HelpBlind review

What to strip for double-blind review

Updated 2026-09

Double-blind is more than a nameless title page. A PDF, a GitHub URL, or “in our prior work” can deanonymize you.

Double-blind vs single-blind

Double-blind: reviewers do not know authors (editors do). NeurIPS, ICML, CVPR, and some Chinese journals use it.

Single-blind: reviewers know authors. Many Chinese journals do; you may still suggest exclusions.

Pre-submit cleanup list

Check every item. One leak can look intentional.

  • Strip names, headers, acknowledgments, and grant IDs for review
  • Do not put author names in the filename
  • Clean supplements, video credits, and license headers
  • Use anonymous code mirrors—or omit identifying links
  • Cite yourself in the third person; put concurrent unpublished work in an anonymized supplement

What about arXiv?

Most ML/vision flagships allow arXiv, but the review PDF must stay anonymous and you must not advertise “under review at X” during review.

A reviewer stumbling on a preprint is usually not an automatic conflict; still do not point the review PDF at the non-anonymous version. Read each CFP.

FAQ

When do acknowledgments return?
Usually at camera-ready. Keep them out of the review PDF.
Chinese journals vs conferences?
Same idea; some Chinese journals also police filenames and signed pledges. Follow the live instructions.

Layouts

Related