M3DocDep：基于大型视觉语言模型的多模态、多页、多文档依赖分块
M3DocDep: Multi-modal, Multi-page, Multi-document Dependency Chunking with Large Vision-Language Models
对做文档视觉理解的人，这篇给出了一个可复用的思路：把 OCR/版面解析结果统一到文档级坐标系，用边界感知的 SoftROI 做多模态块嵌入，再用 biaffine 打分重建跨页文档树——这套「视觉结构线索 + 全局依赖」的建模方式可直接迁移到多页文档检索、版面层级解析等任务。
原文依据 existing text-centric chunkers overlook the visual and structural cues present in real-world documents
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论在包含扫描图像、复杂版面和多个页面的大规模工业文档中，检索增强生成（RAG）的效果高度依赖分块质量。然而，现有以文本为中心的分块器忽略了真实文档中存在的视觉与结构线索，导致分块冗余或歧义，从而损害检索与答案准确率。为解决该问题，我们提出 M3DocDep，它集成了：(i) SharedDet，用于将文档解析与 OCR 输出归一化到文档级坐标系；(ii) 带边界感知 SoftROI 的多模态块嵌入；(iii) 通过双仿射打分进行全局文档树重建；(iv) 结构感知的依赖分块，在保留边界的同时减少冗余。M3DocDep 在文档层级解析（DHP）与语料级 RAG 评测上均取得一致提升：STEDS 提升 +28.5--39.6%，检索 nDCG 提升 +1.1--15.3%，问答 ANLS 提升 +4.5--15.3%。这些结果表明，以多模态、结构感知的分块对文档级依赖进行建模，能够提升 RAG 在长篇、多页工业文档上的性能。
英文原文摘要 1146 字符
In large-scale industrial documents with scanned images, complex layouts, and multiple pages, the effectiveness of retrieval-augmented generation (RAG) is highly dependent on chunking quality. However, existing text-centric chunkers overlook the visual and structural cues present in real-world documents, leading to redundant or ambiguous chunks that impair retrieval and answer accuracy. To address this problem, we propose \ours which integrates (i) SharedDet for normalizing document parsing and OCR outputs into a document-level frame, (ii) Multi-modal block embeddings with boundary-aware SoftROI, (iii) global document-tree reconstruction via biaffine scoring, and (iv) structure-aware dependency chunking that preserves boundaries and reduces redundancy. \ours achieves consistent gains across both Document Hierarchical Parsing (DHP) and corpus-level RAG evaluations, improving STEDS by +28.5--39.6%, retrieval nDCG by +1.1--15.3%, and QA ANLS by +4.5--15.3%. These results demonstrate that modeling document-level dependencies with Multi-modal, structure-aware chunking improves RAG performance on long, multi-page industrial documents.