RSS Review Process

Model how an RSS decision actually gets made. The mechanics below are the verified 2026 design (roboticsconference.org/reviewps/, checked 2026-07-08); the two-stage structure itself is cycle policy, so confirm it survives before relying on it.

The 2026 pipeline

OpenReview submission (double-blind)
        │
   first review round
        │
   threshold gate ── below ──▶ reject (no rebuttal opportunity)
        │ above
   invited one-page rebuttal (due 2026-04-03 AoE)
        │
   Area Chair reads rebuttal
        │  (forwards to reviewers only if extra feedback is needed)
   Area Chair meeting ──▶ decisions (2026-04-27 AoE)

Two properties distinguish this from ML-conference discussion phases:

• The rebuttal is earned, not granted. Papers below the first-round threshold are decided without author input. Consequence: the submission itself must pre-answer foreseeable objections; there may be no second chance (rss-writing-style).
• The audience of the rebuttal is the AC. Reviewers may never see it. Write for the person running the meeting, not for a back-and-forth thread (rss-author-response).

What single-track selectivity does to reviewing

Every accepted paper takes a slot in front of the whole conference, so the operative question in the AC meeting is closer to "must the field hear this?" than "is this sound?" Soundness is necessary, not sufficient. Practical consequences:

Reading a borderline packet

• Separate correctness objections (a flaw in the claim or protocol) from scope objections (the claim is narrower than sold) from significance doubts (sound but not slot-worthy). Only the first two are rebuttal-addressable with existing evidence; the third is won or lost in the submission's framing.
• A lone enthusiastic review does not carry a paper through an AC meeting; a lone unanswered correctness objection can sink one. Triage accordingly.
• Decisions emerge from a synchronous meeting, so the AC's summary understanding is the decision artifact. Everything you write should compress well into two sentences an AC might say aloud.

Timeline anatomy (2026 verified dates)

The compressed April-to-July runway is a distinctive RSS pressure: camera-ready, travel authorization, and artifact release all share eleven weeks.

Calibrating expectations by venue contrast

• Unlike venues with universal author-response phases, RSS's threshold gate means a silent April is itself information — but never certain information, since invitation lists are confidential.
• Unlike rolling-review journals, there is no revise-and-resubmit: the meeting outputs accept or reject, and the improvement loop runs through the next venue slot (rss-workflow).
• Scores, forms, and thresholds are internal; public acceptance statistics were not verified for 2026 (待核实) — do not quote community folklore rates as fact.

Confidentiality and conduct

• Reviewer identities, scores, and thresholds are confidential; do not attempt identification or out-of-band contact.
• Notification wording in 2026 covered "paper and demo acceptance" even though the CFP had no demo-paper category — treat demo mechanics as 待核实 rather than inferring a track.

Who sits in the reviewer pool

• The pool spans planning, control, perception, learning, and mechanism design; topic matching at a smaller venue is tight, so expect at least one reviewer who has personally run experiments like yours — protocol hand-waving gets caught.
• Because the audience is the whole field, at least one review often comes from an adjacent subfield; jargon-dense writing loses exactly the vote the single-track format makes decisive.
• Area Chairs are established researchers appointed per edition; the AC's synthesis carries the meeting, which is why every skill in this pack optimizes for the AC-compressible summary.

Output format

[Stage] under review / rebuttal invited / no rebuttal / AC meeting pending / decided
[Objection triage] correctness: <...> | scope: <...> | significance: <...>
[AC-compressible summary] <two sentences the AC could say in the meeting>
[Rebuttal leverage] high / low / none (below threshold)
[Next move] <one action>