ATC Supplementary
ATC gives you four places to put content, each read by a different reviewer under a different rule:
the two-page extended abstract (round one, two reviewers), the 12-page body (round two, 3-4
reviewers), the appendix (optional reading), and the artifact (opened during round two and
evaluated post-acceptance). Placing content correctly is a strategic decision, not a formatting one,
because reviewers are not required to read the appendix and a round-one reviewer may read only
the abstract.
The placement rule
Anything that decides acceptance goes in a place the deciding reviewer is required to read.
- Round-one acceptance is decided from the extended abstract → the problem, contribution,
headline result, and honest cost must be in the abstract.
- Round-two acceptance is decided from the body → every load-bearing design detail, baseline,
and result must be in the 12 pages.
- The appendix holds support a curious reviewer may consult — extra graphs, proofs, parameter
tables — never a result the verdict depends on.
- The artifact holds runnable evidence and full data — it strengthens trust and earns badges,
but the paper must be judgeable without executing it.
What goes where
| Content |
Home |
Why |
| Problem, contribution, testbed, headline result, key cost |
Extended abstract |
The round-one decision surface; must stand alone |
| Design rationale, implementation, main experiments, fair baselines, tails/variance |
Body (≤12 pp) |
The round-two decision surface; nothing acceptance-critical may leave it |
| Extra sensitivity graphs, parameter sweeps, proofs, extended tables |
Appendix |
Optional depth; reviewers not required to read it |
| Full code, traces/datasets, configs, claim-to-experiment scripts |
Artifact |
Runnable evidence + badge material; paper still judged without running it |
The extended abstract is not a teaser
The most common ATC mistake is treating the extended abstract as a marketing blurb that depends on
the paper. It is a self-contained argument that must survive being read instead of the paper
(see atc-writing-style). Move the real problem, the real contribution, and a real result into it;
do not defer them to "as shown in the paper."
The appendix is not overflow
The second common mistake is using the appendix to dodge the 12-page limit by exiling results a
reviewer must weigh. If a graph is needed to believe a claim, it belongs in the body; if it is
merely reassuring, it belongs in the appendix. Ask: would a reviewer who skips this change their
score? If yes, it is body content.
Consistency across the four places
- The number in the abstract, the body, the appendix, and the artifact output must match.
Divergent numbers across places are a credibility hit and a frequent reviewer catch.
- The claim-to-experiment map (see
atc-experiments, atc-reproducibility) should let a reviewer
trace each body claim to an artifact script and each artifact output back to a body figure.
- Keep all four double-blind: an appendix or artifact often leaks identity (hostnames, owner
strings, system names) even when the body is clean.
Decision procedure
[For each result] which round decides it? -> place it where that round's reviewer must read
[Abstract test] can a reviewer accept/reject from the 2 pages alone? if not, move content in
[Body test] is any acceptance-critical claim only in an appendix/artifact? if yes, move it in
[Appendix test] would skipping this change a score? if yes, it is body content, not appendix
[Consistency] do the numbers match across abstract/body/appendix/artifact? anonymity intact?
Output format
[Extended abstract] self-standing, decision-complete for round one? yes/no
[Body] all acceptance-critical content inside 12 pages? leaks to appendix/artifact?
[Appendix] optional-only? nothing score-changing exiled here?
[Artifact] runnable evidence + full data; paper judgeable without it? anonymized?
[Move queue] <content to relocate, with target location and reason>
1---2name: atc-supplementary3description: Use when deciding what belongs in the ATC (ACM SIGOPS Annual Technical Conference, formerly USENIX ATC) two-page extended abstract, the 12-page paper body, the optional appendix, and the artifact — splitting content by review round and decision-criticality so nothing that decides acceptance lives where a reviewer is not required to read.4---56# ATC Supplementary78ATC gives you four places to put content, each read by a different reviewer under a different rule:9the **two-page extended abstract** (round one, two reviewers), the **12-page body** (round two, 3-410reviewers), the **appendix** (optional reading), and the **artifact** (opened during round two and11evaluated post-acceptance). Placing content correctly is a strategic decision, not a formatting one,12because reviewers are **not required** to read the appendix and a round-one reviewer may read **only**13the abstract.1415## The placement rule1617**Anything that decides acceptance goes in a place the deciding reviewer is required to read.**1819- Round-one acceptance is decided from the **extended abstract** → the problem, contribution,20 headline result, and honest cost must be *in the abstract*.21- Round-two acceptance is decided from the **body** → every load-bearing design detail, baseline,22 and result must be *in the 12 pages*.23- The **appendix** holds support a curious reviewer *may* consult — extra graphs, proofs, parameter24 tables — never a result the verdict depends on.25- The **artifact** holds runnable evidence and full data — it strengthens trust and earns badges,26 but the paper must be judgeable without executing it.2728## What goes where2930| Content | Home | Why |31|---|---|---|32| Problem, contribution, testbed, headline result, key cost | **Extended abstract** | The round-one decision surface; must stand alone |33| Design rationale, implementation, main experiments, fair baselines, tails/variance | **Body (≤12 pp)** | The round-two decision surface; nothing acceptance-critical may leave it |34| Extra sensitivity graphs, parameter sweeps, proofs, extended tables | **Appendix** | Optional depth; reviewers not required to read it |35| Full code, traces/datasets, configs, claim-to-experiment scripts | **Artifact** | Runnable evidence + badge material; paper still judged without running it |3637## The extended abstract is not a teaser3839The most common ATC mistake is treating the extended abstract as a marketing blurb that depends on40the paper. It is a **self-contained argument** that must survive being read *instead of* the paper41(see `atc-writing-style`). Move the real problem, the real contribution, and a real result into it;42do not defer them to "as shown in the paper."4344## The appendix is not overflow4546The second common mistake is using the appendix to dodge the 12-page limit by exiling results a47reviewer must weigh. If a graph is needed to believe a claim, it belongs in the body; if it is48merely reassuring, it belongs in the appendix. Ask: **would a reviewer who skips this change their49score?** If yes, it is body content.5051## Consistency across the four places5253- The number in the abstract, the body, the appendix, and the artifact output must **match**.54 Divergent numbers across places are a credibility hit and a frequent reviewer catch.55- The claim-to-experiment map (see `atc-experiments`, `atc-reproducibility`) should let a reviewer56 trace each body claim to an artifact script and each artifact output back to a body figure.57- Keep all four **double-blind**: an appendix or artifact often leaks identity (hostnames, owner58 strings, system names) even when the body is clean.5960## Decision procedure6162```text63[For each result] which round decides it? -> place it where that round's reviewer must read64[Abstract test] can a reviewer accept/reject from the 2 pages alone? if not, move content in65[Body test] is any acceptance-critical claim only in an appendix/artifact? if yes, move it in66[Appendix test] would skipping this change a score? if yes, it is body content, not appendix67[Consistency] do the numbers match across abstract/body/appendix/artifact? anonymity intact?68```6970## Output format7172```text73[Extended abstract] self-standing, decision-complete for round one? yes/no74[Body] all acceptance-critical content inside 12 pages? leaks to appendix/artifact?75[Appendix] optional-only? nothing score-changing exiled here?76[Artifact] runnable evidence + full data; paper judgeable without it? anonymized?77[Move queue] <content to relocate, with target location and reason>78```
brycewang-stanford