ICDE Author Response
Use this after ICDE reviews arrive. Reopen the current call's review-and-revision instructions
before drafting, because ICDE's outcome model — and whether a Revise & Resubmit exists this
edition — is edition-specific (待核实 for the current cycle).
Two response genres
ICDE can hand you either of two response tasks; identify which one you have before writing.
-
A rebuttal to initial reviews (where the process offers one) — a short, factual reply
that corrects errors and answers the decision-critical objection per reviewer.
-
A revise-and-resubmit change document — when the first reviewing phase returns
Revise & Resubmit (ICDE 2026 model), you get roughly four weeks to make concrete
changes, re-reviewed by the same reviewers before a final Accept/Reject.
Triage (both genres)
- Answer concerns that affect correctness, novelty, evaluation soundness, or reproducibility
first; cosmetic complaints last.
- Because ICDE is single-blind, you may cite your own prior work and name your system
freely — there is no anonymity to protect in the reply.
- Anchor every answer to concrete evidence: a paper section, an added experiment, a disclosed
cost, or a specific edit — not a promise.
- Correct factual misreadings before conceding anything; a reviewer who misread Figure 7
should be pointed to the exact panel, courteously.
The revision change document
For an invited revision, build a requirement ledger the same reviewers can check line by
line:
- Extract every required and suggested change into a numbered list, tagged by reviewer.
- For each, decide: done (point to the new text/figure), partially done (state the limit
honestly), or infeasible in four weeks (say so and explain).
- Write the change document so a reviewer can verify each item without hunting — quote the new
sentence, name the new figure, give the section.
- Do not smuggle in scope changes the reviewers did not ask for; an invited revision is judged
against its list, and surprise additions read as evasion.
Systems-reviewer pushback patterns
| Pushback |
What it signals |
ICDE-ready fix |
| "Baseline X is untuned" |
The reviewer suspects an unfair comparison |
Report the tuning budget you gave X, or re-run it tuned and show the new numbers |
| "Results are single-point" |
No scale or variance shown |
Add a scale curve and report variance across runs, not one median |
| "Cost of the mechanism is hidden" |
The trade-off is not disclosed |
Add the memory/bandwidth/latency cost table so the trade is legible |
| "Only synthetic workloads" |
Doubt about real-world relevance |
Add a real workload, or scope the claim to the regime you actually tested |
| "Prior system Y already does this" |
Novelty challenge |
Name the specific mechanism difference and, if possible, compare against Y directly |
Feasibility discipline
- Before promising a demanded experiment, price it against the four-week window and your
hardware access; a revision plan you cannot finish is worse than an honest partial.
- One decision-critical resolution per reviewer beats an exhaustive reply; the area chair reads
for whether the central objection was actually answered.
- Reply and revise early — the window is short and same-reviewer re-review rewards clarity over
volume.
Output format
[Genre] rebuttal / revise-and-resubmit change document
[Priority issue] <reviewer concern>
[Decision dimension] correctness / novelty / evaluation / reproducibility
[Draft response] <ICDE-ready text with evidence anchors>
[Requirement ledger] <numbered change -> done/partial/infeasible, per reviewer>
[Feasibility check] <fits the 4-week window? y/n + why>
