cikm-author-response
GitHub用于处理CIKM会议作者端各类评审沟通场景，包括撰写匿名回复、相机就绪版修订说明、被拒后的重新投稿简报及向程序主席提交的事实澄清函。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill cikm-author-response -g -y
SKILL.md
Frontmatter
{
"name": "cikm-author-response",
"description": "Use when preparing author-side communication around CIKM reviews — drafting for a response window if the cycle offers one (unconfirmed for 2026), writing camera-ready revision notes that answer reviewer concerns, handling post-decision chair correspondence, and converting rejection reviews into a resubmission brief."
}
CIKM Author Response
CIKM's response mechanics are genuinely cycle-dependent: no official statement confirming or denying a 2026 rebuttal phase could be retrieved (source map, 2026-07-08, 待核实). This skill therefore covers all four author-side writing situations around CIKM reviews, with the no-window case treated as the default — because when there is no rebuttal, the reply to reviewers gets written anyway, just in different documents.
Situation map
| Situation | The document you actually write | Governing constraint |
|---|---|---|
| A response window opens | Anonymous reply inside EasyChair | Announced word/format rules; anonymity persists |
| No window, paper accepted | Camera-ready revision note | Address concerns without exceeding accepted scope |
| No window, paper rejected | Resubmission brief for the next venue | Honest triage, no venue-blaming |
| Factual process error | Confidential note to PC chairs | Facts only; never argue merit this way |
If a window opens: the tri-lane reply
CIKM reviews arrive from up to three community lanes (cikm-review-process), so a
good reply is organized by reviewer, not by theme — each lane's reviewer should find
their own objection answered in their own vocabulary:
- Lead each reviewer block with the concern that most threatens their score, not the one easiest to rebut.
- Answer from submitted material — section, table, appendix, or the artifact the paper already cites. New results pasted into a reply carry no weight a reviewer is obliged to credit and can violate window rules; check the announcement.
- Concede real limits in one sentence and scope them ("holds for static graphs; streaming is future work") — the blended pool punishes evasion harder than limitation, because at least one reviewer is expert enough to see through it.
- Preserve anonymity end to end: no names, no identifying URLs, no "as our earlier system at [company]" slips. The double-blind regime does not lapse at review.
Micro-example: the mining-lane scalability objection
Reviewer: "Complexity analysis suggests quadratic blow-up on dense graphs; the experiments avoid this regime."
Reply skeleton: (a) confirm the reading is correct for worst-case density; (b) point to the submitted Table/appendix entry showing observed scaling on the densest tested graph; (c) state the density regime of the target application and that the claim is scoped to it; (d) offer a one-sentence camera-ready scope clarification. Four moves, no new experiments, no evasion.
No window, accepted: the revision note
The August 7 → August 20 camera-ready sprint (2026 dates) is where reviewer concerns
get answered in practice. Build a concern → edit ledger: every substantive review
point maps to a camera-ready change, a scoped-limitation sentence, or a reasoned
no-change. Keep the ledger; if the proceedings process or a chair asks what changed,
the answer is ready, and cikm-camera-ready consumes it directly.
No window, rejected: the resubmission brief
Write it within a week, while context is fresh:
- Sort review points into: fix-with-evidence, fix-with-writing, disagree-with-reason.
- Tag each point by reviewer lane — lane-clustered objections identify which
community the paper failed, which drives the
cikm-workflowfallback choice. - Draft the one paragraph a future reviewer at the next venue would need to see changed. If that paragraph cannot be written, the problem is the contribution, not the venue.
Tone calibration for a three-community room
CIKM's communities have different argument cultures — IR discussion leans protocol-precise, mining leans delta-and-numbers, KM/database leans systems-pragmatic — but one register works for all three: technical, specific, and unwounded. Practical rules:
- Quote the reviewer's actual words before answering them; paraphrase invites "that is not what I said."
- Numbers over adjectives: "Table 3 row 4 shows 2.9 points at k=10" persuades every lane; "we respectfully believe the improvement is substantial" persuades none.
- Never grade the reviewer ("the reviewer misunderstands...") — write "the submission states this compactly in §4.2; we will make it explicit" and let the evidence do the accusing.
- Budget by leverage: one blocking objection answered completely beats five minor points each answered at 20%.
Verify-before-writing checklist
Response mechanics are exactly the kind of fact this venue changes per cycle, so before drafting anything reviewer-facing:
[ ] Reopen the current cycle's author instructions / notification email:
does a response channel exist at all this year?
[ ] If yes: word/character limit, format, deadline (AoE?), and whether
new results or links are permitted or barred
[ ] Anonymity rules during the window (assume double-blind persists)
[ ] Who may submit: contact author only, or any author?
[ ] For chair notes: the stated contact route on the host site, not a
guessed email address
Every item comes from the current host site or EasyChair instance — the 2026 source map deliberately records the rebuttal question as unresolved (待核实), and this checklist is how it gets resolved for whatever cycle is live.
The one-page ceiling
Whatever the channel — window reply, camera-ready note, or chair correspondence — hold to roughly a page of substance per audience. Beyond that length, the reader is skimming, and a skimmed concession does as much damage as an unmade one. If the honest response needs three pages, the paper had three pages of problems; choose the one whose resolution changes the decision and spend the page there.
Output format
[Situation] response-window / camera-ready note / resubmission brief / chair note
[Per-reviewer leads] <R1/R2/R3: the concern each block opens with>
[Evidence anchors] <submitted section/table/artifact per concern>
[Concessions] <scoped limits admitted, in one line each>
[Deliverable deadline] <window close / Aug-20-style camera-ready gate / next venue gate>
Version History
- 9f86f09 Current 2026-07-19 14:39
