Use when revising an ACM MM (ACM Multimedia) paper for house style — putting the cross-modal contribution on the first page, framing media (figures, video, audio) as evidence rather than decoration, making the fusion the visible claim, and compressing the argument into a 6-8 page ACM sigconf body with references-only overflow.
SkillScore82Maintained
Automatic
◆Writing & Editing pack
Last verified 8/15/2026
What it changes
Passed 18 of 22 test cases with the skill, and 14 of 22 without it.
Without skill64%
With skill82%
Checks it passed that the model alone missed
Marks [First-page verdict] as 'buried' because the cross-modal/multimedia contribution does not appear on page one.
Identifies 'vision-paper voice' as a failure mode for framing the paper purely as a single-modality benchmark race.
Identifies 'modality as garnish' as a key failure mode where secondary modalities are mentioned but not ablated or shown to matter.
and 2 more in the full report
5 fixed13 passed either way3 still failing1 broke
Pass-rate lift
+18 pp
Tokens
+35%
Agent turns
0%
Response time
-26%
Model gemini-3.6-flashMeasured 8/15/2026v118 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 5 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 5
Prompt
In our paper draft, the audio-visual fusion mechanism is introduced on page 4, while page 1 focuses entirely on generic single-modality image classification history. Rate the first-page verdict using:
[First-page verdict]
[Fusion visibility]
[Media evidence]
[Scope]
[Page budget]
[Top three fixes]
What was checked, without and with the skill
Checked
Without skill
With skill
Marks [First-page verdict] as 'buried' because the cross-modal/multimedia contribution does not appear on page one.
Missed
Passed
case-16 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md100 linessingle file
ACM MM Writing Style
Use this to revise an ACM Multimedia draft for the venue's expectations. The reader is a
busy reviewer scanning across sixteen thematic areas; the paper must announce what is
multimedia about it before the model diagram.
The ACM MM first-page arc
Lead the abstract and first paragraph with: problem → why one modality is insufficient →
the cross-modal method or system → media-grounded evidence → why it matters for
multimedia. The multimedia contribution belongs on page one; a paper that opens with a
single-modality benchmark reads as a CVPR or ACL paper that wandered in.
State the fusion or systems mechanism as the contribution, not a backbone swap.
Make every media element do work: a teaser figure that shows the cross-modal signal, a
video that is evidence for a claim, an audio clip that a reader can check.
Scope claims to what the evidence supports; "improves engagement" needs a measured
engagement result behind it.
Media-as-evidence table
Media element
Weak (decoration)
Strong (evidence)
Teaser figure
A pretty system diagram
The moment where modalities disagree and the method wins
Qualitative grid
Cherry-picked successes
Paired success/failure across modalities with captions that state the point
Supplementary video
"See our results"
A clip tied to a specific claim, with the baseline shown alongside
Audio sample
An unlabeled waveform
The case the vision-only baseline misses, annotated
Compression into the sigconf body
The 6–8 page body is short by ACM standards, and figures compete with text for space.
Push proofs, full protocols, extra qualitative media, and hyperparameter tables to the
supplement; keep the body's argument self-contained without them.
Write self-contained captions — a reviewer reading only figures and captions should
grasp the cross-modal claim.
Budget page space before writing: decide which two or three media elements earn body space
and which move to the supplement.
Use when building or auditing the related-work section of an ACM MM (ACM Multimedia) paper — covering the multimedia literature spread across vision, audio/speech, language, HCI/QoE, and systems, handling arXiv-speed concurrency, keeping citations double-blind, and verifying that cited "ACM MM papers" are ACM MM and not ICMR, MMSys, CVPR, or TOMM.