Skip to content

Dataset · allenai

OCR bench for PDF to Markdown

allenai/olmOCR-bench

1,403 PDFs and 7,010 unit tests to evaluate OCR systems on converting PDFs to Markdown while preserving structure.

What's inside

olmOCR-bench contains 1,403 PDF files and 7,010 unit tests reflecting the properties of high-quality OCR output. The benchmark evaluates how accurately a system converts PDFs to Markdown while preserving text and structure. Language is English, modalities are document and text.

What you can do with it

  • Evaluate OCR systems on PDF-to-Markdown conversion
  • Compare document recognition pipelines
  • Verify structure preservation in OCR output
Languages
English
Size
1K–10K rows
Downloads this month
17.9K
Likes
335

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("allenai/olmOCR-bench")