Use this to split content between the reviewed pages and anything external. DAC's constraint is severe and specific: the manuscript is 6 pages of body plus one references-only page, and unlike some venues DAC does not provide an unlimited reviewed appendix or a formal artifact-evaluation track (待核实 per cycle). The governing rule follows directly: anything that decides acceptance must fit inside the six body pages, because a reviewer is neither obligated to nor guaranteed to open external material.
For each piece of content, ask: would a reviewer need this to be convinced the contribution is real and the QoR claim is fair? If yes, it belongs in the body. If it merely lets a motivated reader reproduce or extend the work, it can live externally.
| Content | Where it goes | Why |
|---|---|---|
| The headline QoR results and the mechanism | Body (6 pages) | Decision-critical; reviewers judge on these |
| Per-benchmark result table for the main claim | Body | The fairness of the comparison is the paper |
| The key ablation | Body | Attribution of the gain is decision-critical |
| Full hyperparameter grids, extra corners, exhaustive per-instance logs | External repo | Supports reproduction, does not decide acceptance |
| Source code, scripts, generated datasets | External repo | Reproducibility, not a reviewed deliverable |
| Extra proofs/derivations too long for six pages | External repo, with the result stated in-body | The claim must stand in-body; detail can be external |
DAC's most mechanical rule: the one extra page holds references and nothing else. Do not treat
it as an appendix — an extra figure, a paragraph, or a table there is grounds for desk rejection
(../dac-submission). Plan the technical argument to end cleanly at page six.
If you link a repository, dataset, or artifact at review time, it must be anonymized exactly like the PDF:
../dac-artifact-evaluation, ../dac-reproducibility).Before moving something out, try to earn its space in the body: convert a prose result to a compact table, merge two figures, cut standard EDA background the reviewers already know. Six dense double-column pages hold more than first drafts assume.
[Split audit] each item -> body (decision-critical) / external (reproduction) / cut
[Seventh page] references only, no body content? yes/no
[External anon] linked repo/dataset/artifact anonymized? reviewer-optional acknowledged? yes/no
[Stands alone] does the paper convince on its six pages without external material? yes/no
[Compression] space reclaimed before relegating anything? yes/no