面向真实世界文档解析：逼真场景合成与文档感知训练
Towards Real-World Document Parsing via Realistic Scene Synthesis and Document-Aware Training
对做文档/图表类视觉理解的人，这篇提供了可复用的思路：用布局模板+文档元素合成大规模全页端到端监督数据，配合渐进式学习与结构 token 优化来抑制 MLLM 的重复与幻觉输出；Wild-OmniDocBench 也可直接用于评测随手拍文档的鲁棒性。
原文依据 A Realistic Scene Synthesis strategy constructs large-scale, structurally diverse full-page end-to-end supervision
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论文档解析近来随着多模态大语言模型（MLLM）的发展而进步，这类模型可直接将文档图像映射为结构化输出。传统的级联式流程依赖精确的版面分析，在随手拍摄或非标准条件下往往失效。尽管端到端方法缓解了这一依赖，它们仍会出现重复、幻觉以及结构不一致的预测——主要原因是缺乏大规模、高质量的全页（文档级）端到端解析数据，以及缺少结构感知的训练策略。为应对这些挑战，我们提出一个数据与训练协同设计的框架，以实现鲁棒的端到端文档解析。逼真场景合成策略通过将版面模板与丰富的文档元素组合，构建大规模、结构多样的全页端到端监督；文档感知训练方案引入渐进式学习与结构 token 优化，以提升结构保真度与解码稳定性。我们进一步构建了 Wild-OmniDocBench，这是一个从真实拍摄文档中导出的基准，用于鲁棒性评测。将该方法集成到一个 10 亿参数的多模态大语言模型中，我们的方法在扫描/数字文档与真实拍摄场景下均取得更优的准确率与鲁棒性。所有模型、数据合成流程与基准都将公开释放，以推动文档理解的未来研究。
英文原文摘要 1434 字符
Document parsing has recently advanced with multimodal large language models (MLLMs) that directly map document images to structured outputs. Traditional cascaded pipelines depend on precise layout analysis and often fail under casually captured or non-standard conditions. Although end-to-end approaches mitigate this dependency, they still exhibit repetitive, hallucinated, and structurally inconsistent predictions--primarily due to the scarcity of large-scale, high-quality full-page (document-level) end-to-end parsing data and the lack of structure-aware training strategies. To address these challenges, we propose a data-training co-design framework for robust end-to-end document parsing. A Realistic Scene Synthesis strategy constructs large-scale, structurally diverse full-page end-to-end supervision by composing layout templates with rich document elements, while a Document-Aware Training Recipe introduces progressive learning and structure-token optimization to enhance structural fidelity and decoding stability. We further build Wild-OmniDocBench, a benchmark derived from real-world captured documents for robustness evaluation. Integrated into a 1B-parameter MLLM, our method achieves superior accuracy and robustness across both scanned/digital and real-world captured scenarios. All models, data synthesis pipelines, and benchmarks will be publicly released to advance future research in document understanding.