Reproduce “ATHAAR: A Medieval Scientific Arabic Manuscript Dataset for Handwritten Text Recognition” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Ismail El Jamiy, Khadija Msalek, Guillaume Loizelet, Farah Benamara, Nabil Hathout

ATHAAR introduces a dataset and transcription guide for handwritten medieval scientific Arabic, with aligned copies that support tests of recognition across witnesses, chapters, and treatises. The guide records what scribes wrote rather than correcting it. In a cross-chapter test with three examples, Gemini 3 Flash achieved 18.9% character error rate, compared with 59.3% for CHURRO; the test does not establish fine-tuning performance.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.