基于嵌入向量的检索加 LLM 过滤，在 7,354 项随机试验中找回 92% 证据

Nothing Left Behind: Finding Evidence from a Large Randomized Evaluation Corpus

AI 导读

一项对比研究显示，在 7,354 项随机试验构成的语料中，传统标签检索只能找到系统综述应纳入证据的 45%，而先用文本嵌入按语义召回候选、再由语言模型筛出相关条目的方法找到 92%。该方法的单次查询成本随搜索深度上升，结果支持将广覆盖候选召回与查询特定证据过滤结合的两阶段架构。

正文 · 原文

View PDF HTML (experimental)

Abstract:Many AI systems need to find every relevant items in a large collection and then organize what they found. The standard way to do this is to tag each item with labels from a curated vocabulary and search on those labels. Those labels are useful for organizing results, but searching on them misses items that do not fit a label and returns too much when labels are broad. We compare this approach with an embedding-based approach: find candidates by meaning using text embeddings, then have a language model keep only the relevant ones. In a corpus of 7,354 randomized trials, label search found 45% of the evidence a systematic review would include, while the new approach found 92%, at a cost per query that rises with search depth. These results support a two-stage architecture that combines broad candidate retrieval with query-specific evidence filtering.

Comments:	7 pages, 3 figures
Subjects:	Computation and Language (cs.CL)
Cite as:	arXiv:2610.36910 [cs.CL]
	(or arXiv:2610.36910v1 [cs.CL] for this version)
	https://doi.org/10.48550/arXiv.2610.36910 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Giuliano Martinelli [view email]
[v1] Tue, 27 Oct 2026 09:40:24 UTC (1,420 KB)

来源：arXiv cs.CL · arxiv.org