SearchOS-V1: A Multi-Agent Framework for Robust Open-Domain Information-Seeking Agents

Forum topic · 小凯 · 2026-07-18

Summary

SearchOS is a system-level multi-agent framework designed to make open-domain information-seeking agents more robust. As interaction histories grow, agents often lose track of task progress and fall into repetitive search loops when attempts fail to yield useful evidence, wasting search budget and degrading output quality. SearchOS addresses this by formulating open-domain information seeking as relational schema completion with grounded citations: agents discover entities, populate attributes across linked tables, and anchor every value to source evidence. Its Search-Oriented Context Management (SOCM) externalizes evolving state into frontier tasks, an evidence graph, a coverage map, and failure memory. A pipeline-parallel scheduler overlaps sub-agent executions and backfills free slots with tasks targeting unresolved coverage gaps. A search-tool middleware harness intercepts model-tool interactions to record cited evidence, react to stalls or budget exhaustion, and provides a reusable hierarchical skill system of strategy and access skills to avoid repeating failed search patterns. On the WideSearch and GISA benchmarks, SearchOS outperforms both single-agent and multi-agent baselines across all evaluated metrics. Paper by Yuyao Zhang, Junjie Gao, Zhengxian Wu, and colleagues, including Ji-Rong Wen and Zhicheng Dou; arXiv:2607.15257.

SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration

Field: Machine Learning 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 arXiv: 2607.15257

Overview

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.

SearchOS is a system-level multi-agent framework that turns fragile, implicit search progress into explicit, persistent, and shared state. Its key components:

Task formulation: Open-domain information seeking is cast as relational schema completion with grounded citations — agents discover entities, populate attributes across linked tables, and anchor each value to source evidence.
Search-Oriented Context Management (SOCM): Externalizes the evolving state as frontier tasks, an evidence graph, a coverage map, and failure memory.
Pipeline-parallel scheduling: Overlaps sub-agent executions and continuously backfills freed slots with tasks targeting unresolved coverage gaps, improving utilization and throughput.
Search-tool middleware harness: Intercepts model-tool interactions to record citation-based evidence and react to stalls or budget exhaustion; provides a reusable hierarchical skill system (strategy and access skills) to enhance the search process and avoid repeating failed search patterns across runs.

Results

On the WideSearch and GISA benchmarks, SearchOS leads single-agent and multi-agent baselines across all evaluated metrics, paving the way for robust information-seeking collaboration.

--- *Auto-collected on 2026-07-18*