文档结构是否有助于稠密检索？两个语料库上四种机制的安慰剂对照消融研究
Does Document Structure Help Dense Retrieval? A Placebo-Controlled Ablation of Four Mechanisms Across Two Corpora
浏览论文内容
中文总结 AI 辅助
通过安慰剂对照消融实验，验证文档结构处理对稠密检索的贡献，发现结构对齐分块和真实标题路径有效，而朴素分层检索有害，效应虽小但跨语料库稳健。
中文摘要 AI 辅助
检索增强生成系统越来越依赖文档结构处理：结构对齐的分块、LLM生成的块上下文、标题路径元数据以及分层两阶段检索。已有研究在不同语料库、嵌入器和指标上分别支持每种方法，但均未控制一个共同的混杂因素：任何添加到块前面的文本都会扰动其嵌入。我们提出了一种机制隔离的消融研究，在统一协议下测试所有四种处理，匹配各条件下的块大小，并添加语义上无效的安慰剂——结构有效但在文档间打乱的标题路径。我们使用覆盖率感知的nDCG进行检索评分，并通过文档聚类自助法与Holm校正测试四个预注册对比，在两个相距较远的语料库上进行：200篇维基百科特色文章（951个查询）和1,585篇QASPER论文（4,303个问题）。组织性有帮助，原因是内容而非词元：具有真实标题路径的结构对齐块优于带上下文的固定窗口（+0.022 / +0.012 cov-nDCG@10）和安慰剂（+0.010 / +0.016）。朴素的两阶段分层检索效果较差（-0.033 / -0.015），可追溯到第一阶段章节召回率。在维基百科上，真实结构优于LLM生成的结构，但在QASPER上并非如此。效应量较小（$dz$ 0.06-0.11），但经Holm校正后显著且跨语料库一致。
英文摘要
Retrieval-augmented generation systems increasingly rely on document-structure treatments: structure-aligned chunking, LLM-generated chunk contexts, heading-path metadata, and hierarchical two-stage retrieval. Separate studies support each on different corpora, embedders, and metrics, and none control for a shared confound: any text prepended to a chunk perturbs its embedding. We present a mechanism-isolating ablation testing all four treatments under one protocol, matching chunk sizes across conditions and adding a semantically null placebo---heading paths that are structurally valid but shuffled across documents. We score retrieval with a coverage-aware nDCG and test four pre-registered contrasts via document-clustered bootstrap with Holm correction, on two distant corpora: 200 Wikipedia Featured Articles (951 queries) and 1,585 QASPER papers (4,303 questions). Organization helps, and the cause is content, not tokens: structure-aligned chunks with real heading paths beat contextualized fixed windows (+0.022 / +0.012 cov-nDCG@10) and the placebo (+0.010 / +0.016). Naive two-stage hierarchical retrieval hurts (-0.033 / -0.015), traceable to first-stage section recall. Gold structure beats LLM-induced structure on Wikipedia but not on QASPER. Effects are small ($dz$ 0.06-0.11) but Holm-significant and consistent across corpora.
发表机构
- Center for Research Computing, University of Notre Dame（圣母大学研究计算中心）
机构由 AI 辅助整理，请以论文原文为准。
