arXiv cs.MA· Vishal Bhutani, Prasang Gupta, Sumanth Chundru, Kevin Paul·· 2 天前AI 评分31
MetaShare：面向开放式智能体的失败感知集体自我改进框架
MetaShare: Failure-Aware Collective Self-Improvement for Open-Ended Agents
AI 导读
MetaShare 是一个失败感知的智能体自我改进框架，将成功与失败的智能体修改连同带符号的性能变化一并记录，在每代进化前从其他谱系检索多样化高性能策略与高置信度失败案例，避免重复已知死路。
正文 · 原文
Abstract:Humans often learn fastest not by repeating success, but by remembering which failures not to repeat. Open-ended self-improving agents can share experience across lineages, as in GEA, or edit their own improvement mechanisms, as in DGM-Hyperagents, yet failed improvement attempts are rarely retained as reusable cross-branch knowledge, and search can collapse onto redundant strategies. We introduce MetaShare, a failure-aware framework that records successful and unsuccessful agent modifications together with signed performance deltas. Before each generation, MetaShare retrieves diverse high-performing strategies alongside high-confidence failures from other lineages, letting agents reuse effective mechanisms while avoiding known dead ends. A task-blind auditor periodically injects absent or underrepresented mechanism categories, while embedding-based de-ranking limits redundancy during retrieval. Multiple meta-agents propose modifications in parallel at heterogeneous temperatures behind a deterministic selector that gates applicability, scope, compilability, and size. Experiments across paper review, coding, math grading, robotics reward design, and financial audit-risk assessment show that MetaShare's median matches or exceeds DGM-Hyperagents' final performance and reaches its own best-performing agent in 1.7--2.9$\times$ fewer evolution iterations.
| Comments: | 30 pages, 4 figures. Accepted at NeurIPS 2026 Meta-Agents Workshop (Managing Agents that Manage Agents) |
| Subjects: | Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2610.21887 [cs.MA] |
| (or arXiv:2610.21887v1 [cs.MA] for this version) | |
| https://doi.org/10.48550/arXiv.2610.21887 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Vishal Bhutani [view email]
[v1]
Thu, 15 Oct 2026 13:49:41 UTC (674 KB)
来源：arXiv cs.MA · arxiv.org