2026-10-06T16:10:49Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33974086 2026-10-04T04:32:36Z category_28843 item_type_3 month_year_10_2026

LongHisDoc: A Comprehensive Benchmark for Chinese Long Historical Document Understanding Zeyu Shan (25097293) Knowledge representation and reasoning LLM OCR Historical Document VQA <p dir="ltr">In this repo, we present <b>LongHisDoc</b>, a pioneering benchmark specifically designed to evaluate the capabilities of LLMs and LVLMs in long-context historical document understanding tasks. This benchmark includes 101 historical documents across 10 categories, with 1,012 expert-annotated question-answer pairs covering four types, and the evidence for the questions is drawn from three modalities. Here, we provide all page screenshots of 101 PDF-formatted historical documents (./LongHisDoc_IMG), their OCR results (./OCR_res), and QA pairs (./LongHisDoc.json).</p> 2026-10-04T04:32:36Z Dataset Dataset 10.6084/m9.figshare.33974086.v2 https://figshare.com/articles/dataset/LongHisDoc_A_Comprehensive_Benchmark_for_Chinese_Long_Historical_Document_Understanding/33974086 CC BY 4.0