AI 中文总结
该研究针对长文档检索的结构缺陷，提出无嵌入智能体检索框架READ，在政府财务报告数据集上显著优于密集检索，且性能提升源于操作界面，同时发现BM25与READ无统计差异。
AI 中文摘要
针对长文档的检索增强生成目前主流设计为：将文本分块、对分块进行嵌入，然后返回与查询最接近的Top-K个邻居。我们认为，对于财务报表、审计报告、监管申报表这类重要文档，该设计在结构上存在缺陷，且我们将该论点量化验证。在一份780页的政府财务报告中，86.8%的内容行是表格行，数千个几乎相同的数值在同一嵌入空间中竞争，且一个数值的单位通常继承自其上方中位数13行的表头——因此分块边界常会将一个数值与其所属的“十万卢比（lakh）”或“千万卢比（crore）”单位分隔，产生两个数量级的错误。我们构建了一个作为钢人（steelman）的感知表格分块器，该分块器解决了单位问题，但在我们尝试的所有分块大小下，仍有27%-30%的数值分块没有财年表头。我们提出READ（Reliable Embedding-free Agentic Document-search，可靠的无嵌入智能体文档搜索），其中智能体通过三种确定性操作读取原始文档：归一化词汇搜索、结构导航和限定范围的读取，这些操作通过模型上下文协议（Model Context Protocol）对外暴露，因此其轨迹是可复现的审计追踪，而非不透明的相似度分数。在51个经过验证的问题上，READ的回答准确率为58.8%，而密集检索的准确率仅为15.7%（p_Holm=2×10^-5）；即使对READ进行调优，其准确率为35.3%，仍比密集检索高出23.5个百分点（p_Holm=0.017）。使用相同循环但搭配Top-K工具的智能体仅达到27.5%的准确率，说明性能提升源于操作界面而非迭代方式。我们还报告了证据不支持的结论：BM25与READ的性能在统计上无显著差异，因此我们的结果区分了基于嵌入的检索与无嵌入检索，而非智能体与词汇搜索。
英文摘要
Retrieval-augmented generation over long documents is dominated by one design: chunk the text, embed the chunks, and surface the top-k nearest neighbours of the query. We argue that for an important class of documents -- financial statements, audit reports, regulatory returns -- this design is structurally unsound, and we make the argument measurable. On a 780-page government financial report, 86.8% of content lines are table rows, thousands of near-identical figures compete in one embedding space, and a figure inherits its unit from a header a median of 13 lines above it -- so a chunk boundary routinely separates a number from whether it is in lakh or crore, an error of two orders of magnitude. A table-aware chunker built as a steelman fixes the unit problem but leaves 27-30% of numeric chunks with no fiscal-year header at every chunk size we tried. We propose READ (Reliable Embedding-free Agentic Document-search), in which an agent reads the raw document through three deterministic operations -- normalized lexical search, structural navigation, and bounded span reads -- exposed over the Model Context Protocol, so a trajectory is a replayable audit trail, not an opaque similarity score. On 51 verified questions READ answers 58.8% against dense retrieval's 15.7% (p_Holm = 2 x 10^-5) -- or 35.3% tuned, which READ still leads by 23.5 points (p_Holm = 0.017). An agent given the same loop but a top-k tool reaches only 27.5%, locating the gain in the interface rather than in iteration. We also report what the evidence does not support: BM25 is statistically indistinguishable from READ, so our result separates embedding-based from embedding-free retrieval, not agentic from lexical search.
Comments20 pages, 5 figures, 14 tables. Code, benchmark, and full result trajectories: https://github.com/twospoon/READ
