eacl-supplementary

EACL Supplementary Material

Use this to decide what goes where across an EACL submission: the content pages, the

Limitations section, the appendices, and the anonymized supplementary archive. The governing

rule in ARR is that **reviewers must read the content pages but are not obliged to read

appendices or the archive** — so anything decision-critical that lives outside the body is a

self-inflicted risk. Reopen the current ARR CFP before finalizing structure.

The placement hierarchy

| Location | Reviewer obligation | Put here |

|---|---|---|

| Content pages (8 long / 4 short) | Must read | The claim, method, key results, the argument |

| Limitations (after conclusion, uncounted) | Must read | Honest scope, failure modes, threats to validity |

| References (unlimited) | As needed | Citations |

| Appendices (in-PDF, after refs) | May skip | Proof detail, extra tables, full hyperparameters |

| Anonymized archive (OpenReview) | May skip | Code, data, prompts, model outputs |

The self-containedness test

• Read only the content pages and the Limitations section. Can a reviewer **accept or reject on

that alone**? If a core mechanism, the main baseline, or the headline result lives only in an

appendix, move it into the body.

• "See Appendix X" is fine for detail, never for the argument.

Limitations is body, not backmatter

• The Limitations section is mandatory and read; treat it as part of the argument, not a

disclaimer. Name the languages you did not cover, the settings you did not test, and the

claims you cannot support — reviewers reward this and it is explicitly protected.

• Do not smuggle results into Limitations to dodge the page limit; that is transparent and

counterproductive.

Appendix organization

Appendix A: full hyperparameters + search spaces
Appendix B: additional / per-language results tables
Appendix C: annotation guidelines + agreement
Appendix D: prompt templates + decoding settings (verbatim)
Appendix E: additional error analysis / examples

Label appendices so the body can point precisely, and keep each reproducible from the archive.

Anonymized archive hygiene

• The OpenReview supplement must be fully de-identified (see eacl-artifact-evaluation): no

names in paths, headers, or git history.

• Keep the archive self-contained — a reviewer who does open it should be able to reproduce a

reported table without hunting for external files.

Common misplacements to catch

• Core baseline hidden in an appendix — reviewers may never see it; the comparison then reads

as missing.

• Key multilingual result only in the archive — an EACL reviewer will not go digging;

surface at least a summary table in the body.

• Ethics/data-governance detail buried — for sensitive or lower-resource data, keep the

intended-use and provenance statement visible.

Output format

[Placement audit] Pass / Needs moves
[Decision-critical content outside body] <list, with where it should go>
[Limitations quality] <substantive / ritual>
[Appendix map] <A-E contents>
[Archive hygiene] <anonymized + self-contained: pass/fail>