chi-author-response
GitHub指导ACM CHI论文在收到修改重投邀请后，于五周内完成修订与回复。涵盖审稿意见分类、带追踪变化的稿件生成及结构化回复信撰写，旨在通过清晰的审计轨迹和针对性回应提高二审录用率。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill chi-author-response -g -y
SKILL.md
Frontmatter
{
"name": "chi-author-response",
"description": "Use when an ACM CHI paper receives a revise-and-resubmit invitation and the five-week window opens — planning the revision triage, making tracked changes reviewers can audit, and writing the response letter that carries the paper through round 2 and the PC meeting."
}
CHI Author Response
CHI does not run a short rebuttal where you argue in a text box. An invited paper gets roughly five weeks to deliver two artifacts: a genuinely revised manuscript with changes visibly tracked, and a response letter mapping every reviewer concern to what changed. For CHI 2027 the resubmission deadline is December 3, 2026 (posted on chi2027.acm.org, read 2026-07-08). The stakes are real: the posted process warns up to 50% of revised papers may still be rejected; CHI 2026 accepted 65.5% of resubmissions.
The first 48 hours: triage, not writing
Read all reviews twice, write nothing outward, then sort every distinct criticism:
| Triage class | Test | Response strategy |
|---|---|---|
| Must-fix | Named by the 1AC meta-review or ≥2 reviewers | Change the paper substantively; show the diff |
| Cheap win | Clarity, missing citation, terminology, figure | Fix all of them, every single one |
| Analysis request | Wants a new analysis of existing data | Run it if the pipeline is alive; report honestly either way |
| New-study request | Wants data you cannot collect in five weeks | Do not fake it; scope the claim down instead and say so |
| Misreading | Reviewer misunderstood something actually present | Treat as your writing defect: fix the text, then point to it |
| Genuine disagreement | Legitimate difference of research values | Push back once, respectfully, with evidence and citations |
The meta-review is the priority list. Reviewers at CHI re-read revisions against their own round-1 comments, and the PC meeting discusses the 1AC's summary of how well you responded — not the eloquence of your prose.
Revising so the diff does the arguing
The instructions ask for a revision with tracked changes — highlighted, conventionally in color — plus the response letter. Make the audit trail mechanical:
# LaTeX: generate the tracked-changes PDF from the submitted vs revised source
latexdiff --type=CULINECHASE submitted/main.tex revised/main.tex > diff.tex \
&& latexmk -pdf diff.tex
# Sanity: confirm the revision still respects CHI length norms after additions
texcount -inc -sum revised/main.tex | tail -3
# Keep anonymization intact through round 2 — new text leaks too
pdftotext revised/main.pdf - | grep -nEi 'our (lab|university)|github\.com/[a-z]' | head
Five weeks invites over-revision. Do not restructure sections no reviewer questioned, and do not let the word count balloon past what the contribution justifies — length norms still apply at resubmission, and a bloated revision reads as panic.
The response letter
Structure it per reviewer, with numbered items quoting or precisely paraphrasing each concern, in the reviewers' own order. For each item: what changed, where (section and page in the revised PDF), and — where relevant — one sentence of substance. A workable skeleton:
Response to Reviews — Submission #1234
Summary of changes: <5-8 bullets, the big moves only>
R1.1 "<concern>" → Revised §4.2 (p.7): we now report <what>. <One sentence of
substance or evidence.>
R1.2 "<concern>" → We respectfully keep <choice> because <evidence/citation>,
and now state this rationale explicitly in §3.1.
2AC.1 ...
Tone rules that decide borderline cases at the PC meeting:
- Thank reviewers once, globally; do not lard every item with gratitude.
- Never write "the reviewer is wrong." Write what the paper now says and where.
- Concede real limitations plainly and move them into the limitations section — reviewers reward honesty at CHI more reliably than at most venues, because study limitations are expected in human-subjects work.
- If two reviewers conflict, name the conflict neutrally and state which path you took and why, so the 1AC can resolve it rather than rediscover it.
What you may and may not change
The revision is expected to respond to the reviews; it is not a fresh submission. Safe: new analyses of existing data, reframed claims, restructured argument, improved figures, expanded related work, fixed statistics. Risky without flagging: changing the core claims, swapping the method story, adding an entirely new study — if the paper needed those, discuss withdrawing and resubmitting next cycle instead, because the same review team will judge whether the new material was reviewed properly. Author lists and anonymization rules stay as they were (待核实 for edge cases: check the current cycle's instructions before touching authorship).
If you decide not to resubmit
Withdrawing is legitimate — 27 of 2,603 invited CHI 2026 papers did not return. Do it formally in PCS, keep the reviews, and fold them into the next venue's draft while they are fresh. An RR invitation you decline is still market information: the paper was in the top ~39% of a 6,700-paper pool.
Output format
[Window] opens <date> → due 2026-12-03; days remaining: <n>
[Triage] must-fix: <n> / cheap: <n> / analysis: <n> / out-of-scope: <n>
[Meta-review demand] <the 1AC's central ask, one line>
[Riskiest item] <the concern you cannot fully satisfy + chosen strategy>
[Deliverables] tracked-changes PDF ✓/✗ · response letter ✓/✗ · length re-check ✓/✗
Version History
- 9f86f09 Current 2026-07-19 14:39
