Multimodal LLMs enable low-cost digitization of heterogeneous historical tables by leveraging domain expertise via prompts, enabling cohesive panel data from diverse sources . Using gold-standard data for evaluation, the pipeline achieves 95.4% exact cell matching at roughly 50x lower costs than outsourcing . It also dramatically reduces layout parsing errors (61.4% to 0.35%), and attains 96.7% exact transcription of linked cells with a mean absolute percentage error of 0.7% . In replication studies, the LLM pipeline aligns with human-based header categorization and yields econometric inferences that are largely indistinguishable from gold-standard results across eight models .
一、元数据与概览
二、逐节深度解读
三、图表与表格的深度解读(与文本的联系、数据点与潜在局限)
图1a 与 Figure 1b(管线架构与迭代改进循环)
图2(Performance Detail by Time and Geography)
图3(County-Level Vehicle Adoption Rates by Year, 1910–1961)
表格层面的要点(Table 1–Table 5;A.1–A.5.1 等附录)
四、估值分析(成本与效益评估)
五、风险因素评估
六、批判性视角与细微差别(审慎视角)
七、结论性综合(综合性观点与启示)
附注:图像与关键图表的溯源嵌入
总体而言,此研究为历史数据科研提供了一条可行且可复制的路径:在高异质性、低标准化的历史表格场景中,结合多模态LLMs、领域知识驱动的提示设计、以及严格的金标准端到端评估,可以以更低成本、高质量地构建具有可用于经济分析的面板数据集。其方法论的核心在于以域知识驱动的迭代改进、固定步骤的工作流设计,以及对下游推断的一致性进行实证验证。这些要素共同构成了一个在学术研究与数据生产之间的稳健桥梁。 , ,
2505.11599.pdf