Use when positioning an ACL submission against the NLP literature, covering ACL Anthology citation practice, arXiv-versus-published version citation, concurrent LLM-era preprints, prior-cycle ARR resubmission overlap, anonymity-preserving self-citation, and the fast-moving baseline problem in computational linguistics.
NewMaintained
Automatic
◆Research pack
Last verified 8/15/2026
What it changes
Passed 17 of 22 test cases with the skill, and 17 of 22 without it.
Without skill77%
With skill77%
Checks it passed that the model alone missed
State that citing inaccessible or unreleased documents as supporting evidence is strictly forbidden
Define the two passes as a coverage check for novel claims and a citation hygiene check for venues and self-citations
2 fixed15 passed either way3 still failing2 broke
Pass-rate lift
0 pp
Tokens
+35%
Agent turns
0%
Response time
-30%
Model gemini-3.6-flashMeasured 8/15/2026v117 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 2 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 2
Prompt
What are the two distinct auditing passes needed when reviewing an ACL related work section draft?
What was checked, without and with the skill
Checked
Without skill
With skill
Define the two passes as a coverage check for novel claims and a citation hygiene check for venues and self-citations
Missed
Passed
case-22 · gemini-3.6-flash · 8/15/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md114 linessingle file
ACL Related Work
Use this to audit positioning and citation hygiene. NLP moves fast enough that
related-work sections age between ARR cycles; the section must prove the
authors know the field as of this cycle, not as of the project's start.
The Anthology-first rule
When a paper exists in the ACL Anthology (or another archival venue), cite
that version, not its arXiv preprint — Anthology entries are the community's
citation of record and carry stable BibTeX.
The Anthology is fully open access (post-2016 materials under CC BY 4.0),
so "couldn't access the published version" is never a defense here.
For findings-tier work, cite it as Findings of ACL/EMNLP — it is archival
and reviewers treat miscrediting it as sloppiness.
Freshness pressure unique to this field
Expect at least one reviewer to name a baseline or dataset from the last
two conference rounds. Sweep ACL, EMNLP, NAACL, EACL, TACL, and CL for the
nearest neighbors published since your project began.
LLM-era concurrency is the norm: assume 2-4 similar preprints exist. Cite
the ones findable at submission, contrast on technique rather than date,
and avoid priority wars that reviewers cannot adjudicate.
If a newer model release makes your comparison set look dated mid-cycle,
address it in the author response and plan the upgrade for camera-ready or
resubmission — silence reads as unawareness.
Positioning moves that work at ACL
Contribution type
The contrast reviewers want spelled out
Method
Nearest prior method + the mechanism-level difference, not just scores
What prior evaluations measured incorrectly or could not see
Metric
Failure cases of the incumbent metric, with correlation evidence
Theme-track paper
How the work answers the year's theme question (ACL 2026: explainability)
One paragraph of typed contrast per lane beats an undifferentiated citation
inventory; ACL's compressed format cannot afford taxonomy for its own sake.
Use when positioning a CoRL paper against the literature — the robot-learning lineage across CoRL/RSS/ICRA, the ML-methods stream from NeurIPS/ICLR/ICML, the fast-moving VLA and foundation-model wave, classical robotics baselines, concurrent arXiv work, and PMLR citation hygiene including the year-offset trap.