Hugging Face publication checklist

Published publicly on 2026-08-17 at:

https://huggingface.co/factlogic/phoenix-arabic-manuscript-htr

The public release uses the conservative CC BY-NC-SA 2.0 license because Muharaf contributed to training. The checks below remain the release audit trail.

Before publishing

Confirm the repository name and owner.
Preserve the CC BY-NC-SA 2.0 license and the non-commercial training-data notice unless a new legal review authorizes a change.
Verify SHA256SUMS against model.mlmodel.
Confirm that the README has no placeholder comments.
Confirm that aggregate benchmark JSON contains no line-level ground truth or restricted images.
Do not upload the Athar retrieval corpus or raw training datasets with the model.
Keep the wording distinction between sealed release evidence and development-guard diagnostics.
Do not claim comparison with the published Muharaf 18.1% number unless its official protocol is reproduced exactly.
Keep the Phoenix-vs-Baseer section labelled as an owner-run server benchmark until raw predictions, sample counts, preprocessing, and an executable evaluation manifest are archived.
Preserve the counter-results: Baseer__Nakba wins Omar CER and exact-line accuracy; do not replace the scoped conclusion with “Phoenix is universally better.”

Suggested commands

$hf = "$env:LOCALAPPDATA\Programs\Python\Python313\Scripts\hf.exe"
& $hf auth whoami
& $hf repos create factlogic/phoenix-arabic-manuscript-htr --type model --private
& $hf upload factlogic/phoenix-arabic-manuscript-htr . . --repo-type model

Run the upload command from this directory. The repository is already public; do not change the license metadata without reviewing upstream dataset obligations.