普林斯顿论文：多开发者专用智能体组合在 GAIA 上达 70% 准确率
New Princeton paper finds that combining specialized agents from many developers can beat any single developer's general agent, with GAIA accuracy reaching 70%.
普林斯顿一篇新论文发现，把来自多个开发者的专用智能体组合起来，可以超过任何单一开发者构建的通用智能体，在 GAIA 上准确率达到 70%。作者用 LLM 智能体模拟在固定预算下构建智能体的开发者群体，比较多开发者系统与单一系统，多开发者系统会把任务拆分并路由给不同开发者的智能体。在相同总开发预算下，单独得分最多 35% 的专用智能体组合后达到 70%，比通用智能体的中位数高 12 个百分点；但在仅需网页搜索的 SealQA 上，组合系统表现更差且成本更高。作者建议让智能体填补共享系统中的空缺，因为单独得分 19% 的智能体把组合系统从 40% 提升到 53%。
正文
当前显示原文，中文全文译文尚未就绪。
New Princeton paper finds that combining specialized agents from many developers can beat any single developer's general agent, with GAIA accuracy reaching 70%.
Today most agentic systems are built and controlled by 1 developer, so each team rebuilds the same basic skills. The authors argue this limits scale.
They simulated populations of developers with LLM agents that build agents under a fixed budget, then compared lone systems with a multi-developer system. The multi-developer system splits each task and routes the parts to different developers' agents.
On GAIA with the same total development budget, specialists that each scored at most 35% alone reached 70% together. That is 12 points above the median general agent, while on web-search-only SealQA the combined system did worse at higher cost.
Build agents to fill gaps in shared systems, since an agent scoring 19% alone lifted a combined system from 40% to 53%.