGen, DyLAN, AgentNet), three Qwen3 backbones (4B/8B/14B), two Gemma4 backbones (E2B/E4B) and five benchmarks spanning math, code, QA, and embodied tasks, DecentMem improves average accuracy by up to 23.8% over the strongest centralized memory baseline and by up to 52.5% over the no-memory baseline, while reducing token usage by up to 49%.
format Preprint
id arxiv_https___arxiv_org_abs_2605_22721
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Self-Evolving Multi-Agent Systems via Decentralized Memory
Hao, Guangya
Long, Yunbo
Zhao, Zhuokai
Multiagent Systems
Self-evolving multi-agent systems (MAS) have emerged as a promising route to LLM agents that continually improve from experience, with persistent memory at their foundation. However, existing designs almost exclusively adopt a centralized repository shared across agents, incurring communication and coordination overhead, raising privacy concerns, and collapsing agent diversity. We propose DecentMem, a decentralized memory framework in which each agent maintains its own dual-pool memory -- an exploitation pool of consolidated past trajectories and an exploration pool of LLM-generated candidates for unseen contexts. The two pools are reweighted online based on stage-wise feedback from an LLM-as-a-judge. Theoretically, we prove that this design guarantees global reachability of the solution space and achieves $O(\log T)$ cumulative regret, matching the stochastic bandit lower bound up to constants. In practice, across three MAS frameworks (AutoGen, DyLAN, AgentNet), three Qwen3 backbones (4B/8B/14B), two Gemma4 backbones (E2B/E4B) and five benchmarks spanning math, code, QA, and embodied tasks, DecentMem improves average accuracy by up to 23.8% over the strongest centralized memory baseline and by up to 52.5% over the no-memory baseline, while reducing token usage by up to 49%.
title Self-Evolving Multi-Agent Systems via Decentralized Memory
topic Multiagent Systems
url https://arxiv.org/abs/2605.22721