ACL Supplementary
Use this when splitting an ACL paper between body, appendix, and archive. The
governing ARR principle: reviewers are not required to consider material in
appendices or supplements, so anything decision-critical that lives only
there is effectively invisible.
The ACL page anatomy
[ content pages: 8 long / 4 short ] <- the reviewed argument lives here
[ Limitations (REQUIRED, unlimited) ] <- after conclusion, outside page count
[ Ethics statement (optional) ]
[ References (unlimited) ]
[ Appendices (unlimited, same PDF) ] <- optional reading for reviewers
+ separate .tgz/.zip archive <- software / data supplement
Missing Limitations is a desk-reject condition; treating it as one throwaway
sentence is a review-stage penalty even when it passes the gate.
What must not leave the body
- The main results table and the headline comparison.
- Task definition and enough of the method that a reviewer can judge novelty.
- At least a summary of the error analysis — a pointer-only error analysis
reads as not having one.
- Human-evaluation design in one paragraph: raters, items, agreement.
- The experimental setup at reproduction-outline level; full grids can go down.
What appendices are good at
- Full prompt texts and few-shot exemplars (reference them per experiment).
- Complete hyperparameter tables and search ranges.
- Per-language / per-dataset breakdowns behind an averaged headline number.
- Annotation guidelines and interface screenshots.
- Extended qualitative examples and additional ablations.
- Proofs or derivations for the occasional formal result.
Limitations section that actually works
| Weak pattern |
Stronger ACL pattern |
| "Results may not generalize" |
Name the languages, domains, and model scales actually tested and the nearest untested regime |
| "LLMs can hallucinate" |
State which conclusions depend on a specific model snapshot and API behavior |
| Silent on data |
Note license constraints, demographic skew, or collection-window bias in the corpora used |
| Written last-minute |
Mirrors the risks reviewers will find anyway, defusing them on your terms |
ACL's policy explicitly instructs reviewers not to punish honest limitations,
which makes this section the cheapest goodwill in the whole submission.
Archive rules and hygiene
- Upload software/data as .tgz or .zip in the OpenReview fields; personal
cloud-storage links are barred, and any external page must be anonymous and
untracked.
- Strip authorship trails: git history, notebook metadata, absolute paths with
usernames, license headers, README contact lines.
- Test the archive on a clean machine: it must unpack, the README must state
what maps to which table, and nothing should require credentials just to read.
- Include model outputs where feasible so reviewers can verify scoring without
compute (see
acl-reproducibility).
Body-vs-appendix vignette
A long paper introduces a retrieval-augmented QA method with results on six
benchmarks in three languages. Body: method figure, main table (six benchmarks
averaged + per-language block), two-paragraph error analysis, one ablation that
carries the mechanism claim. Appendix: full per-benchmark tables, prompts,
retrieval index details, remaining ablations, annotation guidelines for the
human study. Archive: code, prompts as files, and all model outputs. The test:
a reviewer who never scrolls past the references can still reconstruct and
believe every claim in the abstract.
Appendix ordering convention that reviewers navigate well
- A: full experimental setup (models, hyperparameters, hardware, budgets).
- B: prompts and few-shot exemplars, one subsection per experiment.
- C: complete results — per-dataset, per-language, per-seed tables behind
every averaged number in the body.
- D: annotation materials — guidelines, interface, pay, agreement detail.
- E: additional analyses and ablations, each forward-referenced from the
body at least once (unreferenced appendix content is invisible).
- F: qualitative examples, marked as random or curated — say which.
Number tables and figures continuously with the body so the author response
can cite "Table 9" unambiguously during the discussion phase.
Ethics statement: when to write one
Write it when the paper involves human subjects or annotators, scraped
user-generated content, demographic inference, dual-use capability, or
release of models/data with realistic misuse paths. Skip it when nothing
applies — the Responsible NLP checklist already covers the routine cases,
and a padded statement invites the very scrutiny it fails to answer.
It shares the unlimited space after the conclusion with Limitations.
Size and dependency guardrails
- Keep the archive small enough to download on conference-hotel wifi;
reviewers abandon multi-gigabyte supplements unopened.
- No credentials, no API keys, no gated-model weights — describe access
paths instead of shipping secrets.
- Pin dependency versions in one environment file; "latest transformers"
is a different codebase every cycle.
- If data cannot be shared (license, privacy), include the loader code and
a synthetic sample with identical schema so scripts still run.
Output format
[Split status] sound / body-overloaded / appendix-dependent
[Limitations quality] substantive / ritual / missing
[Must-move-up] <decision-critical items currently below the fold>
[Archive check] <format/anonymity/clean-machine findings>
[Reviewer-blind spots] <claims visible only outside the body>
Source: brycewang-stanford/Awesome-Journal-Skills → ACL-Skills/skills/acl-supplementary/SKILL.md
1---2name: acl-supplementary3description: Use when organizing appendices and supplementary material for an ACL paper under ACL Rolling Review, covering the mandatory Limitations and optional ethics sections, appendices after references, anonymized software and data archives, the no-cloud-links rule, and deciding what must stay in the 8-page or 4-page body.4---5
6
7# ACL Supplementary
8
9Use this when splitting an ACL paper between body, appendix, and archive. The
10governing ARR principle: **reviewers are not required to consider material in
11appendices or supplements**, so anything decision-critical that lives only
12there is effectively invisible.
13
14## The ACL page anatomy
15
16```text
17[ content pages: 8 long / 4 short ] <- the reviewed argument lives here
18[ Limitations (REQUIRED, unlimited) ] <- after conclusion, outside page count
19[ Ethics statement (optional) ]
20[ References (unlimited) ]
21[ Appendices (unlimited, same PDF) ] <- optional reading for reviewers
22+ separate .tgz/.zip archive <- software / data supplement
23```
24
25Missing Limitations is a desk-reject condition; treating it as one throwaway
26sentence is a review-stage penalty even when it passes the gate.
27
28## What must not leave the body
29
30- The main results table and the headline comparison.
31- Task definition and enough of the method that a reviewer can judge novelty.
32- At least a summary of the error analysis — a pointer-only error analysis
33 reads as not having one.
34- Human-evaluation design in one paragraph: raters, items, agreement.
35- The experimental setup at reproduction-outline level; full grids can go down.
36
37## What appendices are good at
38
39- Full prompt texts and few-shot exemplars (reference them per experiment).
40- Complete hyperparameter tables and search ranges.
41- Per-language / per-dataset breakdowns behind an averaged headline number.
42- Annotation guidelines and interface screenshots.
43- Extended qualitative examples and additional ablations.
44- Proofs or derivations for the occasional formal result.
45
46## Limitations section that actually works
47
48| Weak pattern | Stronger ACL pattern |
49|---|---|
50| "Results may not generalize" | Name the languages, domains, and model scales actually tested and the nearest untested regime |
51| "LLMs can hallucinate" | State which conclusions depend on a specific model snapshot and API behavior |
52| Silent on data | Note license constraints, demographic skew, or collection-window bias in the corpora used |
53| Written last-minute | Mirrors the risks reviewers will find anyway, defusing them on your terms |
54
55ACL's policy explicitly instructs reviewers not to punish honest limitations,
56which makes this section the cheapest goodwill in the whole submission.
57
58## Archive rules and hygiene
59
60- Upload software/data as .tgz or .zip in the OpenReview fields; personal
61 cloud-storage links are barred, and any external page must be anonymous and
62 untracked.
63- Strip authorship trails: git history, notebook metadata, absolute paths with
64 usernames, license headers, README contact lines.
65- Test the archive on a clean machine: it must unpack, the README must state
66 what maps to which table, and nothing should require credentials just to read.
67- Include model outputs where feasible so reviewers can verify scoring without
68 compute (see `acl-reproducibility`).
69
70## Body-vs-appendix vignette
71
72A long paper introduces a retrieval-augmented QA method with results on six
73benchmarks in three languages. Body: method figure, main table (six benchmarks
74averaged + per-language block), two-paragraph error analysis, one ablation that
75carries the mechanism claim. Appendix: full per-benchmark tables, prompts,
76retrieval index details, remaining ablations, annotation guidelines for the
77human study. Archive: code, prompts as files, and all model outputs. The test:
78a reviewer who never scrolls past the references can still reconstruct and
79believe every claim in the abstract.
80
81## Appendix ordering convention that reviewers navigate well
82
83- A: full experimental setup (models, hyperparameters, hardware, budgets).
84- B: prompts and few-shot exemplars, one subsection per experiment.
85- C: complete results — per-dataset, per-language, per-seed tables behind
86 every averaged number in the body.
87- D: annotation materials — guidelines, interface, pay, agreement detail.
88- E: additional analyses and ablations, each forward-referenced from the
89 body at least once (unreferenced appendix content is invisible).
90- F: qualitative examples, marked as random or curated — say which.
91
92Number tables and figures continuously with the body so the author response
93can cite "Table 9" unambiguously during the discussion phase.
94
95## Ethics statement: when to write one
96
97Write it when the paper involves human subjects or annotators, scraped
98user-generated content, demographic inference, dual-use capability, or
99release of models/data with realistic misuse paths. Skip it when nothing
100applies — the Responsible NLP checklist already covers the routine cases,
101and a padded statement invites the very scrutiny it fails to answer.
102It shares the unlimited space after the conclusion with Limitations.
103
104## Size and dependency guardrails
105
106- Keep the archive small enough to download on conference-hotel wifi;
107 reviewers abandon multi-gigabyte supplements unopened.
108- No credentials, no API keys, no gated-model weights — describe access
109 paths instead of shipping secrets.
110- Pin dependency versions in one environment file; "latest transformers"
111 is a different codebase every cycle.
112- If data cannot be shared (license, privacy), include the loader code and
113 a synthetic sample with identical schema so scripts still run.
114
115## Output format
116
117```text
118[Split status] sound / body-overloaded / appendix-dependent
119[Limitations quality] substantive / ritual / missing
120[Must-move-up] <decision-critical items currently below the fold>
121[Archive check] <format/anonymity/clean-machine findings>
122[Reviewer-blind spots] <claims visible only outside the body>
123```
124
125---
126
127**Source:** [`brycewang-stanford/Awesome-Journal-Skills`](https://github.com/brycewang-stanford/Awesome-Journal-Skills) → `ACL-Skills/skills/acl-supplementary/SKILL.md`
thedixitjain