新闻
SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration
arXiv cs.AI · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- SearchOS-V1 is a multi-agent framework that tracks search progress explicitly to help AI agents find and cite information without getting stuck in repetitive loops.
- 为何重要
- Matters for engineers building information-seeking systems where multiple agents collaborate on web search and need to avoid wasted queries and incomplete results.
- 注意
- Paper is recent academic work; real-world robustness beyond benchmark datasets WideSearch and GISA remains to be demonstrated in production environments.
收听本摘要
文章节选
-->
Computer Science > Artificial Intelligence
arXiv:2607.15257v1 (cs)
[Submitted on 16 Jul 2026]
Title: SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration
Authors: Yuyao Zhang , Junjie Gao , Zhengxian Wu , Jiaming Fan , Jin Zhang , Shihan Ma , Yao Yao , Weiran Qi , Chuyan Jin , Guiyu Ma , Xingzhong Xu , Kai Yang , Ji-Rong Wen , Zhicheng Dou
View a PDF of the paper titled SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration, by Yuyao Zhang and 13 other authors
View PDF HTML (experimental)
Abstract: Recent advances in Tool-Integrated Large Language Models have made web search a core capability of information-seeking agents. However, as interaction histories grow, agents increasingly struggle to track task progress. When search attempts fail to yield useful evidence, current single- and multi-agent systems can become trapped in repetitive loops, wasting search budgets and ultimately compromising the quality and completeness of the final output. We introduce SearchOS, a system-level multi-agent framework that turns fragile, implicit search progress into explicit, persistent, and shared state. First, we formulate open-domain information seeking as relational schema completion with grounded citations, where agents discover entities, populate attributes across linked tables, and anchor each value to source evidence. Then
节选自原文。请前往来源阅读全文。
- agent
- language model
- rag
- multi-agent
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。