ecai-supplementary

ECAI Supplementary Material

ECAI's defining constraint is the 7-page body (references are the only overflow: 1 page in a

standalone ECAI, 2 pages in IJCAI-ECAI 2026). There is no unlimited in-paper appendix: the

paper is exactly its 7 pages plus references. A separate supplement can carry proofs, extra

tables, and code — but it is optional reading for reviewers, so the split must be made by

decision-criticality, not convenience.

The one rule

> Anything a reviewer needs in order to judge the paper must be in the 7-page body.

> The supplement holds support and detail, never the claim, the core argument, or the key

> evidence.

Reviewers may or may not open the supplement, and they read it under time pressure. If your central

result only becomes convincing after page 7, the paper is under-argued.

Belongs in the body (7 pages)

• The problem statement, contribution, and the claim in precise form.
• The core of the argument: the main theorem statement and proof idea, or the main experimental

result with the comparison that supports the claim.

• The key table/figure a reviewer uses to decide, with enough of the protocol to trust it.
• Threats/limitations that bear on whether the claim holds.

Belongs in the supplement (anonymized)

• Full proofs the body sketches (typical for KR/theory/argumentation/planning work).
• Additional experiments, ablations, and per-domain/per-dataset breakdowns beyond the headline.
• Code, domain files, seeds, configs, and cached outputs (ecai-reproducibility).
• Extended related-work discussion that does not fit the reference budget.
• Hyperparameter grids, dataset statistics, and preprocessing details.

The decision-criticality test

For each block ask: "If a reviewer skipped this, could they still fairly judge the contribution?"

• No → it is decision-critical → keep it in the body, compress elsewhere to fit.
• Yes → it supports/details the claim → supplement.

A frequent ECAI mistake is exiling the proof idea or the baseline comparison to the

supplement to save space. Those are decision-critical; cut prose, merge figures, or tighten

notation instead.

Compression tactics before you exile content

The tension is real at 7 pages. Recover space editorially before demoting anything critical:

• Merge related figures; use one multi-panel figure instead of three.
• State lemmas compactly; move only routine proof steps to the supplement, keep the idea in-body.
• Cut roadmap/signposting sentences; the body is too short to preview itself section by section.
• Tighten related work to delta-first positioning (ecai-related-work).
• Move verbose hyperparameter tables out; keep the one setting that matters in the text.

Double-blind and anonymity

The supplement is read under double-blind review — anonymize it exactly as the PDF (strip

repository owners, institution and funding lines, system names). A de-anonymizing supplement risks

a summary reject (ecai-submission, ecai-reproducibility).

Format and mechanics (verify per cycle)

• Confirm on the current call whether and how a supplement is submitted (a separate file/archive

vs. an appendix mechanism), the size limit, and whether reviewers are told it is optional.

These differ between a standalone ECAI and the joint IJCAI-ECAI 2026 — treat as 待核实 until

the call confirms.

• Do not use the reference pages as a spillover for body text; they are for references (plus, in

2026, an optional ethics statement).

Output format

[Body budget] 7 pages; decision-critical content all inside? yes/no
[In body] problem, contribution, main theorem/idea or main result + supporting table
[In supplement] full proofs / extra experiments / code+data / extended related work
[Criticality check] anything decision-critical currently outside the body? <list -> move in>
[Anonymity] supplement clean / leaks: <where>
[Mechanics] supplement format/size per current call confirmed (待核实 if not)