brycewang-stanford FOCS Submission
Anchor cycle: FOCS 2026 (67th edition), submissions due Wednesday, April 1,
2026, 5:00 PM EDT — a US-Eastern wall clock equal to 21:00 UTC, not an
anywhere-on-earth deadline — via focs26.hotcrp.com (CFP checked 2026-07-08;
that deadline has passed, so this audit now serves the next cycle: reopen the
live CFP before trusting any number here).
What the 2026 CFP actually required
| Requirement | 2026 wording (verify per cycle) |
|---|---|
| Length | No bound; full version encouraged as the submission |
| Attention rule | Abstract + references + first ten pages guaranteed; the rest at the committee's discretion |
| Typography | 11-point or larger font, single column, single-spaced with ample spacing, ≥1-inch margins, letter-size paper |
| File | PDF without security restrictions on copying or printing |
| Anonymity | Double-blind: no names/affiliations/emails, third-person self-references, acknowledgements omitted |
| Publication policy | SIGACT rules: nothing published or scheduled before the conference month elsewhere; no simultaneous proceedings/journal submission |
| Exceptions | Prior appearance in a venue with significantly different format/content/audience: contact the program chair before submitting |
Note the guaranteed-read set: FOCS names the references as always-read
material alongside the ten pages. Your bibliography is a reviewed object —
stale, sloppy, or misattributed citations are visible even to the laziest
reading (see focs-related-work).
Mechanical pre-flight
Run these on the exact PDF you will upload, not on a draft:
# Page geometry and count of the guaranteed window
pdfinfo submission.pdf | grep -E 'Pages|Page size' # letter = 612 x 792 pts
# FOCS-specific: the CFP forbids copy/print security restrictions
qpdf --show-encryption submission.pdf # want: "File is not encrypted"
# Identity leakage in embedded metadata
exiftool submission.pdf | grep -iE 'author|creator|company'
# Self-reference sweep in the source
grep -rn --include='*.tex' -iE '\b(we|our) (recent|previous|earlier|prior)\b|acknowledg|thank' .
The encryption check is easy to fail accidentally: some institutional PDF tooling and some print-to-PDF drivers stamp permission flags by default, and a restricted PDF violates the letter of the CFP.
Deadline-week sequence
- T-7 days — content freeze. No new theorems. Anything proved this week goes to the arXiv version later, not into the submission.
- T-5 days — window audit. Confirm every result is stated by page ten and the technical overview reflects the proofs as they now exist.
- T-3 days — first upload. HotCRP allows re-uploads until the deadline; an early upload converts a hard cliff into a soft one. Fill conflicts, topics, and co-author records now — PC conflict data affects assignment quality directly.
- T-1 day — blind and format pass using the commands above, on the compiled final PDF.
- Deadline day — clock arithmetic. 5 PM EDT is late evening in Europe and early morning in Asia; compute your local equivalent once, write it down, and stop trusting mental arithmetic under stress.
HotCRP metadata is part of the submission
Theory authors treat the web form as an afterthought; the committee's machinery does not. Three fields shape your paper's fate before anyone reads a word:
- Topics. Assignment starts from your topic selections. Choosing only your micro-area starves the paper of breadth readers; choosing everything looks unserious. Pick the two or three areas whose reviewers you would actually want in the room.
- Conflicts. Declare advisors, recent co-authors, and current-institution colleagues per the form's definitions. Under-declared conflicts discovered later can void a review cycle; over-declared conflicts (tagging everyone senior in the area) can be read as reviewer-shopping.
- Author records. Complete them accurately even though the PDF is blind — the system uses them for conflict resolution, and errors are painful to repair after decisions.
Policy traps specific to this venue
- The IEEE/SIGACT seam. FOCS is an IEEE Computer Society (TCMF) event that enforces ACM SIGACT's publication policy. Authors who read only IEEE's generic submission rules miss the simultaneous-submission bar entirely.
- arXiv is safe, proceedings are not. Posting the full version to arXiv or ECCC before submitting is normal community practice and does not violate the prior-publication rule; a workshop with published proceedings does.
- The invited-journal overlap. If a journal version of a companion result is already in press with overlapping material, disclose to the chair before the deadline rather than hoping the overlap goes unnoticed.
Readiness verdict
[FOCS submission audit]
Window : all theorems stated by p.10 PASS/FAIL
Format : 11pt+/single-column/1in/letter PASS/FAIL
PDF : unencrypted, metadata clean PASS/FAIL
Blind : third-person self-refs, no acks PASS/FAIL
Policy : no simultaneous/prior publication PASS/FAIL
HotCRP : conflicts, topics, authors complete PASS/FAIL
Verdict : UPLOAD / FIX(ordered list) / CONSULT CHAIR
Any FAIL in the first three rows is fixable in hours; a policy FAIL is not fixable at all after upload — resolve those with the program chair first.