micro-review-process
GitHub解析MICRO会议2026年双盲评审流程，涵盖从截稿、反驳到通知的时间线。说明评分维度（创新性、方法论可信度等）及行业赛道差异，强调格式合规与职业行为准则，辅助作者应对审稿。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill micro-review-process -g -y
SKILL.md
Frontmatter
{
"name": "micro-review-process",
"description": "Use when reasoning about how a MICRO submission is judged — the double-blind HotCRP pipeline from the April deadline through the June rebuttal\/revision window to the early-July notification, what reviewers score in a microarchitecture paper, the professional-conduct rules, and how the new Industry Track review differs."
}
MICRO Review Process
Model of the pipeline a MICRO submission enters, anchored to the 2026 (59th) cycle
as verified 2026-07-08. Committee composition and internal mechanics vary by year
and are only partly public — items marked 待核实 must be reconfirmed on
microarch.org/micro59/ before being repeated as fact.
2026 cycle timeline (verified)
| Gate | Date | Notes |
|---|---|---|
| Abstract registration | March 31, 2026, 11:59 PM EDT | Hard gate for paper upload |
| Full paper | April 7, 2026, 11:59 PM EDT | HotCRP: micro2026.hotcrp.com |
| Rebuttal / revision window | June 3–17, 2026 | Two weeks — unusually long |
| Notification | July 7, 2026 | |
| Camera-ready | September 11, 2026 | |
| Symposium | Oct 31 – Nov 4, 2026, Athens |
The eight-week gap between upload and rebuttal is when reviews are written; the three weeks after the window close is discussion and the decision meeting.
What the process is known to enforce
- Double-blind, end to end. Anonymity obligations persist through the rebuttal/revision window; artifact links stay anonymized the whole time.
- Format policing with teeth. Submissions are inspected visually and
automatically; violations can reject a paper even after the HotCRP checker
passed it (see
micro-submission). - A professional-conduct clause covering both directions. The 2026 guidelines state that abusive or inappropriate language in reviews, comments, or rebuttals can be grounds for dismissing reviewers, rejecting papers, and reporting to ACM and IEEE. Anger in a rebuttal is not just ineffective at MICRO — it is enumerated as a rejection ground.
- A separate Industry Track review. The inaugural 2026 track advertises a review process tailored to production work — evidence standards that accommodate proprietary constraints. Main-track novelty framing and industry-track deployment framing are different games; know which one your paper entered.
What reviewers weigh in a microarchitecture paper
Composite of the community's public reviewing norms (not a quoted rubric — 待核实 against any published review form for the current cycle):
- Mechanism novelty and clarity — is there a new hardware idea, stated at implementable precision?
- Methodology credibility — simulator, baseline strength, workloads, sampling; the fastest score-killer in this community.
- Evidence-claim match — simulation claims sized to simulation fidelity
(
micro-experimentsinstrument ladder). - Cost honesty — storage/area/power/latency accounting present and plausible.
- Significance — does the headroom captured justify silicon change; would anyone build it?
A practical corollary: reviewer expertise at MICRO skews toward methodology forensics. Expect at least one review that reconstructs your simulator config from the paper and interrogates the discrepancies.
Reading the outcome space
MICRO 2026's public timeline names one rebuttal/revision window and one notification date; whether the cycle offered conditional accepts or shepherding is 待核实. Plan against the simple model:
accept -> camera-ready sprint (Sept 11) + artifact evaluation decision
reject -> triage: fixable-methodology vs wrong-venue vs insufficient-headroom
next slots: ASPLOS Sept 9 '26 · ISCA Nov 17 '26 · HPCA July '27 · MICRO Apr '27
resubmit rule: fix what reviews falsified, re-verify every date on live pages
Keep the reviews regardless of outcome — architecture PCs overlap heavily across MICRO/ISCA/HPCA/ASPLOS, and a resubmission that ignores prior reviews may meet its own reviewer again.
Anatomy of a typical architecture review
Expect each review to contain, in some order: a summary paragraph (check it — a wrong summary means the paper failed to communicate, and the rebuttal's first job is a polite correction), a strengths list, a weaknesses list where methodology items dominate, and detailed questions. Useful triage signals:
- A review whose weaknesses are all questions is winnable in rebuttal; a review whose weaknesses are all assertions needs numbers to move.
- The most senior-sounding review is not automatically the decision driver; the review with the most specific methodology engagement usually is.
- Uniform mid-scores with shallow text often mean the paper landed outside every reviewer's exact expertise — a topic-checkbox lesson for the resubmission.
Score scales, reviewer counts, and whether an online discussion precedes the PC meeting are internal mechanics that vary by year: 待核实, do not assume.
Leverage points for authors
- Before submission: topic checkboxes in HotCRP (steer toward mechanism expertise), abstract quality (bidding happens on it), format cleanliness.
- During review: nothing. Do not update linked repos in identifying ways; the submission is frozen.
- Rebuttal window: the only mid-process input — see
micro-author-responsefor how to spend two weeks well. - After notification: AE participation (badges strengthen the published record), Top Picks nomination season (IEEE Micro selects from the year's architecture papers).
Calibrating expectations
Flagship-architecture acceptance rates historically sit in the high-teens to
low-twenties percent (exact per-year figures 待核实 from PC-chair reports or the
proceedings front matter — do not quote a number without a source). The planning
consequence is not pessimism but lattice thinking: a strong paper's expected
path to publication is one to two cycles across the four venues, which is why
micro-workflow treats every submission as having a plan B with a date on it,
and why reviews — even from a rejection — are the highest-value artifact the
process returns.
Output format
[Stage] pre-submission / under review / rebuttal window / post-notification
[Days to next gate] <N days to date, per verified timeline>
[Reviewer-lens self-score] mechanism / methodology / evidence-match / cost / significance: 1-5 each
[Conduct check] rebuttal tone professional: yes / revise
[Outcome plan] accept path + reject path both written: yes / no
[待核实 before quoting] <items to reconfirm on microarch.org>
Version History
- 9f86f09 Current 2026-07-19 16:58
