长文档的检索增强生成主要由一种设计: 块、文本, 嵌入块, 并显示查询的前k 个最近邻居组成。我们认为，对于一类重要的文件——财务报表,审计报告,监管回报——这种设计在结构上是不健全的,，我们使论点变得可衡量。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字在一个嵌入空间中竞争,，并且数字从其上方 13 行中位数的标题继承其单位 - 因此块边界通常将数字与十万或千万, 两个数量级的误差分开。作为钢铁侠构建的表感知分块器解决了单位问题，但在我们尝试的每个块大小中留下了 27-30% 的数字块，没有财政年度标题。我们建议 READ (Reliable Embedding-free Agentic Document-search),，其中代理通过三个确定性操作读取原始文档 - 标准化词汇搜索, 结构导航, 和有界跨度读取 - 通过模型上下文协议, 公开，因此轨迹是可重播的审计跟踪, 而不是不透明的相似性分数。在 51 个经过验证的问题上，READ 的答案为 58.8%，反对密集检索的 15.7% (p_Holm = 2 x 10^-5) - 或 35.3% 调整,，其中 READ 仍领先 23.5 分 (p_Holm = 0.017)。给定相同循环但使用 top-k 工具的代理仅达到 27.5%,，将增益定位在界面而不是迭代中。我们还报告了证据不支持的内容: BM25 在统计上与 READ, 无法区分，因此我们的结果将基于嵌入的检索与无嵌入的检索分开, 与词汇搜索无关。
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的 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.
科目: 人工智能 (cs.AI); 计算和语言 (cs.CL); 信息检索 (cs.IR)
Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Information Retrieval (cs.IR)