brycewang-stanford EuroSys Author Response
Use this once EuroSys reviews arrive, or when a revision decision lands. The decisive EuroSys fact: every submission ends as accept, reject, or a one-shot revision offer (EuroSys 2026/2027 CFPs, rendered 2026-07-08). A response is not just damage control — it is the input that decides which of the three bins the PC discussion puts you in. Rebuttal mechanics (window length, word caps, whether a response phase exists at all in the current round) are set per cycle in HotCRP — 待核实 against the round's own instructions before drafting.
Read the reviews as a bin-sorting problem
| Review pattern | Likely bin | Response goal |
|---|---|---|
| Positive scores, factual misreadings | Accept-adjacent | Correct the record crisply; give the champion ammunition |
| "Promising but evaluation incomplete" | Revision candidate | Show the missing experiment is bounded and specific |
| Disputed novelty or contribution | Contested | Name the delta against the exact cited system |
| Fundamental design objection | Reject-adjacent | Concede scope honestly; salvage what is defensible |
The revision-candidate row matters most at this venue: reviewers who believe the fix is enumerable can argue for a revision offer instead of a reject, and the condition list they write becomes your contract.
Drafting rules for systems reviewers
- Lead with the one correction that changes a score, not with thanks.
- Quote the paper by section and figure number; EuroSys reviewers re-open the PDF.
- Answer measurement objections with the measurement's own terms: workload, hardware, baseline version, repetition count — not adjectives.
- Never promise results you have not already produced; a EuroSys revision offer will convert vague promises into explicit conditions you must then meet.
- Stay anonymous. The double-blind contract holds through the response phase.
Response skeleton
R2's central concern: the comparison against <baseline> uses its default config.
- Fact: §5.2 tunes <baseline> per the authors' published guidance (Fig. 7 caption).
- New clarification: at the tuned setting, the gap is 1.8x, not the 3x headline;
the 3x case is the untuned default, which §5.2 labels as such.
- Text change if accepted/revised: promote the tuned number to the abstract.
Minor points: [R1-a] typo confirmed. [R3-b] the trace is public; link withheld
under double-blind, provenance stated in §4.1.
If the outcome is a one-shot revision
The offer arrives with a list of necessary conditions — extra analysis, experiments, or comparisons that substantiate the claims. Treat it like a shepherded contract with a hard deadline: the next EuroSys submission round.
- Convert each condition into a task with an owner and a machine reservation.
- Decide within a week whether every condition is physically satisfiable before that round's paper gate; declining early beats failing late.
- Write the revision note as condition → change → location-in-paper, one row per condition, so the returning reviewers can verify without rereading everything.
- Do not smuggle in a new contribution; the revision is judged on the list, and scope creep reads as instability.
A revision accepted at the following round is presented at the edition it lands
in — plan travel and camera-ready effort accordingly (see eurosys-workflow).
Word-budget allocation
Whatever cap the round sets, spend it in this ratio:
| Share | Spend it on |
|---|---|
| ~50% | The single score-moving objection, answered with evidence |
| ~25% | Factual corrections that change how reviews read each other |
| ~15% | The most likely revision conditions, pre-scoped and bounded |
| ~10% | Minor points, batched one line each with review IDs |
Resist symmetric per-reviewer allocation: the PC discussion is one conversation, not four, and the meta-question it settles is "does the author team understand what is actually wrong?"
Timeline discipline
- Draft the bin-sort within 24 hours of reviews landing; the first reading under emotion misclassifies severity in both directions.
- Circulate the score-moving answer to all authors before polishing minor points — co-author disagreement about the central concession is the thing you cannot fix on deadline morning.
- If the round runs reviewer-author discussion rather than a one-shot rebuttal (待核实 per round), answer early; late replies reach reviewers who have already argued their position in the PC thread.
- Keep a copy of exactly what was submitted; a later revision offer will be interpreted against the rebuttal's promises.
What sinks responses here
- Arguing taste ("we believe the design is elegant") instead of evidence.
- Burying the score-moving answer under ten minor-point replies.
- Disputing a reviewer's competence — the PC discussion is a room of their peers.
- Treating a revision offer as a soft accept; unmet conditions produce a reject and the same-season resubmission ban that follows it.
Output format
[Bin estimate] accept-adjacent / revision candidate / contested / reject-adjacent
[Score-moving issue] <the one objection worth most of the word budget>
[Fact corrections] <misreading -> section/figure evidence>
[Revision-condition plan] <condition -> task -> feasible by next round? yes/no>
[Anonymity check] passed / issues