Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› icassp-author-response
icassp-author-response
GitHub用于撰写ICASSP论文作者回复。针对单盲评审和纯文本回复限制，指导如何纠正事实错误、回应信号处理领域的常见质疑，并聚焦技术委员会决策需求，提供专业且简洁的回复策略与模板。
Trigger Scenarios
需要撰写ICASSP会议 rebuttal
收到ICASSP审稿意见后准备回复
处理单盲评审下的作者回应
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill icassp-author-response -g -y
SKILL.md
Frontmatter
{
"name": "icassp-author-response",
"description": "Use when drafting an ICASSP rebuttal or author response to reviews, covering the recently added and short author-response window, the single-blind setting where reviewers already know you, answering signal-processing reviewers from the four-page record without a revised PDF, and writing for the technical-committee decision rather than for tone."
}
ICASSP Author Response
Use this after ICASSP reviews are released. First confirm whether the current cycle even runs a rebuttal: ICASSP historically had no author-response stage, and a short window is a comparatively recent addition (in the 2026 cycle authors could reply until ~22 December 2025). Reopen the current call before promising reviewers anything.
What the rebuttal is and is not
- It is a short, text-based reply to the reviews, not a chance to upload a revised PDF or new experiments. Argue from what you already submitted.
- ICASSP is single-blind, so the reviewers know your identity; write professionally and do not attempt to re-establish anonymity or drop identifying hints to sway a reviewer.
- The audience is the area/technical-committee decision: give whoever aggregates the reviews a clean reason to accept, not a point-by-point defense that wins arguments but not the paper.
Triage before drafting
- Separate factual errors in the review (a metric misread, a missed baseline that is in the paper) from legitimate gaps (a missing comparison, an unreported condition).
- Fix the factual errors first with an exact pointer: "Table 2, row 3 already reports this."
- For real gaps, concede cleanly and scope a camera-ready clarification that adds no unsupported new claim — the four pages you submitted are the evidence of record.
- Rank concerns by decision weight; one resolved correctness objection outranks five answered style notes.
Signal-processing reviewer pushback patterns
| Pushback | What it signals | ICASSP-ready fix |
|---|---|---|
| "The baseline is not the current strong method" | Reviewer knows the task's state of the art | Name the specific stronger baseline you did compare, or concede and scope it for camera-ready |
| "Which SNR / condition does this hold at?" | The reported number hides its operating point | Point to the sweep or table cell; if only one condition was tested, say so plainly |
| "Metric is not the one this task uses" | Measurement mismatch (e.g., accuracy where SI-SDR belongs) | Report the field-standard metric if you have it, or explain why yours is defensible |
| "No significance / spread over runs" | One-seed result | Cite the multi-run spread if present; otherwise concede it as a limitation |
| "This is a known result / prior art" | Novelty challenge | Cite the exact delta versus the named prior work in one sentence |
Drafting pattern
- Open with the single most decision-critical correction or concession.
- Anchor each point to an exact location in the submitted paper (table, figure, equation, line).
- State the signal-processing consequence, not just the fact.
- Promise only camera-ready wording or clarification fixes that stay within the accepted claim.
Micro-example
Reviewer objection: the enhancement gain is reported only at one SNR, so the improvement may not generalize. Reply skeleton:
- Concede the main table fixes SNR at one value for space.
- Point to Fig. 3, which already sweeps 0-15 dB and shows the gain persists.
- Note the trend is monotone, so the single-SNR table is representative, not cherry-picked.
- Offer one camera-ready sentence stating the operating range explicitly.
Calibration for a short window
- Reply early; ICASSP rebuttal windows are brief and a late, exhaustive reply is worth less than a prompt, precise one.
- Do not paste new derivations or fresh result tables the reviewers cannot verify against the submitted PDF.
- Keep within any stated length/format cap; if none is stated, be concise anyway — committee members read many rebuttals in one sitting.
Output format
[Rebuttal exists?] yes (dates) / no this cycle / unverified
[Priority issue] <reviewer concern>
[Decision dimension] correctness / novelty / evidence / clarity / metric-fit
[Draft response] <ICASSP-ready text, anchored to submitted paper>
[Evidence anchor] <table/figure/equation/line>
[Camera-ready promise] <scoped fix that adds no new claim>
Currency note
The 2026 rebuttal window (reported open to ~22 December 2025) was checked 2026-07-09 via
official-page renderings (see ../../resources/official-source-map.md). Whether a rebuttal
recurs, and its dates and length, are cycle-specific — verify on the current call before drafting.
Version History
- 9f86f09 Current 2026-07-19 15:53
