soda-submission

name	soda-submission
description	Use when preparing or auditing a SODA (ACM-SIAM Symposium on Discrete Algorithms) submission — the no-page-limit full-version norm, the title-page abstract, 11-point single-column letter format, lightweight double-blind rules, HotCRP entry, and the July AoE deadline with its summary-rejection triggers.

SODA Submission

Anchor cycle: SODA 2027, the ACM-SIAM Symposium on Discrete Algorithms in Philadelphia, January 24-27, 2027, jointly sponsored by the SIAM Activity Group on Discrete Mathematics and ACM SIGACT. Facts below were checked 2026-07-08 against the SIAM SODA27 submission page (read via search rendering; siam.org refused direct fetches) and soda27.hotcrp.com. The 2027 paper deadline is July 9, 2026, anywhere on Earth — if you are reading this in early July 2026, the clock is already inside the final week. Re-verify every number on the live SIAM page for any later cycle.

Submit the full version — SODA means it

SODA imposes no page limit and the CFP encourages authors to submit the full version of the paper. This inverts the instinct trained by page-capped venues:

• Do not amputate proofs to look "conference-length." A SODA submission missing a proof invites the referee to assume the gap is real.
• The document still needs an act structure. Committee members triage hundreds of submissions; the ones they champion make the result and the technique graspable before page 5, with the complete proofs behind that front matter.
• The title page carries the title plus an abstract of one to two paragraphs summarizing the contributions — this abstract is the bidding surface for the PC, so it must state the theorem, the prior bound it beats, and the new idea.

Format floor (2027 wording)

Typeset in 11-point or larger font, single column, single-spaced between lines with ample spacing throughout, 1-inch margins on all sides, letter-size (8.5 x 11 inch) paper. The CFP warns that a submission deviating from the guidelines "risks summary rejection without consideration of its merits." A compliant skeleton:

\documentclass[11pt,letterpaper]{article}
\usepackage[margin=1in]{geometry}
\usepackage{amsmath,amssymb,amsthm,thm-restate}
% Lightweight double-blind: omit \author entirely; keep all citations intact.
\title{Faster Algorithms for X via Y}
\date{}
\begin{document}
\maketitle
\begin{abstract}
One to two paragraphs: problem, prior bound, new bound, technique.
\end{abstract}
% Sections 1-3: result, technique overview, comparison to prior bounds.
% Remainder: full proofs. No page limit; completeness is expected.
\end{document}

Lightweight double-blind, SODA flavor

The 2027 cycle uses what the CFP calls a fairly lightweight double-blind process. The operational rules differ from heavyweight-anonymity venues and getting the flavor wrong hurts in both directions:

Rule	What it means in practice
No author names, affiliations, or emails on or in the submission	Delete the author block; scrub \thanks, grant lines, and PDF metadata
Dissemination stays free	Posting to arXiv, giving talks, and circulating drafts are all explicitly permitted
Important references must not be omitted or anonymized	Cite your own prior work normally, in third person — hiding it is the violation
Anonymity is best-effort, not forensic	Reviewers who recognize the arXiv version are not a policy problem; a signed PDF is

Sweep before upload: pdfinfo paper.pdf for metadata, acknowledgements section removed, self-references rewritten from "we showed in [7]" to "as shown in [7]."

Prior and concurrent work constraints

Submissions must be original: not previously published and not simultaneously under review at another conference with published proceedings or a journal. arXiv posting does not count as prior publication. The dangerous overlap cases for an algorithms group are ESA, ICALP, STOC, and FOCS windows — a paper still alive at one of these when SODA's deadline closes is a dual submission (exact 2027 policy wording: 待核实 against the live CFP before relying on edge cases).

HotCRP mechanics

• Site: soda27.hotcrp.com. Create the record days early; first-time HotCRP accounts, conflict-of-interest entry, and topic tagging consume an evening.
• The form's title and abstract must match the PDF's title page; the form abstract is what PC members see first while bidding.
• Declare conflicts against the PC honestly and completely — the algorithms community is small, and an undeclared advisor relationship is an integrity flag.
• Upload, then download the stored PDF and re-open it. Confirm the compile you intended is the compile they have.

The final 48 hours, hour-ordered

AoE means the deadline passes only when July 9 ends at UTC-12 — but treating that as slack is how groups discover HotCRP outages at the worst moment.

1. T-48h: freeze mathematics. Any theorem edit after this point requires re-running the drift check in soda-reproducibility.
2. T-36h: compile the candidate PDF; run the mechanical audits below.
3. T-24h: upload a complete draft to HotCRP. From here on, every improvement is a re-upload, never a first upload.
4. T-12h: least-involved coauthor reads title page + overview cold and states the headline bound back to you. Mismatch means the abstract fails.
5. T-2h: final re-upload; download the stored PDF; page through it.

Mechanical audit commands worth scripting once per group:

pdfinfo paper.pdf | grep -iE 'author|title|producer'   # metadata leaks
pdffonts paper.pdf | awk 'NR>2 && $5=="no"'            # unembedded fonts
pdftotext paper.pdf - | grep -inE 'full version|supplementary|anonym'
pdftotext paper.pdf - | grep -inE 'acknowledg|grant no|we thank'

Each grep hit is either a leftover from another venue's template or an anonymity leak; both are five-minute fixes at T-36h and unfixable at T+1.

Failure triage

Failure	Consequence	Recoverable after July 9?
Missed AoE deadline	No submission; next stop STOC (November) or wait a year	No
Two-column or 10-point formatting	Summary-rejection risk per CFP	No
Author names in PDF or metadata	Double-blind violation, chair discretion	No
Key lemma unproved "for space"	Reviewers treat the result as unestablished	No — there was no space constraint to blame
Abstract oversells vs. theorem	Credibility damage during rebuttal	Partially, in the September response

Output format

[SODA readiness] Ready / Needs fixes / Not ready
[Format audit] <font/column/margins/title-page findings>
[Anonymity audit] <names, metadata, self-citation phrasing>
[Completeness audit] <unproved claims, missing comparisons>
[HotCRP audit] <record, conflicts, form-PDF mismatches>
[Fix order] <ranked actions before the AoE cutoff>