iros-review-process

IROS Review Process

Use this to reason about review-stage strategy. Reopen the current CFP, PaperPlaza instructions, and any

reviewer guidelines before making process claims — IROS review mechanics are set per edition by the

co-sponsoring IEEE/RSJ committee.

Process model

• Submission and review run on PaperPlaza. Papers are handled by an Associate Editor under an

Editor, who assign and synthesize reviews; there is no standing editor-in-chief for the conference.

• Reviewers are embodied-systems researchers. They weigh hardware, task distributions, trial counts,

failure analysis, and sim-to-real honesty at least as much as a clean plot.

• The video is read as evidence: many reviewers watch it before the text, so a weak or misleading

video colors the whole review.

• The 2026 cycle is double-anonymous; author identity should not be inferable from text, figures, or

video.

• Accepted papers go to IEEE Xplore, so camera-ready compliance and the copyright form matter as much

as the initial decision.

The no-rebuttal consequence

IROS has traditionally delivered reviews together with the decision and no author-response phase.

This changes strategy fundamentally: you cannot answer an objection after the fact, so **every

foreseeable objection must be pre-answered in the paper and the video**. The video functions as the only

asynchronous "rebuttal" a reviewer gets — a shown-and-labeled failure disarms the "did it really work?"

reviewer before they can write it down. Confirm the current cycle's policy before assuming a rebuttal

exists or does not.

Scoring leverage table

| Review dimension | What raises it | What sinks it |

|---|---|---|

| Technical merit | A system that integrates cleanly and runs under a real constraint | A component swap dressed up as a system |

| Evidence | Trials, resets, failures, and a stated sim-to-real gap | A hero demo with no counts and no failures |

| Clarity | A system diagram and figures that carry the argument | An argument that needs an appendix IROS does not provide |

| Reproducibility | Hardware ledger, logs, honest availability | "Code coming soon" with no specs |

Stage-by-stage realism

• Submission to notification is a long silence (spring to mid-June in 2026); there is no interaction

to manage, only evidence to preserve for camera-ready.

• Notification arrives with the reviews; with no rebuttal, the lever is the camera-ready revision and

the meta-review's requested changes.

• A single unresolved "I don't believe the robot did this" concern outweighs several resolved clarity

nits, because embodied credibility is the currency here.

Reject-to-next-venue realism

Because IROS is on the fall calendar, a reject can route to the next ICRA, to CoRL, or to an RA-L

submission with revision — decide by the reviews' substance, not by prestige.

Reading a review packet:
  [ ] separate "the robot didn't convince me" from "writing unclear"
  [ ] list objections that a camera-ready edit can address vs cannot
  [ ] note every evidence gap the reviewers named for the next run
  [ ] pick the re-route (ICRA / CoRL / RA-L) if rejected, by substance

Output format

[Current stage] submitted / in-review / notified / camera-ready
[Decision actors] <Associate Editor / Editor / reviewers>
[Likely leverage] <integration / evidence / clarity / reproducibility>
[No-rebuttal implication] <what must be pre-answered in paper + video>
[Next move] <camera-ready edit or re-route>