跳到正文
arXiv:cs.AI· Zhexi Feng·· 22 小时前

Fan Out 还是 Relay?面向 LLM 智能体查询孤岛表的 Ownership-Aware Routing

Fan Out or Relay? Ownership-Aware Routing for LLM Agents Querying Siloed Tables

AI 导读

针对 LLM 协调者查询多个数据所有者私有表的场景,研究指出 fan-out 与 relay 两种策略孰优取决于数据归属:在 PeopleJoin 和 BIRD 的十三个设置中,fan-out 在单一所有者可回答的问题上于十二个设置更准确,但在涉及两位所有者表连接时最多只答对 4.3%。

正文

View PDF HTML (experimental)

Abstract:Inside organizations, a question's data often sit with several owners, who query their own tables but never hand them over. An LLM coordinator can fan out, sending the question to every owner at once, or relay, querying owners in turn and reusing earlier rows. We show that neither is safe in advance, because data ownership decides which one is right. Across thirteen settings on PeopleJoin and BIRD with three model families, fan-out is more accurate in twelve on questions one owner can answer, yet answers at most 4.3% of those joining two owners' tables. The flip follows ownership, not query shape: on BIRD two-table joins, fan-out answers 32-44% when one owner holds both and 1-4% otherwise. Relay wins by conditioning on returned rows, and hiding them from an otherwise identical planner cuts two-owner accuracy from 23.2% to 5.1%. Since ownership is public, we propose Ownership-Aware Routing (OAR), which drafts a schema-only SQL query, counts the owners whose tables it joins, and routes the question before any private row is read. In ten settings OAR is never significantly less accurate than the best fixed protocol, and beats every fixed protocol on PeopleJoin, by up to 9.3 points on held-out organizations.
Comments: 19 pages, 4 figures, 14 tables
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Multiagent Systems (cs.MA)
ACM classes: I.2.7; I.2.11; H.2.4
Cite as: arXiv:2610.18157 [cs.CL]
  (or arXiv:2610.18157v1 [cs.CL] for this version)
  https://doi.org/10.48550/arXiv.2610.18157

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Zhexi Feng [view email]
[v1] Tue, 13 Oct 2026 11:01:37 UTC (207 KB)

来源:arXiv:cs.AI · arxiv.org