Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› ipsn-review-process
ipsn-review-process
GitHub解析IPSN论文评审流程，涵盖双盲、分赛道（IP/SPOTS）审稿、反驳阶段及最佳论文评选。对比SenSys等会议差异，指导作者根据审稿信号优化证据真实性与基线公平性，应对拒稿风险。
Trigger Scenarios
询问IPSN或相关会议的审稿标准
分析论文被拒原因及改进建议
了解双盲评审和反驳阶段策略
区分IP与SPOTS赛道的关注点
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill ipsn-review-process -g -y
SKILL.md
Frontmatter
{
"name": "ipsn-review-process",
"description": "Use when reasoning about how an IPSN-lineage submission is evaluated, covering double-blind review, the per-track (IP \/ SPOTS) program committees, the rebuttal, Best Paper and Best Research Artifact judging, and how IPSN's process differs from SenSys's revision model, OpenReview venues, and CPS-IoT Week neighbors."
}
IPSN Review Process
Model the pipeline before interpreting any single review. IPSN's process is double-blind, per-track, and conference-style (not journal-style): a submission is read by a program committee matched to its track — IP or SPOTS — and returns an accept/reject decision, usually with a rebuttal opportunity. Because IPSN merged into SenSys, confirm the successor's exact mechanics on the current call; the structure below is the IPSN-lineage model and what to expect.
Process model
- Submission and review run on HotCRP with double-blind anonymity: author identities are hidden from reviewers and reviewer identities from authors.
- Papers are matched to reviewers by track. An IP-track paper is read by method reviewers
(estimation, signal processing, learning, localization); a SPOTS-track paper by
platform/systems reviewers (hardware, embedded software, tools, deployment). This is why the track
choice in
ipsn-submissionmatters so much. - Reviewers weigh: the soundness of the information-processing method or platform design; whether the evidence is real (real sensors, ground truth, measured energy/latency, honest deployment numbers); novelty against the sensing literature; and reproducibility / artifact support.
- A rebuttal typically lets authors correct factual misreadings before the decision (verify the window and format on the current call).
- Accepted papers appeared in both ACM DL and IEEE Xplore; the successor publishes regular papers in the ACM proceedings and demos/posters in the IEEE proceedings.
What each track's reviewers check first
| Track | Reviewer's first question | Common reject trigger |
|---|---|---|
| IP | Is the estimator/inference sound, and is the baseline fair? | Simulation-only, or a proxy metric standing in for the real sensing outcome |
| SPOTS | Is the platform/tool reusable, and are the design trade-offs measured? | A one-off build with a datasheet but no measured power/robustness story |
| Either (deployment) | Are the real-world hardships reported honestly? | Yield, synchronization, and energy numbers missing or idealized |
Reading a decision against the criteria
| Signal in the reviews | What it means | Author move |
|---|---|---|
| "Only simulated / no real hardware" | Evidence-realism doubt (fatal at IPSN) | If possible add a real-sensor result in rebuttal; otherwise reroute |
| "Baseline is not a real alternative / untuned" | Soundness doubt | Add or justify a fair baseline; report the comparison |
| "Deployment numbers look idealized" | Honesty/realism doubt | Report yield, sync error, energy as measured, with limits |
| "Artifact would strengthen this" | Reproducibility gap | Commit to (and anonymize) a firmware+dataset artifact |
| "Wrong track / out of scope" | Track or venue mismatch | Hard to fix in rebuttal; a ipsn-topic-selection lesson for next time |
How IPSN differs from its neighbors and successor
- vs. SenSys (pre-merger): SenSys is the sibling embedded-networked-sensing flagship; IPSN's distinctive move was the IP/SPOTS split and its information-processing/estimation flavor. Post-merger the two communities share one venue — but an IP-track style paper is still judged on its estimation/inference soundness.
- vs. OpenReview ML venues: IPSN is not open-review, not score-thread public, and not leaderboard-driven. Offline accuracy on a clean dataset does not carry a paper here; on-device or in-field evidence does.
- vs. CPS-IoT Week neighbors (RTAS/HSCC/ICCPS): those reviewers want timing guarantees, control theory, or hybrid-systems verification. An IPSN paper is judged on sensing/information-processing soundness and real measurement, not worst-case schedulability.
Where author leverage actually exists
[Before submission] track choice + topic tags -> reviewer pool (largest lever)
[Initial reviews] factual corrections, a real-hardware number a reviewer said was missing
[Rebuttal] narrow, evidence-backed answers to soundness/realism doubts
[After reject] no journal-style guaranteed revision round; reroute or resubmit next cycle
A rebuttal moves borderline papers when it corrects a misread table or supplies a measured number a reviewer flagged; it does not move papers that argue taste or promise experiments not yet run.
Best Paper and Best Research Artifact judging
IPSN gave a Best Paper Award and a Best Research Artifact Award. The artifact award is a
distinct incentive: a firmware+dataset package that an evaluator can actually run and reuse is judged
on more than the paper's claims. Target it deliberately (ipsn-artifact-evaluation) — verify whether
it persists under the successor (待核实).
Misreadings to avoid
- Expecting a journal-style Major Revision — IPSN is conference-style accept/reject with a rebuttal, not a guaranteed revise-and-resubmit round (unlike a journal or a Major-Revision venue).
- Treating the rebuttal as a debate — the PC discussion decides; the rebuttal is evidence for an advocate, not a closing argument.
- Assuming the successor keeps IPSN's exact mechanics — the merged SenSys may differ; confirm.
Output format
[Process stage] pre-submission / awaiting reviews / rebuttal / decision / accepted
[Track] IP / SPOTS (or successor category)
[Criterion map] each review point -> soundness | evidence-realism | novelty | reproducibility | track-fit
[Leverage plan] the next-stage action that can actually change the outcome
[Forbidden moves] identity leak / unrun experiments promised as done / arguing taste
Version History
- 9f86f09 Current 2026-07-19 16:04
