datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
chunkr-reading-order-bench-oss
Chunkr Reading Order Bench - Open Source Subset
Open-source subset of the Chunkr Reading Order benchmark dataset, containing 733 professionally annotated documents with detailed reading order annotations across diverse document layouts.
This dataset benchmarks reading order detection models on complex, real-world documents including financial reports, legal contracts, research papers, medical records, and more. Each document includes ground truth reading order sequences essential… See the full description on the dataset page: https://huggingface.co/datasets/ChunkrAI/chunkr-reading-order-bench-oss.NayanaBench-with-reading-orderNayanaBench-with-reading-order-deepseek-ocr-enhistorical-reading-order-benchmark
Detection is easy, reading order is the tax
A multilingual benchmark of line detection and reading order on historical documents
Author: Abhishek Jha (GitHub · Hugging Face)
Code & harness (canonical): https://github.com/AbhiPandit1/historical-reading-order-benchmark
What this is, in one paragraph
On historical documents, finding the text lines on a page is close to solved.
Deciding the order in which to read them is not. On a single column the two… See the full description on the dataset page: https://huggingface.co/datasets/abhishekjha1008/historical-reading-order-benchmark.