corvicai/FinTabNet_ComTQA

该数据集包含来自财富500强公司财务收益报告的大规模表格集合。它源自ComTQA数据集，而ComTQA数据集又是基于IBM研究院发布的原始FinTabNet数据集。此版本经过修改，使每个问题特定于单个文档。这种改变将数据集转变为一个针对不同PDFs的大型语料库进行问题回答的基准，旨在评估系统首先检索正确文档，然后提取和推理其内容的能力。

This dataset contains a large-scale collection of tables from the financial earnings reports of Fortune 500 companies. It is derived from the ComTQA dataset, which in turn is based on the original FinTabNet dataset released by IBM Research. This version has been modified to make each question specific to a single document, transforming the dataset into a benchmark for question answering over a large corpus of different PDFs, designed to evaluate a systems ability to first retrieve the correct document and then extract and reason over its contents.