Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →brycewang-stanford/acmmm-reproducibility
Use when strengthening the reproducibility of an ACM MM (ACM Multimedia) paper or preparing for the ACM MM Reproducibility track and ACM artifact badging — capturing environments, media/data access, seeds, and multimodal pipelines so an independent reviewer can rebuild results and reach Artifacts Evaluated or Results Reproduced badges.
Passed 21 of 22 test cases with the skill, and 3 of 22 without it.
and 19 more in the full report
18 fixed3 passed either way1 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 18 test cases the skill fixed. Same model, same prompt, with and without the skill.
We set up an anonymous data link for double-blind review on an internal university server that requires a password request via email to the lab technician. Evaluate our data access strategy for ACM Multimedia main track review and present the findings using bracketed output lines.
| Checked | Without skill | With skill |
|---|---|---|
| Flags requiring password requests via email as a double-blind anonymity leak and a broken access path. | Passed | Passed |
| Uses the tag '[Data access]' in the diagnostic report. | Missed | Passed |
The instructions your model reads when the skill loads.
Use this to make an ACM Multimedia result rebuildable — both for main-track credibility and for the dedicated Reproducibility track, which routes artifacts through ACM's badging pipeline. Multimedia adds a wrinkle: the data is often video, audio, or interactive media, and "run the code" is not enough if a reviewer cannot obtain or render the media.
ACM's artifact model distinguishes availability, evaluation, and reproduction. Map your goal to the badge you are actually pursuing:
| Badge (ACM terminology) | What it asserts | What you must ship |
|---|---|---|
| Artifacts Available | The artifact is publicly, permanently retrievable | A DOI/archived repository with the code and media pointers |
| Artifacts Evaluated (Functional/Reusable) | Reviewers ran it and it works / is reusable | Build + run instructions, environment, documentation |
| Results Reproduced | An independent team reproduced the paper's results | A pipeline that regenerates the reported numbers/media |
Confirm the exact badge set offered for the current cycle on the Reproducibility-track call; ACM's badge names and criteria evolve.
Keep a single record that ties each reported result to the code, data, and config that produced it:
result: Table 2, row "full model"
code commit: <hash>
config: configs/full.yaml
data: <dataset name + version + anonymous mirror for review>
media preprocessing: <fps, sample rate, caption source>
seed(s): <list>
hardware: <GPU/CPU, hours>
expected output: results/table2_full.json
test cases passed, vs 3 / 22 without
Published skills whose descriptions closely match this one, with their headline benchmark lift.
More skills from the same source and nearby in the registry, with their headline benchmark lift.