返回全部动态
代理治理何时有效：GAMPO框架与能力门控证据
原标题：When Agent Governance Helps
AI 摘要
该研究提出了GAMPO框架，基于321篇文献综合，用于设计受治理的自生成目标AI代理组织。在CHI-Bench医疗基准上的实验表明，治理收益受模型剩余能力限制：弱模型上完整程序无效甚至有害，而一句“验证写入”提示使pass@1翻倍；前沿模型上，基于案例的答案盲完成定义比通用程序更有效，将预授权成功率从40%提升至68%（单次）或84%（五次自洽）。研究为探索性，样本量小，结果需谨慎解读。
以上摘要由 AI 生成，可能存在误差。事实请以原文为准。
正文节选
When Agent Governance Helps Abstract No specification says how a governed autotelic AI agent organization, where agents pursue self-generated goals inside guardrails, should be designed and evaluated. We answer in two parts. First, we synthesize the Governed Autotelic Multi-Agent Product Organization (GAMPO) framework from a document-based qualitative evidence synthesis of 321 sources, integrating agency, agile, platform, and governance theory into a runnable specification. Second, we probe a pr
发布时间：2026-09-09 12:00
抓取时间：2026-09-09 12:39
来源机构：arXiv