Agent Skills
› brycewang-stanford/Awesome-Journal-Skills
› hri-supplementary
hri-supplementary
GitHub规划ACM/IEEE HRI论文补充材料，重点指导视频制作、匿名化处理及双盲审查合规。依据决策关键性原则，区分正文与附录内容，确保核心论证在8页内完整呈现，补充材料包含仪器、结果扩展等辅助信息。
触发场景
规划HRI会议论文的补充材料
准备双盲审查所需的视频和附录
决定哪些内容放入正文或补充文件
安装
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill hri-supplementary -g -y
SKILL.md
Frontmatter
{
"name": "hri-supplementary",
"description": "Use when planning the supplementary materials for an ACM\/IEEE HRI full paper — above all the video figure that shows the embodied interaction, plus study instruments, extended results, and appendices — deciding what a reviewer must see to judge the interaction versus what belongs outside the 8-page body, all anonymized for double-blind review."
}
HRI Supplementary Materials
HRI is unusual among conferences in how much the video matters. The interaction it studies is temporal and embodied, and prose cannot convey what a robot's timing, motion, and presence do to a person. A well-made supplementary video figure materially changes how reviewers understand the contribution — and its absence is often read as a gap. This skill plans the video and the rest of the supplement, and splits content between the body and the supplement by decision-criticality.
The video figure comes first
Treat the video as a first-class artifact, not an afterthought:
- Show the interaction, not just the robot: the person and robot together, the conditions being compared, and the behavior your contribution designs or studies. A reviewer should see the difference your paper claims.
- Keep it short and legible — a focused clip (often a couple of minutes; check the current call for length/size limits, 待核实 per cycle) beats a long unedited recording. Caption or narrate what to look at.
- Anonymize it for review — this is the most-missed double-blind leak. Blur or omit identifiable
faces, mute or remove identifying voices and lab/institution logos, and strip filename/metadata
that names an author. If participants did not consent to public video, use a review-only,
de-identified cut (see
hri-reproducibility). - Format for reviewers — a widely playable codec/container and a size the submission portal accepts; a video a reviewer cannot open is a video that does not exist.
The decision-criticality rule
Ask of every piece of content: could a reviewer reach the accept/reject decision without seeing this?
- No → it belongs in the 8-page body. The core design, the study design, the primary results and their effect sizes, the main figures, and the central limitation must be in the paper. Reviewers are not required to read the supplement to judge the paper, and decision-critical material hidden in an appendix is effectively invisible.
- Yes → it can go to the supplement. Full instrument text, extended tables, additional analyses, qualitative excerpts, and implementation detail live outside the body.
The 8-page budget (incl. figures, excl. references) forces this discipline; use it to keep the body sharp, not to smuggle the argument into an appendix.
What the supplement typically holds
- Video figure (the centerpiece) and, if relevant, a short demo clip.
- Study instruments — full questionnaires with items and scoring, interview guides, task instructions, consent-form summary.
- Coding schemes and qualitative appendices — codebooks, extended excerpts, additional themes.
- Extended results — full statistical tables, robustness/secondary analyses, per-condition breakdowns that would crowd the body.
- Robot-behavior details — parameters, WoZ protocol specifics, additional apparatus figures.
- Reproducibility artifact pointer — the (anonymized) archive of de-identified data and analysis
code (see
hri-reproducibility).
Keep the body self-sufficient
- Every claim that decides the paper must be supported in the body; the supplement enriches, it does not carry.
- Reference supplementary items explicitly ("full scale in the supplement," "see supplementary video"), so reviewers know they exist — but never make a core inference depend on following the link.
- Do not use the supplement to exceed the page limit by relocating essential text; reviewers and chairs notice.
Anonymization sweep for the supplement
Double-blind covers everything uploaded, not just the PDF:
- Video — faces, voices, logos, on-screen names, file metadata.
- Data files — free-text fields, participant identifiers, embedded author metadata.
- Code/materials — author names in headers, institution paths, repository owners.
- Archive — host behind an anonymizing view; a personal domain both rots and reveals identity.
See hri-submission for the mechanical checklist to run on the final files.
Anti-patterns HRI reviewers flag
- No video for a paper about embodied, temporal interaction.
- A video that leaks identity (faces/voices/logos) in double-blind review.
- Decision-critical results only in the supplement, leaving the body unable to stand alone.
- An unplayable or oversized video the portal or reviewer cannot open.
- Appendix used to dodge the page limit by relocating essential argument.
Output format
[Video] present? shows the interaction/conditions? length/format within limits?
[Video anonymization] faces/voices/logos/metadata clean for double-blind?
[Body sufficiency] all decision-critical content in the 8 pages? yes/no
[Supplement contents] instruments · coding schemes · extended results · artifact pointer
[Anonymization] video + data + code + archive all swept?
[Fix queue] <ordered>
版本历史
- 9f86f09 当前 2026-07-19 15:51
