brycewang-stanford ICDT Author Response
Write the response for a venue where the referees checked your proofs. ICDT's central author-interaction mechanism is the Cycle-1 Revise-and-resubmit: not a short rebuttal that argues taste, but a returned paper in which each gap the referees named has been closed and the change documented. The second read is done by the same referees; a point raised and not addressed is what turns a revision into a reject.
Confirm the exact interaction model for the live cycle: ICDT's primary lever is the revision; whether a separate short rebuttal/author-comment phase runs in a given edition is 待核实 — read the current call. The advice below applies to the revision and, where one exists, to a brief comment phase.
Triage the referee points first
Sort every point by kind, because the fix differs:
| Point kind | Example | Response |
|---|---|---|
| Correctness gap | "The proof of Lemma 4 omits the cyclic case" | Fix it in the paper; show the added case; this is top priority |
| Missing bound | "Only the upper bound is proved" | Prove the matching lower bound or scope the claim honestly |
| Model ambiguity | "Is this data or combined complexity?" | Pin the model in the body; correct any statement that was ambiguous |
| Positioning | "How does this differ from [PODS year]?" | Add the precise delta to related work (icdt-related-work) |
| Clarity | "Definition 2 is hard to parse" | Rewrite; a confused referee re-reads the revision |
| Genuine misreading | The referee misunderstood a quantifier | Correct it and fix the prose that allowed the misreading |
Correctness first, then bounds, then model, then clarity and positioning.
Write the revision, not a debate
- Close the gap in the paper, then point to where: "Lemma 4 now covers the cyclic case (Appendix A, para 3)." The change lives in the PDF; the response is a map to it.
- Prefer a fix over an argument. "We prove the missing lower bound in the new Section 5" beats "we believe the lower bound is not essential."
- Where you decline a request, say why, technically. "A matching lower bound would require resolving [open problem]; we instead scope the claim to [restricted class] (Theorem 3)." An explicit, reasoned decline is respected; a silent omission is not.
- Never argue taste. "We think the model is natural" moves nothing; a referee who found a gap wants it closed, not defended.
Anonymity in the response (since 2024)
The regular track is anonymous through review, so the revised paper and any response text must not de-anonymize you:
- No author names, no "our prior work," no acknowledgments reintroduced in the revision.
- Do not add an arXiv/full-version link that names you into the revised PDF.
- Keep self-citations third-person in the revised text (see
icdt-related-work).
A point-by-point skeleton for the revision letter
Summary: what changed at the top level (e.g., "added the matching lower bound; fixed the cyclic
case in Lemma 4; clarified the complexity measure throughout").
R1.1 [correctness] "Lemma 4 omits the cyclic case."
-> Fixed. The cyclic case is now handled in Lemma 4, with the argument in Appendix A.
R1.2 [bound] "Only an upper bound is shown."
-> Added. Theorem 5 now proves a matching lower bound by reduction from <problem>.
R2.1 [model] "Data or combined complexity?"
-> Clarified. The bound is combined complexity; stated explicitly in Section 3 and the abstract.
R2.2 [positioning] "Relation to [X, PODS year]?"
-> Added to Related Work: their result is the special case of bounded arity; ours removes it.
The cross-cycle rule (do not misuse the response)
If your paper was rejected in Cycle 1 (not invited to revise), a response cannot resurrect it,
and you may not resubmit to Cycle 2 unless the referees explicitly invited it. Reroute instead
(icdt-topic-selection). The response letter belongs to the revision path, not the reject path.
Output format
[Interaction type] Cycle-1 revision / short comment phase / na (rejected)
[Point triage] correctness / bound / model / positioning / clarity, each -> concrete change
[Changes made] <list of paper edits with locations>
[Declines] <each declined request + technical reason>
[Anonymity] revised PDF and letter non-identifying? yes/no