CharlieWei5183/academic-english-editing

英文潤稿站（academic-english-editing）

本 plugin 內建的英文學術潤稿站（0.8.0 起隨附）。在 translate-en 產出英文稿之後、人工英文確認之前選用：把 AI 生成的英文痕跡清掉、對齊學術慣例並貼近作者與期刊語氣，同時保住學術寫作本來就正確的精確、證據綁定語氣。技術寫作裡「中性而精準」就是人的聲音，不要注入意見、幽默或第一人稱個性。

掛點合約〔來源：academic-pipeline 潤稿站合約〕

• 這是選用站：可略過，但不能取代人工英文確認。
• 只可動語言呈現；不得改數字、<!-- results.* -->／<!-- table:* -->／<!-- claims.* --> markers、[@citekey]、claim strength、validation boundary、方法或限制語意。
• 潤稿產生新英文版本 manuscript/en/vN.md（有 SI 時同步新版本）。
• 潤稿後重跑 check_markers.py、fact-checker 跨語意回查（每節抽 2–3 句回譯比對）與 reviewer-language；之後仍須人工英文確認與 full check-consistency(en)。
• 任何涉及 claim 強度、數字解讀、方法或限制的更動，回上游來源 artifact（claim-audit／interpret-results／validation），不在本站裁定。

為何不含 claim-evidence 改寫（原 academic-humanizer Layer 4）

academic-humanizer 的 Layer 4（把動詞降級到證據強度、替無依據主張補指標或 softening）會改變 claim strength，違反上面「只動呈現」的合約；而且走到本站時，claims 已在 claim-audit 審定並經人工確認。因此本站不做 Layer 4 改寫：發現 over-claim 只標記並回報給 claim-audit／interpret-results，不自行改稿。原 Layer 4 的精神在 pipeline 內已由 claim-audit 的 entailment／強度攻擊與 check_numbers.py 落實。

何時用

編修英文學術散文：論文各節、摘要、rebuttal、related work。不用於部落格、行銷、個人隨筆，也不是規避 AI 使用揭露的工具。若要處理經費／研究提案（NSF、NIH 等），改用獨立 skill grant-proposal-editing（提案語域不同，賣的是願景與可行性，論文層的緊縮不適用）。

核心原則〔來源：academic-humanizer v0.3.3 · Core principle〕

學術寫作已經有正確的人聲：中性、精準、第三人稱複數（"we"）、每個主張都綁到證據。本站的工作是（1）在不口語化的前提下清掉 AI tells；（2）守住一般 humanizer 會漏掉的紀律——保留每個數字、結果與引用，不讓任何動詞強過它的證據（但改稿只在「呈現」層；證據強度的裁定屬 claim-audit）。

流程〔來源：academic-humanizer v0.3.3 · Process，依掛點合約調整〕

1. Read：讀英文稿與（若有）作者既有論文樣本；記下目標期刊語域。
2. Audit（先不改）：逐一列出偵測到的 pattern（位置＋建議改法）。涉及 claim 強度或數字解讀者標「→ 回 claim-audit」，不進本站改寫清單。
3. Rewrite：結構與內容不變，所有數字、markers、citekey、claim 標記保留，去除 tells、保留合法 hedging。
4. Report：輸出潤飾稿＋簡短 change log（移除的 pattern 類型、語氣／語域註記），並確認未動任何數字、方程式或引用。涵蓋原文涵蓋的一切：原本五段，改寫也五段。

---

以下 Layer 1／2／3／5 內容源自 academic-humanizer v0.3.3，逐層標註來源；僅去除與提案模式（原 Layer 6）相關的交叉引用。

Layer 1：通用 AI-tell 清單〔來源：academic-humanizer v0.3.3 · Layer 1〕

掃描並修正這些通用 pattern，但受 Layer 3 的學術例外約束： 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; 見 2.11）；以及 em-dashes（整個移除；改以逗號、冒號、括號或分句重寫）。

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：學術 AI tells（移除或修正）〔來源：academic-humanizer v0.3.3 · Layer 2〕

2.1 Over-claiming verbs

實證工作 shows / provides evidence，不 prove / demonstrate 普世真理。 Watch: demonstrate, prove, establish, confirm, guarantee；"significantly" 無檢定／數字。 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. （註：本站只在「呈現」層調語氣；若實際證據不足以支持動詞，屬 claim 強度問題，標記回 claim-audit，不在本站改寫。）

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" 每節超過一次；"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

別讓連續句子都以 Moreover／Furthermore／Additionally／In particular 起頭；讓邏輯自己承接。 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

每條貢獻點名具體結果，不是重述摘要。 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

引一兩篇真正相關的並說明為何，而非一串方括號。 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. 量化或刪掉。 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,". 若真重要，句子自己會顯示。 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 偏好用 "which"／"that"／"while"／"with" 串三四個子句的長句。拆開：一句一個意思，砍掉不承重的子句。Watch: 超過約 30 字、或 3＋子句的句子。 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：保留、不要過度矯正〔來源：academic-humanizer v0.3.3 · Layer 3〕

一般 humanizer 會抹平正當的學術構造，這些要留：

• Evidence-tied hedging 是正確且必要的。 當主張確實不確定時，保留 "suggests", "is consistent with", "we hypothesize that", "may indicate", "appears to"。*錯誤修法：*把 "the results suggest X" 改成 "the results prove X"——那是製造 over-claim。保留校準過的動詞。
• 被動語態在動作者無關時沒問題："Samples were normalized to total protein."
• 第一人稱複數 "we" 是標準，不要為了避開它而重寫。
• 分號與偶爾的三連適量可用。Em-dashes 是例外：整個移除（Layer 1），改以逗號、冒號、括號或分句重寫。
• 形式定義、具名方法／指標、術語、方程式與符號照原樣保留。
• 絕不發明、刪除或更動任何數字、方程式或引用；保留每個 cite key。（本站由 pipeline 的 check_markers.py／check_numbers.py／check_citations.py 與 full check-consistency(en) 機械把關。）

Layer 5：語氣與期刊語域對齊〔來源：academic-humanizer v0.3.3 · Layer 5〕

若作者提供既有論文，先讀樣本，記下句子節奏、連接詞習慣、hedging 的程度與位置、如何開節、記號與慣用語，然後對齊。也對齊期刊語域（例如 ICLR／NeurIPS：簡練、直接、results-forward；Nature／PNAS：較說明性）。沒有樣本時，預設乾淨、精準、合乎期刊語域的散文，不要一般 humanizer 那種口語、帶意見的聲音。

---

來源與衍生說明（Provenance）

• Layer 1／2／3／5、核心原則、流程、何時用：源自 user skill academic-humanizer v0.3.3（MIT），逐節以〔來源：academic-humanizer v0.3.3 · …〕標註；文字大致照搬，僅移除與提案模式相關的交叉引用，並在 2.1／Layer 3 補上「改稿只在呈現層、證據強度回 claim-audit」的掛點註記。
• 「掛點合約」與「為何不含 claim-evidence 改寫」：academic-pipeline 專屬新增〔來源：academic-pipeline 潤稿站合約〕，對齊 translate-en／status 對潤稿站的既有規範。
• 未納入：academic-humanizer 的 Layer 4（claim-evidence discipline）不進本站；其職責在 pipeline 內由 claim-audit／interpret-results／check_numbers.py 承擔。
• 另行抽出：academic-humanizer 的 Layer 6（NSF／NIH 提案模式，含 6.1–6.5）抽為獨立 skill grant-proposal-editing，不隨本 plugin（正交於研究→投稿流程）。

Output

輸出潤飾後英文稿與簡短 change report：移除的 pattern（分類）、claim 相關但未改而標記回 claim-audit 的項目、語氣／語域註記，並確認無數字、方程式或引用遭更動。