icdt-submission
GitHub用于在上传至Microsoft CMT前审计ICDT论文投稿，涵盖双周期选择、摘要与正文章节截止日期、lipics-v2021格式及15页限制、匿名提交要求及完整证明规范。
触发场景
安装
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill icdt-submission -g -y
SKILL.md
Frontmatter
{
"name": "icdt-submission",
"description": "Use when auditing an ICDT (International Conference on Database Theory) regular-paper submission for Microsoft CMT readiness, covering the two-step abstract-then-paper deadline in the correct submission cycle, the lipics-v2021 15-page limit excluding references, the clearly-marked appendix read at the PC's discretion, anonymous submission since 2024, the complete-proofs \/ full-version expectation, and which submission problems are unfixable after the AoE cutoff."
}
ICDT Submission
Run this audit before uploading to Microsoft CMT for the International Conference on Database
Theory. ICDT is a pure database-theory venue: reviewers read for a precise theorem, a correct
proof, and a genuine advance in the foundations of data management, and they publish the result
open access in LIPIcs (Schloss Dagstuhl). Every number below was read on 2026-07-09 from the
ICDT 2026/2027 calls at databasetheory.org via search renderings (see
resources/official-source-map.md); treat them as a one-cycle snapshot and reopen the live call
first.
Pick the right cycle first
ICDT runs two submission cycles per year, and choosing the cycle is the first decision, not a detail:
- Cycle 1 closes in the (northern) spring and offers a revision option — a first decision can be Accept, Revise-and-resubmit, or Reject.
- Cycle 2 closes in the (northern) autumn; a paper rejected in Cycle 1 may not be resubmitted to Cycle 2 unless the reviewers explicitly invited it. Do not treat the two cycles as a free second attempt.
For ICDT 2027 (Lille), Cycle 1 papers were due 10 March 2026 and Cycle 2 papers 10 September 2026 (AoE); confirm the live dates. If your proof might still move, the earlier cycle's revision round is worth more than a later raw submission.
The two-step deadline
Each cycle separates an abstract deadline from the paper deadline, both AoE and about a week apart (ICDT 2027 Cycle 1: abstracts 3 March 2026, papers 10 March 2026):
- Register the abstract with the real title and abstract — it drives PC bidding and conflict handling. A placeholder abstract worsens your reviewer match.
- Miss the abstract deadline and CMT will not accept the PDF later.
Format and page budget
lipics-v2021LIPIcs document class, unmodified. This is the Dagstuhl LaTeX style, notacmartand notIEEEtran— do not carry a PODS/SIGMODacmarthabit across.- At most 15 pages excluding references. References do not count against the 15; nothing else escapes it except the appendix below.
- A clearly marked appendix is allowed but read at the PC's discretion. The main 15 pages must contain everything needed to assess the contribution; the appendix holds the full proofs a referee may check. Online/external appendices are not allowed — everything reviewed lives in the single submitted PDF.
Complete proofs and the full version
ICDT expects complete, checkable proofs. A theorem stated without a proof a referee can verify is the classic desk-level weakness at a theory venue:
- Put proof sketches and the main argument in the body; put the full details in the marked appendix so the paper is self-contained within one PDF.
- Maintain a full version (typically on arXiv) that will carry the complete proofs after review; reference it in the camera-ready, not in an anonymized submission if the link deanonymizes you.
Anonymous-submission sweep (since 2024)
ICDT regular papers are anonymous: the PDF must not reveal authorship.
# Mechanical pass on the submission PDF
pdfinfo paper.pdf | grep -Ei 'author|creator|producer'
pdftotext paper.pdf - | grep -nEi 'acknowledg|grant|funded by|our (previous|earlier) (work|paper)|\bwe .* in \[[0-9]+\]' | head
- Remove author names, affiliations, and acknowledgments to funding sources and collaborators.
- Cite your own prior work in the third person ("Building on the construction of [17]…"), not "our earlier paper [17]."
- If you post a full version, do not let an arXiv link or an identifying system name reintroduce your identity into the anonymous PDF.
- Note the exception: the Database Theory in Action short-paper track is not anonymous — the anonymity rule is for the regular research track.
Desk-risk triage
| Finding at audit time | Severity | Real fix |
|---|---|---|
| Body over 15 pages (excl. refs) | Format-reject-grade | Move detail to the marked appendix; tighten the body — refs do not absorb it |
Class altered (margins, font, \vspace hacks) |
Named format ground | Recompile clean lipics-v2021; recover space by editing, not by shrinking |
| A theorem with no verifiable proof anywhere in the PDF | Fatal at a theory venue | Add the full proof to the appendix before upload |
| Author identity leaks (names, acks, "our tool") | Anonymity violation | Re-anonymize; scrub PDF metadata; neutralize self-citations |
| Abstract not registered by the earlier deadline | No paper slot exists | Nothing fixes this post-AoE — calendar it now |
| Resubmitting a Cycle-1 reject into Cycle 2 uninvited | Against the cross-cycle rule | Do not; only resubmit if reviewers invited it |
| Same result under review at PODS/a journal | Concurrent-submission exposure | Withdraw one; check the current dual-submission wording |
Final-week order of operations
- Freeze the theorem statements early; the exposition can churn, the results cannot.
- Register the abstract in the correct cycle before the earlier deadline.
- Confirm every claimed theorem has a complete proof in the body or the marked appendix.
- Run the anonymity sweep on the final PDF, including self-citation phrasing.
- Fill every CMT field — subject areas that match your result, conflicts for each coauthor's institution and recent collaborators — a day early.
- Re-download the uploaded PDF and read it cold to confirm it is the file you meant.
Reverify each cycle
- The cycle dates, which cycle you are targeting, and the revision-window length.
- The page limit and the required LIPIcs class revision.
- The anonymous-submission wording and any generative-AI disclosure rule (待核实).
- Whether the Database Theory in Action track runs and its separate rules.
Output format
[ICDT submission status] ready / blocked / needs work
[Cycle] 1 (revision-eligible) / 2 (no uninvited carry), abstract+paper registered? yes/no
[Format] pages used (body/refs), lipics-v2021 compliance, appendix marked?
[Proofs] every theorem has a checkable proof in-PDF? yes/no
[Anonymity] clean / leaks: <where>
[Fix queue] <ordered, with dates before the AoE cutoff>
版本历史
- 9f86f09 当前 2026-07-19 15:56
