English Academic Editing Station (academic-english-editing)

What does English Academic Editing Station (academic-english-editing) do?

This built-in English academic editing station is included with the plugin (available from v0.8.0). Use it after the translate-en step produces an English draft, but before final human English confirmation: remove AI-generated English traces, align with academic conventions, match the author's and journal's tone, while preserving the precise, evidence-bound tone that academic writing already has. In technical writing, 'neutral and precise' is the human voice—do not inject opinions, humor, or a first-person personality.

Posting Contract (Source: academic-pipeline Editing Station Contract)

• This is an optional station: it can be skipped, but it cannot replace human English confirmation.
• Only language presentation may be changed; do not modify numbers, <!-- results.* --> / <!-- table:* --> / <!-- claims.* --> markers, [@citekey], claim strength, validation boundary, or the semantics of methods or limitations.
• After editing, create a new English version manuscript/en/vN.md (sync version if SI exists).
• After editing, re-run check_markers.py, fact-checker cross-semantic review (sample 2–3 sentences per section and back-translate for comparison), and reviewer-language; human English confirmation and a full check-consistency(en) are still required.
• Any changes involving claim strength, numerical interpretation, methods, or limitations must be sent back to the upstream source artifact (claim-audit / interpret-results / validation); this station does not adjudicate them.

Why Claim-Evidence Rewriting Is Excluded (Formerly academic-humanizer Layer 4)

Layer 4 of academic-humanizer (downgrading verbs to the strength of the evidence, or adding indicators or softening for unsupported claims) changes claim strength, violating the above 'only modify presentation' contract. Moreover, by the time the manuscript reaches this station, claims have already been reviewed by claim-audit and confirmed by a human. Therefore, this station does not perform Layer 4 rewriting: if it finds an over-claim, it only marks it and reports it to claim-audit/interpret-results, without rewriting the text. The spirit of the original Layer 4 is already enforced in the pipeline by claim-audit's entailment/strength attack and check_numbers.py.

When to Use

Edit English academic prose: paper sections, abstracts, rebuttals, related work. Do not use for blogs, marketing, personal essays, and it is not a tool to evade AI-use disclosure. For grant/research proposals (NSF, NIH, etc.), use the separate skill grant-proposal-editing (the proposal register is different—it sells vision and feasibility; the tightness of a paper is not appropriate).

Core Principle (Source: academic-humanizer v0.3.3 · Core principle)

Academic writing already has a correct human voice: neutral, precise, third-person plural ('we'), with each claim tied to evidence. This station's job is to (1) remove AI tells without becoming colloquial; (2) maintain discipline that general humanizers miss—preserve every number, result, and citation, and never let any verb exceed its evidence (but rewriting is only at the 'presentation' level; adjudication of evidence strength belongs to claim-audit).

Process (Source: academic-humanizer v0.3.3 · Process, adjusted for the posting contract)

1. Read: Read the English draft and (if available) the author's existing paper samples; note the target journal's register.
2. Audit (no rewriting yet): List all detected patterns (location + suggested fix) one by one. Items involving claim strength or numerical interpretation are marked '→ send to claim-audit' and are not included in this station's rewriting list.
3. Rewrite: Keep structure and content unchanged; preserve all numbers, markers, citekeys, and claim indicators; remove tells; retain legitimate hedging.
4. Report: Output the polished draft plus a brief change log (types of removed patterns, tone/register notes), and confirm that no numbers, equations, or citations have been altered. Cover everything the original text covers: if the original had five paragraphs, the rewrite also has five.

---

The following Layers 1/2/3/5 content is derived from academic-humanizer v0.3.3, with the source marked layer by layer; only cross-references related to the proposal mode (formerly Layer 6) have been removed.

Layer 1: Generic AI-tell Checklist (Source: academic-humanizer v0.3.3 · Layer 1)

Scan and correct these generic patterns, subject to Layer 3's academic exceptions: inflated significance ('marking a pivotal moment'); superficial '-ing' tails that fake depth ('..., highlighting...'); promotional/figurative language ('rich', 'vibrant', 'groundbreaking'); vague attributions ('experts argue' with no cite); AI vocabulary (delve, underscore, intricate, tapestry, testament, landscape (abstract), pivotal, showcase, foster, leverage (filler), realm, seamless); copula avoidance ('serves as' → 'is'); negative parallelisms ('not just X, but Y'); rule-of-three padding; elegant variation (cycling synonyms for one referent); filler ('it is worth noting that', 'in order to'); overlong, clause-stacked sentences (split them; see 2.11); and em-dashes (remove entirely; rewrite with commas, colons, parentheses, or sentence breaks).

Before: Additionally, an enduring testament to the method's value is its ability to delve into intricate dependencies, showcasing a seamless integration that underscores its pivotal role. After: The method also captures higher-order dependencies, which the baselines miss (Table 2).

Layer 2: Academic AI Tells (Remove or Correct) (Source: academic-humanizer v0.3.3 · Layer 2)

2.1 Over-claiming verbs

Empirical work shows / provides evidence, it does not prove / demonstrate universal truths. Watch: demonstrate, prove, establish, confirm, guarantee; 'significantly' without a test or number. Before: We prove that our method significantly outperforms all prior approaches. After: Our method improves held-out accuracy by 4–7 points over the strongest prior approach (Table 3); the gain is significant at p < 0.01 by a paired test. (Note: This station adjusts tone only at the 'presentation' level; if the actual evidence does not support the verb, that is a claim-strength issue, marked for claim-audit, not rewritten here.)

2.2 Significance hype

Watch: paves the way for, a crucial/pivotal step toward, has the potential to revolutionize, opens new avenues, sheds light on, of paramount importance, bridges the gap. Before: This work paves the way for a new paradigm and sheds light on a problem of paramount importance. After: This work addresses one failure mode of prior methods: error accumulation under long-horizon rollout (Section 4).

2.3 Empty intensifiers

Watch: extensive/comprehensive/thorough experiments, a wide range of, numerous, various. Before: We conduct extensive experiments on a wide range of datasets. After: We evaluate on three datasets (ImageNet, CIFAR-100, and COCO).

2.4 Novelty padding

Watch: 'novel' more than once per section; 'to the best of our knowledge'; 'for the first time'. Before: We propose a novel framework and, to the best of our knowledge, are the first to study this. After: We study online calibration under delayed labels, which prior calibration work (offline) does not address.

2.5 Formulaic openers

Watch: 'In recent years, X has attracted increasing attention'; 'With the rapid development of...'; 'Despite recent advances,...'. Before: In recent years, tabular deep learning has attracted increasing attention. After: Tabular deep learning has a structural limitation: most models discard feature-type metadata and must relearn it from data.

2.6 Connective overuse

Don't let consecutive sentences all start with Moreover/Furthermore/Additionally/In particular; let the logic carry itself. Before: Moreover, the method is fast. Furthermore, it is simple. Additionally, it scales. After: The method is fast and simple, and it scales to one million rows (Section 5).

2.7 Contribution-list cliches

Each contribution should name specific results, not restate the abstract. Before: Our contributions are: (1) a novel method; (2) extensive experiments; (3) strong results. After: We (1) introduce a metadata-aware encoder that reaches 0.91 AUROC vs 0.86 for the strongest baseline; (2) show it stays within 2 points under 20% label noise where the baseline drops 9; (3) release the benchmark.

2.8 Citation dumping

Cite one or two truly relevant papers and explain why, not a string of brackets. Before: Many methods exist [3, 7, 9, 12, 15]. After: The closest prior method is TabNet [7], which encodes all features jointly; we instead condition on feature-type metadata.

2.9 Hedging-by-vagueness

Watch: somewhat, relatively, fairly, to some extent, quite. Quantify or remove. Before: Performance is somewhat better and relatively robust. After: Accuracy is 3 points higher and varies by less than 1 point across five seeds.

2.10 Boilerplate emphasis

Watch: 'It is worth noting that', 'It should be emphasized that', 'Notably,', 'Importantly,'. If it is truly important, the sentence will show it. Before: It is worth noting that, importantly, the gain holds across scenarios. After: The gain holds across all three scenarios (Table 4).

2.11 Overlong, clause-stacked sentences

AI likes to string together three or four clauses with 'which'/'that'/'while'/'with'. Split them: one idea per sentence, cut clauses that don't bear weight. Watch: sentences over about 30 words or with 3+ clauses. Before: Existing methods, though promising, are largely empirical, with unclear principles underpinning their behavior, which limits their reliability and further progress. After: Existing methods stay empirical. Their principles are unclear, which limits reliability and progress.

Layer 3: Preserve—Do Not Over-Correct (Source: academic-humanizer v0.3.3 · Layer 3)

General humanizers tend to flatten legitimate academic structures; keep these:

• Evidence-tied hedging is correct and necessary. When the claim is indeed uncertain, preserve 'suggests', 'is consistent with', 'we hypothesize that', 'may indicate', 'appears to'. Wrong correction: changing 'the results suggest X' to 'the results prove X'—that creates an over-claim. Keep calibrated verbs.
• Passive voice is fine when the actor is irrelevant: 'Samples were normalized to total protein.'
• First-person plural 'we' is standard; don't rewrite just to avoid it.
• Semicolons and occasional triples are acceptable in moderation. Em-dashes are an exception: remove them entirely (Layer 1), rewriting with commas, colons, parentheses, or sentence breaks.
• Formal definitions, named methods/metrics, terminology, equations, and symbols stay as-is.
• Never invent, delete, or change any number, equation, or citation; preserve every cite key. (This station is mechanically guarded by the pipeline's check_markers.py / check_numbers.py / check_citations.py and the full check-consistency(en).)

Layer 5: Tone and Journal Register Alignment (Source: academic-humanizer v0.3.3 · Layer 5)

If the author provides existing papers, read the sample first; note sentence rhythm, connective habits, degree and placement of hedging, how sections open, trademarks, and idiomatic expressions, then align. Also align with the journal register (e.g., ICLR/NeurIPS: concise, direct, results-forward; Nature/PNAS: more expository). When no sample is available, default to clean, precise, journal-appropriate prose, not the colloquial, opinionated voice of a general humanizer.

---

Provenance and Derivation Notes

• Layers 1/2/3/5, Core Principle, Process, When to Use: Derived from the user skill academic-humanizer v0.3.3 (MIT), marked section by section with [Source: academic-humanizer v0.3.3 · …]; text is largely copied verbatim, only removing cross-references to the proposal mode and adding 'rewriting only at the presentation level, evidence strength goes to claim-audit' annotations in 2.1/Layer 3.
• Posting Contract and 'Why Claim-Evidence Rewriting Is Excluded': Newly added for the academic-pipeline [Source: academic-pipeline Editing Station Contract], aligning with the existing specifications of translate-en/status for the editing station.
• Not Included: Layer 4 of academic-humanizer (claim-evidence discipline) is not part of this station; its responsibilities in the pipeline are handled by claim-audit/interpret-results/check_numbers.py.
• Separately Extracted: Layer 6 of academic-humanizer (NSF/NIH proposal mode, including 6.1–6.5) is extracted as the separate skill grant-proposal-editing, not included with this plugin (orthogonal to the research-to-submission workflow).

Output

Output the polished English draft plus a brief change report: removed pattern (category), claim-related items (not changed but marked for claim-audit), tone/register notes, and a confirmation that no numbers, equations, or citations have been altered.