icdt-review-process
GitHub解析ICDT数据库理论会议的审稿流程，涵盖双周期提交、匿名评审及修订机制。重点强调证明正确性审查标准，区分可修复漏洞与根本性错误，并对比PODS和EDBT在日历、出版商及评审侧重点上的差异，指导作者策略。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill icdt-review-process -g -y
SKILL.md
Frontmatter
{
"name": "icdt-review-process",
"description": "Use when reasoning about how an ICDT (International Conference on Database Theory) submission is evaluated, covering the two-submission-cycle model, the first-cycle revision (Accept \/ Revise \/ Reject) decision, anonymous review since 2024, the cross-cycle resubmission restriction, proof-correctness scrutiny by a database-theory PC, and how ICDT's process differs from PODS and from the co-located EDBT systems track."
}
ICDT Review Process
Model the pipeline before interpreting any single review. ICDT's process is built around two submission cycles per year and a first-cycle revision option, and the reviewing bar is proof correctness: the referees are database theorists who will check the argument, not just weigh the idea. The most consequential mental shift for authors arriving from a systems venue is that a wrong or incomplete proof is not a "revise" — it is often a reject, because the contribution is the theorem.
Process model
- Submission and review run on Microsoft CMT with anonymous review (regular track, since 2024): the PC does not see author identities.
- Each cycle has an abstract deadline then a paper deadline about a week later; the PC bids and is assigned from the registered abstracts.
- Cycle 1 decisions are Accept, Revise-and-resubmit, or Reject. A "revise" is a genuine invitation to fix identified gaps and return within the cycle for a second read.
- Cycle 2 reaches a final Accept/Reject. A paper rejected in Cycle 1 cannot be resubmitted to Cycle 2 unless the reviewers explicitly invited it — the cycles are not two independent lotteries.
- Accepted papers publish open access in LIPIcs (Schloss Dagstuhl), so camera-ready proof completeness and metadata matter as much as the verdict.
Reading a decision against the categories
| Decision | What it means | Author move |
|---|---|---|
| Accept | The theorem holds and the advance is real; polish only | Camera-ready; finalize the full-version proofs; do not reopen scope |
| Revise-and-resubmit (Cycle 1) | A fixable gap: a case in a proof, a missing bound, an unclear model | Treat as a deadline: fix each identified point and document it in the revised version |
| Reject | Structural: a broken proof, a known result, a model with no consequence | Do not resubmit to Cycle 2 uninvited; repair deeply or reroute (PODS / a journal) |
The strategic reading: an ICDT referee distinguishes a gap you can close (a missing lemma, a special case) from a claim that is wrong (the main theorem fails). Write so that any weakness a referee finds is of the first kind — an argument you can complete in the revision — not the second.
How ICDT differs from its siblings
- vs. PODS: PODS is the other database-theory flagship, but North-American-leaning, co-located with SIGMOD, and now published in ACM PACMMOD. ICDT is the European venue, co-located with EDBT, published open access in LIPIcs, and runs a two-cycle calendar with a revision option rather than PODS's ACM-journal framing. The scope overlaps heavily; the calendar, publisher, and revision mechanics do not. Never assume a shared deadline or template.
- vs. EDBT (the co-located systems track): EDBT reviews systems and experimental data-management work with reproducibility expectations; ICDT reviews theorems. Being in the same building and the same week does not mean the same reviewer culture — an ICDT referee wants a proof, not a benchmark.
- vs. LICS / ICALP / STACS: these EATCS-adjacent theory venues share ICDT's proof rigor but not its data-management framing; ICDT rewards a result whose consequence for querying, constraints, or data is clear, not theory for its own sake.
Who reads you
Expect referees from your subarea of database theory — query languages, finite model theory, complexity of query evaluation, consistent query answering, provenance, or data-integration theory. They will read the proofs, look for the model of computation and the exact complexity class, check whether upper and lower bounds match, and notice an unstated assumption that makes a theorem easier than advertised. Vague "it can be shown that" statements are caught, not skimmed.
Where author leverage actually exists
[Before submission] precise abstract + subject areas -> the right referees (largest lever)
[Cycle choice] target Cycle 1 if the proof might still move (revision option)
[Revision round] the strongest lever: close every identified gap and say where you did
[After reject] no uninvited Cycle-2 carry; reroute to PODS or a journal (TODS/LMCS/TheoretiCS)
A revision succeeds when it closes the specific gap the referees named and documents the change; it fails when it argues that the gap does not matter. Silent non-fixes — an identified missing case left unaddressed — turn the second read into a reject.
Misreadings to avoid
- Treating "revise" as an accept — the second read re-checks the proof; budget the revision window like a deadline.
- Treating Cycle 2 as a retry lane — the cross-cycle rule blocks uninvited resubmission of a Cycle-1 reject.
- Assuming a systems-style rebuttal exists — ICDT's mechanism is the revision, not a short rebuttal letter (confirm the current cycle's exact author-interaction rules; 待核实).
- Confusing ICDT with PODS process — different publisher, calendar, and revision model.
Output format
[Process stage] pre-submission / awaiting Cycle-N decision / revising / final / accepted
[Decision category] accept / revise / reject, with the criterion (correctness | novelty | consequence)
[Referee map] each point -> proof gap | model unclear | known result | bound not tight | scope
[Leverage plan] the next-stage action (which gap to close) that can change the outcome
[Forbidden moves] uninvited Cycle-2 resubmission / identity leak / unproved claim left standing
Version History
- 9f86f09 Current 2026-07-19 15:56
