HotCRP, EasyChair, OpenReview, CMT

Your paper does not go to the conference. It goes to a submission system: a website where you upload the PDF, fill in metadata, and later read your reviews. There are four you will meet again and again, and each has a personality.

The big four

System	Personality	You will meet it at
HotCRP	Fast, minimal, beloved by systems folks	Systems and security conferences (SOSP, USENIX)
EasyChair	The veteran. Utilitarian, everywhere	Workshops, many CS conferences worldwide
OpenReview	Public-facing, reviews often visible	ML venues (ICLR, NeurIPS)
CMT	Microsoft’s workhorse, form-heavy	Vision and ML conferences (CVPR)

They all do the same core job. The differences are in the details: OpenReview may show your reviews to the world, HotCRP will email you about every comment, CMT has a topic-selection grid that takes ten minutes to click through.

Make your account early

Do this the week you decide to submit, not deadline night:

• Create the account with the email you check daily
• Confirm the verification email actually arrives
• Check that co-authors have accounts too (OpenReview profiles can take days to approve for new users)
• Start a draft submission so you know what fields exist

That last one matters. Some venues ask surprise questions: ethics statements, reproducibility checklists, conflict domains. Better to discover them a week out.

PDF-only vs source upload

Most venues want a PDF at submission time. Some, and almost all publishers at camera-ready time, want your LaTeX source too. Keep your project tidy enough to zip up on demand. Oleafly exports a source ZIP directly, which is exactly the shape submission systems expect.

The metadata fields

Beyond the PDF, expect to fill in:

• Title and abstract, as plain text. Rewrite any LaTeX math in words or Unicode.
• Topics, checkboxes that route your paper to reviewers. Pick honestly.
• Conflicts, people and institutions who should not review you: advisors, recent co-authors, colleagues. Getting this wrong can void a review, so be thorough.
• Anonymity confirmation, at double-blind venues. See anonymization for double-blind review.

The 11:59 AoE reality

Every deadline night, thousands of people submit in the final hour, and the systems get slow. Uploads that took seconds at lunch take minutes at 11:50 PM. The fix is simple and nobody follows it: submit a working version early.

You can update until the deadline

Here is the liberating part: submission is not one shot. On every major system you can replace your PDF as many times as you like before the deadline closes. So submit a complete-but-imperfect version the day before, and keep polishing. If the network dies at 11:58, you have a real submission in the system instead of a tragedy.

Next stop after submission: understanding what happens on the other side, in peer review, explained.