rebuttal-writing

name	rebuttal-writing
description	Use when the user wants to draft concise conference rebuttals under response-period constraints and prioritize which reviewer concerns to address.

Rebuttal Writing

Overview

Drafts concise, strategic rebuttals for conference paper reviews under tight word limits. Optimized for the specific constraints of ML/CS conference rebuttal periods (NeurIPS, ICML, ICLR, ACL, etc.) where brevity, clarity, and prioritization are critical.

When to Use

• User has received conference reviews and needs a rebuttal draft
• User asks "how do I respond to this reviewer in 500 words?"
• User wants to prioritize which concerns to address in a rebuttal
• Lab is preparing for NeurIPS / ICML / ICLR author response period
• User needs help deciding what can be addressed vs. what requires new experiments

Key Capabilities

• Parse and rank reviewer concerns by impact on accept/reject decision
• Draft concise point responses within strict word budgets (NeurIPS: 500 words, ICLR: unlimited, etc.)
• Identify "fatal flaw" concerns that must be addressed vs. minor concerns to acknowledge briefly
• Suggest which concerns to push back on vs. accept
• Draft responses that promise specific additions to the paper revision
• Generate a strategy document before drafting: "here's how we recommend allocating your word budget"

Usage Examples

Draft a full conference rebuttal

rebuttal_writing.draft(
    reviews={"R1": "...", "R2": "...", "R3": "..."},
    paper_abstract="...",
    conference="NeurIPS",
    word_limit=500,
    strategy="address_fatal_flaws_first"
)

Generate rebuttal strategy first

rebuttal_writing.strategy(
    reviews={"R1": "...", "R2": "...", "R3": "..."},
    conference="ICML",
    output=["concern_ranking", "word_budget_recommendation", "pushback_opportunities"]
)

Draft response to a single critical concern

rebuttal_writing.single_response(
    concern="The comparison to baseline X is unfair because different hyperparameter budgets were used",
    response_budget=80,  # words
    supporting_evidence="We used the same compute budget as reported in [BaselineX paper] Appendix B"
)

Output Format

Strategy document is a prioritized table with: concern, severity, recommended response type, word budget. Rebuttal draft is plain text respecting word limits, structured as numbered responses. Each response is ≤ allocated word budget.

Notes

• NeurIPS/ICML rebuttals: prioritize fatal flaws, be extremely concise, promise specific revisions
• ICLR rebuttals: more space for detailed responses, engage with all substantive concerns
• Do NOT promise experiments in the rebuttal that cannot realistically be completed before the camera-ready deadline
• Combine with review-response for the full revision cycle after acceptance