HPCA Supplementary
Use this to triage material across the destinations HPCA actually provides. HPCA has
no open-ended supplementary upload culture: material lives in one of four places, and
putting it in the wrong one either wastes body pages or leaks anonymity.
The four destinations
- The 11-page body. Only what the argument needs to stand: the characterization,
the mechanism, and the headline evidence. Everything a reviewer must read to
believe the claim.
- The appendix (within the submitted PDF's allowance). Extended methodology,
additional sensitivity studies, proofs of any modeled bounds — material a reviewer
may consult, kept out of the core argument.
- The anonymized artifact mirror. Code, configs, scripts, and reduced-input
runs, with all identity scrubbed and links anonymized, so a reviewer can inspect
during the window.
- The post-acceptance public release. The de-anonymized, badge-ready artifact on
the AE HotCRP after acceptance, plus any licensed-workload recipes.
Where does each thing go?
| Material |
Destination |
Note |
| Mechanism description, headline results |
Body |
The paper's core |
| Extra sensitivity sweeps, extended methodology |
Appendix |
Referenced from the body |
| Simulator, configs, regen scripts |
Anonymized mirror |
Fully scrubbed for review |
| Machine-state logs, raw output |
Anonymized mirror / appendix |
Provenance for silicon |
| Licensed-workload recipes + checksums |
Post-acceptance release |
Cannot ship the suite itself |
| De-anonymized artifact + README |
Post-acceptance release |
Badge round on AE HotCRP |
Keep anonymity intact across destinations
The most dangerous leaks happen in the material around the paper:
- Scrub author names, institution paths, and repo owners from the mirror; anonymize
every artifact URL in the PDF.
- Strip identifying metadata from config files, commit authorship, and notebook
fields.
- Keep acknowledgments, funding, and any AI-use disclosure phrased so they do not
de-anonymize (the disclosure stays anonymous until camera-ready).
- Never blind the bibliography or drop a reference's authors to save space — that is
a reject trigger, not an anonymity fix.
Stage for the window
Between submission and the rebuttal/revision window, keep a staging shelf: extra
experiments a reviewer might request, pre-run and ready to fold in. The window is
short and combined (respond and revise), so material assembled in advance is the
difference between answering an objection and deferring it.
Triage pass
1. Body carries only what the claim needs — move the rest to appendix/mirror
2. Appendix holds referenced extras, not orphan material
3. Mirror is fully anonymized (names, paths, URLs, metadata)
4. Licensed workloads reduced to recipes+checksums for later release
5. Staging shelf holds pre-run likely-requested experiments
6. Bibliography intact: no blinding, all authors listed
Output format
[Body discipline] core-only? (Y/N)
[Mirror anonymity] fully scrubbed? (Y/N)
[Destination errors] <material in the wrong place>
[Window readiness] staging shelf populated? (Y/N)
[Bibliography] intact and author-complete? (Y/N)
[Fixes] <ordered>
Reopen the current CFP and AE page for the submitted-PDF appendix allowance and the
release mechanics — both are per-edition.
Source: brycewang-stanford/Awesome-Journal-Skills → HPCA-Skills/skills/hpca-supplementary/SKILL.md
1---2name: hpca-supplementary3description: Use when deciding where HPCA material belongs across the four destinations the venue offers — the 11-page body, the anonymized artifact mirror, the staging shelf for the rebuttal/revision window, and the post-acceptance public release — while keeping anonymity intact and every reference author-complete.4---5
6
7# HPCA Supplementary
8
9Use this to triage material across the destinations HPCA actually provides. HPCA has
10no open-ended supplementary upload culture: material lives in one of four places, and
11putting it in the wrong one either wastes body pages or leaks anonymity.
12
13## The four destinations
14
151. **The 11-page body.** Only what the argument needs to stand: the characterization,
16 the mechanism, and the headline evidence. Everything a reviewer must read to
17 believe the claim.
182. **The appendix (within the submitted PDF's allowance).** Extended methodology,
19 additional sensitivity studies, proofs of any modeled bounds — material a reviewer
20 *may* consult, kept out of the core argument.
213. **The anonymized artifact mirror.** Code, configs, scripts, and reduced-input
22 runs, with all identity scrubbed and links anonymized, so a reviewer can inspect
23 during the window.
244. **The post-acceptance public release.** The de-anonymized, badge-ready artifact on
25 the AE HotCRP after acceptance, plus any licensed-workload recipes.
26
27## Where does each thing go?
28
29| Material | Destination | Note |
30|---|---|---|
31| Mechanism description, headline results | Body | The paper's core |
32| Extra sensitivity sweeps, extended methodology | Appendix | Referenced from the body |
33| Simulator, configs, regen scripts | Anonymized mirror | Fully scrubbed for review |
34| Machine-state logs, raw output | Anonymized mirror / appendix | Provenance for silicon |
35| Licensed-workload recipes + checksums | Post-acceptance release | Cannot ship the suite itself |
36| De-anonymized artifact + README | Post-acceptance release | Badge round on AE HotCRP |
37
38## Keep anonymity intact across destinations
39
40The most dangerous leaks happen in the material *around* the paper:
41
42- Scrub author names, institution paths, and repo owners from the mirror; anonymize
43 every artifact URL in the PDF.
44- Strip identifying metadata from config files, commit authorship, and notebook
45 fields.
46- Keep acknowledgments, funding, and any AI-use disclosure phrased so they do not
47 de-anonymize (the disclosure stays anonymous until camera-ready).
48- Never blind the bibliography or drop a reference's authors to save space — that is
49 a reject trigger, not an anonymity fix.
50
51## Stage for the window
52
53Between submission and the rebuttal/revision window, keep a **staging shelf**: extra
54experiments a reviewer might request, pre-run and ready to fold in. The window is
55short and combined (respond *and* revise), so material assembled in advance is the
56difference between answering an objection and deferring it.
57
58## Triage pass
59
60```text
611. Body carries only what the claim needs — move the rest to appendix/mirror
622. Appendix holds referenced extras, not orphan material
633. Mirror is fully anonymized (names, paths, URLs, metadata)
644. Licensed workloads reduced to recipes+checksums for later release
655. Staging shelf holds pre-run likely-requested experiments
666. Bibliography intact: no blinding, all authors listed
67```
68
69## Output format
70
71```text
72[Body discipline] core-only? (Y/N)
73[Mirror anonymity] fully scrubbed? (Y/N)
74[Destination errors] <material in the wrong place>
75[Window readiness] staging shelf populated? (Y/N)
76[Bibliography] intact and author-complete? (Y/N)
77[Fixes] <ordered>
78```
79
80Reopen the current CFP and AE page for the submitted-PDF appendix allowance and the
81release mechanics — both are per-edition.
82
83---
84
85**Source:** [`brycewang-stanford/Awesome-Journal-Skills`](https://github.com/brycewang-stanford/Awesome-Journal-Skills) → `HPCA-Skills/skills/hpca-supplementary/SKILL.md`
thedixitjain