Group-Evolving Agents: Open-Ended Self-Improvement via Experience Sharing

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Weng, Zhaotian, Antoniades, Antonis, Nathani, Deepak, Zhang, Zhen, Pu, Xiao, Wang, Xin Eric
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866917248291045376
author Weng, Zhaotian
Antoniades, Antonis
Nathani, Deepak
Zhang, Zhen
Pu, Xiao
Wang, Xin Eric
author_facet Weng, Zhaotian
Antoniades, Antonis
Nathani, Deepak
Zhang, Zhen
Pu, Xiao
Wang, Xin Eric
contents Open-ended self-improving agents can autonomously modify their own structural designs to advance their capabilities and overcome the limits of pre-defined architectures, thus reducing reliance on human intervention. We introduce Group-Evolving Agents (GEA), a new paradigm for open-ended self-improvements, which treats a group of agents as the fundamental evolutionary unit, enabling explicit experience sharing and reuse within the group throughout evolution. Unlike existing open-ended self-evolving paradigms that adopt tree-structured evolution, GEA overcomes the limitation of inefficient utilization of exploratory diversity caused by isolated evolutionary branches. We evaluate GEA on challenging coding benchmarks, where it significantly outperforms state-of-the-art self-evolving methods (71.0% vs. 56.7% on SWE-bench Verified, 88.3% vs. 68.3% on Polyglot) and matches or exceeds top human-designed agent frameworks (71.8% and 52.0% on two benchmarks, respectively). Analysis reveals that GEA more effectively converts early-stage exploratory diversity into sustained, long-term progress, achieving stronger performance under the same number of evolved agents. Furthermore, GEA exhibits consistent transferability across different coding models and greater robustness, fixing framework-level bugs in 1.4 iterations on average, versus 5 for self-evolving methods.
format Preprint
id arxiv_https___arxiv_org_abs_2602_04837
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Group-Evolving Agents: Open-Ended Self-Improvement via Experience Sharing
Weng, Zhaotian
Antoniades, Antonis
Nathani, Deepak
Zhang, Zhen
Pu, Xiao
Wang, Xin Eric
Artificial Intelligence
Open-ended self-improving agents can autonomously modify their own structural designs to advance their capabilities and overcome the limits of pre-defined architectures, thus reducing reliance on human intervention. We introduce Group-Evolving Agents (GEA), a new paradigm for open-ended self-improvements, which treats a group of agents as the fundamental evolutionary unit, enabling explicit experience sharing and reuse within the group throughout evolution. Unlike existing open-ended self-evolving paradigms that adopt tree-structured evolution, GEA overcomes the limitation of inefficient utilization of exploratory diversity caused by isolated evolutionary branches. We evaluate GEA on challenging coding benchmarks, where it significantly outperforms state-of-the-art self-evolving methods (71.0% vs. 56.7% on SWE-bench Verified, 88.3% vs. 68.3% on Polyglot) and matches or exceeds top human-designed agent frameworks (71.8% and 52.0% on two benchmarks, respectively). Analysis reveals that GEA more effectively converts early-stage exploratory diversity into sustained, long-term progress, achieving stronger performance under the same number of evolved agents. Furthermore, GEA exhibits consistent transferability across different coding models and greater robustness, fixing framework-level bugs in 1.4 iterations on average, versus 5 for self-evolving methods.
title Group-Evolving Agents: Open-Ended Self-Improvement via Experience Sharing
topic Artificial Intelligence
url https://arxiv.org/abs/2602.04837