Revise the manuscript for the venue where "the robot worked" is the beginning of a paper, not its conclusion. RSS expects a scientific claim; the writing's job is to make that claim visible, falsifiable, and already defended.
By the end of page one, a reader should be able to fill in:
If the draft's first page instead inventories modules and integrations, it is a
systems report; either surface the insight the integration proves, or reroute
(rss-topic-selection).
The cardinal RSS writing sin is claim-evidence inflation: general words resting on narrow trials. Enforce the match at sentence level:
| Sentence habit | Failure it causes | Repair |
|---|---|---|
| "robust manipulation of everyday objects" | Generality unearned by 5 objects | Name the object distribution and count |
| "significantly outperforms" | No statistics behind the adverb | Effect size + interval + trials per condition |
| "works in the real world" | One lab bench ≠ the world | State the environments actually tested |
| "can be applied to X, Y, Z" | Speculation formatted as result | Move to a labeled limitations/outlook sentence |
| "novel framework" | Novelty asserted, not located | One sentence naming the delta vs the nearest method |
Only above-threshold papers got a 2026 rebuttal; write as if you will not. For each major claim, plant the defense where the objection arises:
The 8-page limit (excluding references, 2026) is a ceiling the CFP pointedly frames as not a floor. Compression tactics that preserve rigor:
A workable allocation for a claim-plus-hardware paper (adjust, don't worship):
p.1 claim, stakes, falsification test, contribution list
p.2 related work as contrasts (not a tour)
p.3-4 method: the mechanism, then the system carrying it
p.5-6.5 experiments: protocol, results, ablations, failure table
p.6.5-7.5 analysis: what the failures say about the claim
p.7.5-8 limitations with scope sentences; conclusion in 4 lines
refs excluded from the count (2026 rule)
Two structural rules the budget encodes: analysis gets real space (RSS reviewers reward interpretation over raw scoreboards), and limitations are body text, not a camera-ready afterthought.
| Draft reflex | RSS register |
|---|---|
| "To the best of our knowledge, first to..." | Name the nearest prior and state the delta |
| "Extensive experiments demonstrate..." | "600 trials across 6 conditions show..." |
| "We leave X for future work" (for a load-bearing X) | Test X or narrow the claim now |
| "Our approach is general" | "The mechanism assumes A and B; within them..." |
| "Qualitative results are in the video" | Put the quantitative version in the body |
RSS's reviewer culture reads failure reporting as competence, not weakness. The stylistic move: report failures with the same grammatical confidence as successes. "CSD fails on high-friction fabrics (9/40 trials), consistently at the release step" is stronger writing than a defensive footnote — it demonstrates the authors know where the boundary is.
RSS papers span perception, planning, control, and learning in eight pages, and vocabulary collisions cost belief:
[Claim sentence] <one sentence, falsifiable, system-name-free>
[Claim-evidence mismatches] <sentence -> repair>
[Unplanted defenses] <objection -> where to pre-answer it>
[Compression targets] <cut / merge / demote-to-supplement>
[Failure register] present and confident / defensive / missing