arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.08400cs.AI
收藏

SCOUT-RAG:可扩展且成本效益高的代理图式检索框架用于分布式领域

SCOUT-RAG: Scalable and Cost-Efficient Unifying Traversal for Agentic Graph-RAG over Distributed Domains

A*STAR

Longkun Li, Yuanben Zou, Jinghan Wu, Yuqing Wen, Jing Li, Hangwei Qian, Ivor Tsang

AI总结:

SCOUT-RAG是一种分布式代理图式检索框架,通过跨领域检索和增量效用目标提升推理性能,同时降低成本和延迟。

摘要:

图式检索(Graph-RAG)通过结构化知识提升大语言模型的推理能力,但传统设计依赖于集中式知识图谱。在分布式和访问受限的环境中(例如医院或跨国组织),检索必须选择相关领域并确定适当的遍历深度,而无需全局图谱可见性或全面查询。为解决这一挑战,我们引入了SCOUT-RAG(可扩展且成本效益高的统一遍历),一种分布式代理图式检索框架,通过逐步跨领域检索并指导增量效用目标。SCOUT-RAG采用四个协作代理:(i)估计领域相关性;(ii)决定何时扩展检索到其他领域;(iii)调整遍历深度以避免不必要的图谱探索;(iv)合成高质量答案。该框架旨在最小化检索遗憾,即遗漏有用领域信息,同时控制延迟和API成本。在多领域知识设置中,SCOUT-RAG的性能与集中式基线(如DRIFT和全面领域遍历)相当,同时显著减少了跨领域调用、总处理令牌数和延迟。

英文摘要:

Graph-RAG improves LLM reasoning using structured knowledge, yet conventional designs rely on a centralized knowledge graph. In distributed and access-restricted settings (e.g., hospitals or multinational organizations), retrieval must select relevant domains and appropriate traversal depth without global graph visibility or exhaustive querying. To address this challenge, we introduce \textbf{SCOUT-RAG} (\textit{\underline{S}calable and \underline{CO}st-efficient \underline{U}nifying \underline{T}raversal}), a distributed agentic Graph-RAG framework that performs progressive cross-domain retrieval guided by incremental utility goals. SCOUT-RAG employs four cooperative agents that: (i) estimate domain relevance, (ii) decide when to expand retrieval to additional domains, (iii) adapt traversal depth to avoid unnecessary graph exploration, and (iv) synthesize the high-quality answers. The framework is designed to minimize retrieval regret, defined as missing useful domain information, while controlling latency and API cost. Across multi-domain knowledge settings, SCOUT-RAG achieves performance comparable to centralized baselines, including DRIFT and exhaustive domain traversal, while substantially reducing cross-domain calls, total tokens processed, and latency.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑