Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› podc-review-process
podc-review-process
GitHub解析ACM PODC会议论文评审流程，涵盖轻量双盲审稿、无修改轮次、证明优先原则及接受/拒绝决策。对比DISC、SPAA等会议差异，指导作者确保初始提交包含完整证明与模型假设，以应对严格审查。
Trigger Scenarios
询问PODC论文评审标准
了解PODC与DISC/SPAA区别
确认是否需要准备修改稿
查询PODC审稿匿名机制
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill podc-review-process -g -y
SKILL.md
Frontmatter
{
"name": "podc-review-process",
"description": "Use when reasoning about how an ACM PODC submission is evaluated — covering lightweight double-blind review, a program committee that reads proofs, the accept\/reject decision with no journal-style revision round, the Brief Announcements track's lighter bar, and how PODC's process differs from DISC, SPAA, and the sequential-theory venues STOC\/FOCS\/SODA."
}
PODC Review Process
Model the pipeline before interpreting any single review. PODC's process is proofs-first and single-shot: a specialized program committee reads your model and your proof, and the decision is accept or reject — there is no Major Revision, no journal-style R&R. The most consequential mental shift for authors arriving from a software-engineering or ML venue is that there is no second read to fix a gap in; everything that would go into a revision must be in the initial submission.
Process model
- Submission and review run on HotCRP with lightweight double-blind anonymity: identities are hidden from reviewers (no author names/affiliations/emails in the submission), a change from PODC's historically single-blind practice.
- Each paper is read by program-committee members (often with external subreviewers for specific subareas) who weigh: the significance of the result, the correctness of the proof, the precision and honesty of the model/assumptions, novelty against known bounds, and clarity.
- The committee reads the abstract and the first 10 pages as guaranteed; deeper proofs are read
at its discretion — so the merits case must land in those pages (
podc-supplementary). - Decisions are accept / reject. Accepted regular papers publish in the ACM proceedings (≤10 pages); Brief Announcements (≤3 pages) are a separate, lighter-bar track.
- There is typically no revision round; whether a rebuttal / author-response phase runs varies by cycle (待核实 for 2026 — confirm on the call).
What a PODC reviewer actually checks
| Dimension | The reviewer's question | How to pass it |
|---|---|---|
| Correctness | Does the proof close in the stated model? | Self-contained proofs; no assumption creep |
| Significance | Does this advance a problem the community cares about? | Position against the closest known bounds |
| Model honesty | Are the timing/fault/adversary assumptions declared and consistent? | An explicit model box the proofs never violate |
| Novelty | Is the result / technique new relative to PODC/DISC/journals? | A precise delta, not a topic list |
| Tightness | Is an "optimal" claim backed by a matching lower bound? | Prove or cite the lower bound, or don't claim optimal |
A reviewer who finds a proof gap or an undeclared assumption will reject on soundness, and
there is no revision to repair it — which is why podc-experiments (the assumption audit) and
podc-reproducibility (self-contained proofs) are pre-submission, not post-notification, work.
How PODC differs from its siblings
- vs. DISC: DISC is the other distributed-theory flagship (EATCS, LIPIcs proceedings, fall cycle) and shares the Dijkstra Prize and Dissertation Award with PODC. The review culture is close; the differences are calendar, proceedings format, and community center of gravity, not the proofs-first bar.
- vs. SPAA: SPAA centers parallel computation; a message-passing fault-tolerance result is native at PODC and a visitor at SPAA (and vice versa for shared-memory throughput/scheduling).
- vs. STOC/FOCS/SODA: these judge sequential algorithms and complexity; a PODC paper's value must live in the distributed difficulty (faults, asynchrony, locality, concurrency) or it will be read as a mislabeled sequential result.
- vs. software-engineering / ML venues: no artifact evaluation, no rebuttal-driven Major Revision, no empirical-benchmark culture. The object of review is the proof.
Where author leverage actually exists
[Before submission] topic tags + a precise abstract -> reviewers who know your model (largest lever)
[Track choice] regular vs. Brief Announcement -> match the bar to the result's completeness
[In the submission] a self-contained proof and an honest model box -> pre-empt the soundness reject
[Rebuttal, if any] correct a factual misreading or supply a missing lemma/bound (待核实 it exists)
[After reject] no appeal; revise the proof and reroute to DISC/SPAA/a focused venue or a journal
Because there is no revision round, the leverage is almost entirely pre-submission. The single highest-value action is closing every proof and declaring every assumption before 16 February.
Reading a decision
| Outcome | What it means | Author move |
|---|---|---|
| Accept | Result, proof, and model hold | Camera-ready (≤10 pages) + arXiv full version; do not weaken claims |
| Accept as Brief Announcement (if offered) | The idea is valued but not a full regular paper this cycle | Publish the ≤3-page BA; develop the full paper for a later venue |
| Reject — soundness | A proof gap or undeclared assumption | Fix the proof; do not resubmit unchanged |
| Reject — significance/novelty | Known bound, or the delta is too small | Strengthen (a matching lower bound?) or reroute |
Misreadings to avoid
- Expecting a revision round — there is none; the submission is the whole case.
- Treating a preprint as forbidden — arXiv posting is allowed; only concurrent submission elsewhere is the dual-submission problem.
- Assuming a rebuttal exists — confirm per cycle (待核实 for 2026).
- Under- or over-anonymizing — follow the lightweight double-blind wording, neither leaking identity nor stripping citations.
Output format
[Process stage] pre-submission / under review / (rebuttal if any) / decided
[Anonymity model] lightweight double-blind — submission clean of author identity?
[Reviewer-axis map] each review point -> correctness | significance | model-honesty | novelty | tightness
[Leverage plan] the pre-submission or (rebuttal) action that can actually change the outcome
[Forbidden moves] identity leak / claiming a revision round that does not exist / unsupported optimality
Version History
- 9f86f09 Current 2026-07-19 17:16
