Certify：基于 Horn 闭包证书的需求导向搜索，判断证据是否充分
Knowing When the Evidence Is Enough: Requirement-Directed Search with Horn-Closure Certificates
AI 导读
arXiv 论文提出 Certify 方法，将搜索增强模型的证据充分性判断结构化：按问题的需求超图逐项搜索，经实体检查后用 Horn 闭包得出证书，给出最小充分证据或说明缺失需求的弃权。
正文 · 原文
Abstract:After each Web search, a search-augmented model must decide whether its evidence suffices: answer, search again, or abstain. Prevailing systems leave this to holistic model judgement, which can accept a context naming every entity yet lacking one required fact. We instead make sufficiency structural: a per-query requirement hypergraph states which facts jointly yield another fact. This raises two challenges: deriving requirements from the question alone, and linking each requirement to passages about the right entity. Certify therefore decomposes the question, searches per requirement, and accepts links only after entity checks. Horn closure over the links yields a certificate, the minimal sufficient passages and their derivation, or an abstention naming open requirements. Given correct retrieval, requirements and links, the decision is sound, and complete up to the entity check. Pooled over three corpora with unanswerable twins, Certify decides significantly more accurately than search agents, graph retrieval and sufficiency judges in default configurations. On the same evidence, it matches a proprietary model's sufficiency label without an additional model call. Over 21M Wikipedia passages, entity checks add 4 to 13 points of decision accuracy where gold evidence exists. Finally, answers from its certificate are 80% correct, against 57% from all retrieved passages.
| Comments: | 12 pages, 2 figures, 5 tables |
| Subjects: | Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.34684 [cs.IR] |
| (or arXiv:2610.34684v1 [cs.IR] for this version) | |
| https://doi.org/10.48550/arXiv.2610.34684 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Weiwei Sun [view email]
[v1]
Mon, 26 Oct 2026 01:20:18 UTC (87 KB)
来源：arXiv cs.AI · arxiv.org