Mobisys review process

MobiSys Review Process

Use this to reason about review-stage strategy. Reopen the current CFP, the HotCRP site, and the organizing-committee page before making process claims — the mechanics are cycle-specific.

Process model

• MobiSys uses HotCRP for submission and review, with a paper-registration step preceding the paper upload and double-blind review (author identities hidden since 2017).
• Review runs in two rounds. Papers that do not advance past round 1 receive an early rejection with reviews, so authors can re-plan before the process ends. Round-2 survivors reach reviews and a rebuttal window.
• The rebuttal is scope-limited: correct factual errors and answer specific reviewer questions; new results are admissible only when directly responsive, or may be promised for camera-ready (mobisys-author-response).
• Unlike MobiCom and NSDI, the rendered 2026 CFP describes accept/reject through the rebuttal, not a separate one-shot revision channel; treat any revision path as 待核实 per cycle.
• Accepted papers are published in the ACM Digital Library, so camera-ready compliance and artifact badges matter as much as the initial accept.

Who reviews here

• The pool is systems-and-services researchers: people who build and measure mobile runtimes, on-device ML, sensing services, and platforms, and who read an energy or latency claim as something to be re-derived, not admired.
• Because MobiSys is specialized, topical matches are close; a hand-wavy device claim or an undefined energy boundary gets caught rather than skimmed past.
• Borderline mobile-systems papers usually fall on one of three edges: an evaluation on a single device or single run, an energy/thermal claim without a described instrument, or a contribution that is really a better model dressed as a system.

Scoring leverage table

Review dimension	What raises it	What sinks it
Contribution	A system mechanism forced by the device constraint	A model or algorithm with an incidental phone demo
On-device evidence	Distributions on real hardware across devices/states	Single-device, single-run, simulation-only numbers
Energy/latency rigor	Named instrument, boundary, and sustained-load behavior	"Efficient" with no measured joules or throttle trace
Clarity	An explicit operating point (device, workload, budget)	Latency numbers with no device or workload context

Stage-by-stage realism

• Round 1: triage by what a systems reviewer would weigh — is the device evidence real and is the contribution a system? A thin evaluation is the classic early reject.
• Rebuttal: windows are short and scope is narrow; a precise reply that fixes a factual error and answers the one decision-critical question beats an exhaustive one.
• Decision: one unresolved evidence objection (single-device, undefined energy boundary) outweighs several resolved clarity complaints.
• After accept: artifact evaluation and camera-ready are their own gates (mobisys-artifact-evaluation, mobisys-camera-ready).

Output format

[Current stage] submitted / round-1 / round-2 reviews / rebuttal / decision / camera-ready
[Decision actors] <reviewers / AC / chairs>
[Likely leverage] <contribution / on-device evidence / energy-latency rigor / clarity>
[Forbidden moves] <identity leak / out-of-scope new results in rebuttal>
[Next response move] <one action>