brycewang-stanford CoNEXT Supplementary
Decide where each piece of content lives: the reviewed body, the appendix budget, or the artifact. The governing rule at CoNEXT is decision-criticality — anything a reviewer must read to be convinced the paper should be accepted belongs in the body, not in an appendix a reviewer may skip or an artifact they may not open. CoNEXT's appendix budget is small (long papers ≤4 pages, short papers ≤2), so this is a real allocation problem, not a dumping ground.
The three tiers
| Tier | Belongs in | Test |
|---|---|---|
| Decision-critical | Main body (≤16 / ≤10 pages) | Would a reviewer's accept/reject flip without it? -> body |
| Supporting | Appendix (≤4 / ≤2 pages) | Strengthens or reassures but does not decide -> appendix |
| Reproducibility / bulk | Artifact | Needed to reproduce but not to judge -> artifact |
What must stay in the body
- The central claims and the measurements that back them — a headline result relegated to an appendix reads as hidden weakness.
- The key baseline comparison and how baselines were tuned.
- The central limitation and how you bound it (argue it where the result lives).
- The evaluation platform description sufficient to trust the numbers (topology, hardware, scale).
Reviewers are not required to read appendices or run artifacts to reach a decision; if acceptance hinges on it, it is body content.
What the appendix is for
- Extra results that reinforce a body claim (additional paths, parameter sweeps, secondary metrics).
- Detailed protocol/algorithm listings, proofs, or derivations that support but do not carry the contribution.
- Extended methodology detail (full trace-extraction steps, additional testbed configurations).
Keep it within the acmart appendix budget and clearly cross-referenced from the body; an appendix the body never points to is wasted.
What the artifact is for
- Full traces, configs, testbed scripts, and the raw-to-figure pipeline (see
conext-reproducibility). - Large tables, per-run logs, and anything whose purpose is reproduction rather than judgment.
- Interactive or bulky material that cannot fit the page budget.
The artifact supports reproducibility and (if you opted in) badging; it is not a place to smuggle decision-critical evidence past the page limit.
Common misallocations
| Misallocation | Why it hurts | Fix |
|---|---|---|
| Headline result only in the artifact | Reviewers may not open it; reads as hidden | Bring the result into the body |
| Central limitation buried in an appendix | Misses the chance to pre-empt the objection | Argue it in the body where the result lives |
| Key baseline tuning only in a repo README | Soundness cannot be judged from the paper | Summarize the tuning in the body |
| Appendix overflowing the budget | acmart non-compliance / desk risk | Move bulk to the artifact, keep supporting material |
| Body padded with material that belongs in the artifact | Wastes pages you need for evidence | Push reproduction bulk to the artifact |
Double-anonymity across tiers
Every tier is a leak surface: an appendix figure showing an internal hostname, or an artifact with
commit metadata, breaks anonymity as surely as the body. Run the anonymity sweep (see
conext-submission) across the body, appendix, and artifact
together.
Output format
[Allocation] body / appendix / artifact for each major piece of content
[Decision-critical check] anything acceptance depends on living outside the body? -> move it in
[Budget] appendix pages used vs. limit (long ≤4 / short ≤2); acmart compliant?
[Cross-refs] appendix + artifact pointed to from the body? yes/no
[Anonymity] body + appendix + artifact swept together? yes/no