{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"ArXiv上超30%新投稿文本特征与AI撰写一致","description":"一项对12，750篇ArXiv论文全文的检测显示，截至2026年7月，约32%的新投稿文本特征与AI撰写一致，该比例在2026年初峰值接近39%。计算机科学领域最高（65%），数学领域最低（0.7%）。检测器在0.4%假阳性率下可识别85%的AI学术文本。","url":"https://www.aioga.com/news/cmrtkxcfk49mmbitl5rnaga7n/","mainEntityOfPage":"https://www.aioga.com/news/cmrtkxcfk49mmbitl5rnaga7n/","datePublished":"2026-07-20T18:28:23.600Z","dateModified":"2026-07-20T18:28:23.600Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://unslop.run/blog/measuring-ai-writing-on-arxiv","https://aihot.virxact.com/items/cmrtkxcfk49mmbitl5rnaga7n"],"canonicalUrl":"https://www.aioga.com/news/cmrtkxcfk49mmbitl5rnaga7n/","directAnswer":{"@type":"Answer","text":"研究对12,750篇arXiv论文的第一版全文进行评分。按其校准阈值，截至2026年7月，最近完整季度约32%的新投稿超过检测阈值，2026年初一度接近39%，但这表示文本特征与机器撰写一致，并非直接证明论文由AI生成。","url":"https://www.aioga.com/news/cmrtkxcfk49mmbitl5rnaga7n/","dateCreated":"2026-07-20T18:28:23.600Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"unslop.run source article","url":"https://unslop.run/blog/measuring-ai-writing-on-arxiv","datePublished":"2026-07-20T18:28:23.600Z","provider":{"@type":"Organization","name":"unslop.run","url":"https://unslop.run/blog/measuring-ai-writing-on-arxiv"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrtkxcfk49mmbitl5rnaga7n","datePublished":"2026-07-20T18:28:23.600Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrtkxcfk49mmbitl5rnaga7n"}}],"aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","originalPublisher":{"name":"unslop.run","url":"https://unslop.run/blog/measuring-ai-writing-on-arxiv"},"article":{"id":"cmrtkxcfk49mmbitl5rnaga7n","slug":"cmrtkxcfk49mmbitl5rnaga7n","url":"https://www.aioga.com/news/cmrtkxcfk49mmbitl5rnaga7n/","title":"ArXiv上超30%新投稿文本特征与AI撰写一致","title_en":"目前，ArXiv上超过30%的新投稿读起来像是人工智能撰写的","summary":"一项对12，750篇ArXiv论文全文的检测显示，截至2026年7月，约32%的新投稿文本特征与AI撰写一致，该比例在2026年初峰值接近39%。计算机科学领域最高（65%），数学领域最低（0.7%）。检测器在0.4%假阳性率下可识别85%的AI学术文本。","source":"Hacker News 热门（buzzing.cc 中文翻译）","sourceUrl":"https://unslop.run/blog/measuring-ai-writing-on-arxiv","aiHotUrl":"https://aihot.virxact.com/items/cmrtkxcfk49mmbitl5rnaga7n","publishedAt":"2026-07-20T18:28:23.600Z","category":"论文研究","score":74,"selected":true,"articleBody":["We scored the full text of 12,750 arXiv papers and found that about a third of new ones read as machine-written. Here is the method, the results, and an honest account of the limitations.","There is a genre of headline that says \"N% of X is now AI,\" and most are not worth reading, because the detector behind the number also flags some share of genuine human writing. If a tool marks 40% of new papers as machine-written but also marks 20% of papers written before ChatGPT existed, the real story is the 20% nobody mentioned.","So we built the study around that objection. Our detector, described here：/blog/how-our-ai-text-detector-works, is calibrated for academic writing; at a 0.4% false-positive rate it clears 99.6% of genuine pre-LLM scientific text and recovers 85% of AI academic text. We made that false-positive rate the anchor. We took papers submitted in 2021 and 2022, before ChatGPT, treated them as ground-truth human, and set the flag threshold so that exactly 0.4% of them trip it. That line is the floor. Every number we report is a share of papers above a threshold where genuine pre-LLM writing sits, by construction, at 0.4%. The pre-ChatGPT years then act as a built-in control: if the rise were an artifact of the detector, 2021 and 2022 would flag as high as 2026. The first figure shows they do not.","We sampled ten field groups, roughly 25 papers per field per month, from January 2023 to July 2026, plus eight control months across 2021 and 2022, for 12,750 papers in total. For each one we pulled the version-1 PDF, so a paper revised in 2026 cannot leak modern text back into its 2023 slot. We scored the full body text instead of the abstract, because abstracts understate the signal: we have seen the same paper score under 20% on its abstract and over 70% on its body. Every reported figure carries a bootstrap 95% confidence interval.","The flagged share is flat at 0.4% through 2021 and 2022, lifts off within months of ChatGPT, and climbs in two waves to about 32% over the most recent complete quarter, peaking near 39% in early 2026. The spread across fields is large, and it is the table and the second figure that carry it. The values below are each field's flagged share over the 12 months to July 2026, alongside its pre-LLM control level.","Computer science leads at about 65%. Mathematics is lowest, near 0.7%, and the limitations section explains why its low value is hard to interpret. The control column is each field's 2021 to 2022 flag rate averaged over three sensitivity settings; the fields that rise most are not the ones with the highest pre-LLM control level, so an elevated starting point does not explain the rise.","Control sample size. Each field's pre-ChatGPT control is 200 papers. At a 0.4% flag rate only eight papers flag across the entire 2,000-paper control, spread thinly over ten fields, so a single-threshold per-field control rate is coarse. The pooled floor is well estimated and is what the study is anchored to, but the per-field control levels are only approximate, and a larger control would not fix this: pinning a fraction-of-a-percent rate per field would require thousands of control papers per field that pre-2023 arXiv volume does not contain.","A low score can indicate low adoption or a detector blind spot. Mathematics is the clearest case. Mathematics papers are dominated by notation and theorem-proof structure, and once equations and references are removed the remaining prose is sparse and unlike the scientific English the detector was trained on. A mathematics paper drafted with heavy model assistance may score low because its prose is out of distribution for the detector, so a low score in mathematics is weak evidence that a human wrote the paper. The result is consistent with two very different explanations, lower adoption or reduced detector sensitivity in that register, and this data cannot separate them. The fields with the strongest in-distribution assumption, the prose-heavy ones, are also the ones that rise most, so this confound does not account for the aggregate trend. But in the low-scoring fields the ranking should be read as a lower bound on adoption.","Detector coverage. The detector is more sensitive to some generators than others, and we cannot evaluate it against the exact, private mixture of models and prompts that authors actually use. Incomplete coverage lowers the flag rate, so the reported prevalence is a lower bound: the true share is at least what we measured. The detector write-up：/blog/how-our-ai-text-detector-works reports the per-generator performance.","A flag is not authorship. The detector estimates whether text reads as machine-written, at a calibrated probability with a known error rate. It cannot separate a lightly-edited document from a wholly-generated one, and a single score is never grounds to accuse a specific person. We report the prevalence of machine-like writing, which includes heavy AI-assisted editing.","The detector is cheap to run and we make no money from it. You can try it for free on any arXiv paper here：/check, and on your own text here：/app.","Run unslop on your own text →：/app"],"articleImages":[{"sourceUrl":"https://unslop.run/blog/assets/paper-trend.png","alt":"Share of new arXiv papers flagged as machine-written, 2021 to 2026, at a threshold calibrated so pre-ChatGPT papers flag at 0.4%. The eight slate points are the pre-LLM control months; the band is a bootstrap 95% interval.","afterParagraph":0,"url":"/media/articles/cmrtkxcfk49mmbitl5rnaga7n/26c8129f65553227.png"},{"sourceUrl":"https://unslop.run/blog/assets/paper-fields.png","alt":"Share of papers flagged as machine-written by field, over the 12 months to July 2026 (about 300 papers per field).","afterParagraph":0,"url":"/media/articles/cmrtkxcfk49mmbitl5rnaga7n/b94de4d8be0cace2.png"}],"mediaStatus":"ok","articleBodyZh":["我们评分了12,750篇arXiv论文的全文，发现大约三分之一的新论文读起来像是机器写的。以下是方法、结果以及对局限性的诚实说明。","有一种标题类型叫做“X的N％现在是AI”，大多数都不值得读，因为数字背后的检测器也会误判一些真正的人类写作作品。如果一个工具标记40%的新论文为机器写作，但也标记了20%的ChatGPT出现之前的论文，那么真正值得关注的故事其实是没人提到的那20%。","所以我们围绕这一质疑建立了研究。我们的检测器，如此处描述：/blog/how-our-ai-text-detector-works，针对学术写作进行了校准；在0.4%的假阳性率下，它可以清除99.6%的真实预LLM科学文本，并检测出85%的AI学术文本。我们以这个假阳性率为基准。我们选取了2021年和2022年提交的论文，这些是在ChatGPT之前，被视为真实的人类作品，并设置标记阈值，使得正好0.4%的论文触发它。这个线就是底线。我们报告的每个数字都是超过预LLM真实写作阈值的论文比例，这个阈值构造上为0.4%。ChatGPT之前的年份则充当内置对照：如果增长是检测器的假象，2021年和2022年的标记率会和2026年一样高。第一张图显示事实并非如此。","我们抽样了十个领域组，每个领域每月大约25篇论文，从2023年1月到2026年7月，加上2021年和2022年的八个月控制数据，总计12,750篇论文。对于每篇论文，我们提取了第1版PDF，因此2026年修订的论文不会将现代文本泄露到其2023年的位置。我们评分的是全文，而非摘要，因为摘要低估了信号：我们曾看到同一篇论文摘要得分不到20%，而正文得分超过70%。每个报告的数据都带有自助法95%的置信区间。","标记比例在2021年和2022年保持在0.4%的水平，在ChatGPT发布后的几个月内开始上升，并以两个波段攀升到最近一个完整季度约32%，在2026年初达到接近39%的峰值。各领域间的差异很大，可以通过表格和第二张图看出。下列数值分别为各领域截至2026年7月的12个月内被标记的比例，以及其预LLM控制水平。","计算机科学领先约65%。数学最低，接近0.7%，限制部分解释了其低值难以解释的原因。控制列是每个领域2021到2022年标记率在三个敏感性设置下的平均值；上升幅度最大的领域并非预LLM控制水平最高的领域，因此较高的起点并不能解释这种上升。","控制样本量。每个领域的ChatGPT前控制为200篇论文。在0.4%的标记率下，整个2,000篇控制论文中仅有八篇被标记，分布在十个领域中，因此每个领域单一阈值的控制率是粗略的。汇总下限估计良好，也是研究的锚定点，但每个领域的控制水平只是近似值，更大的控制样本也无法解决这个问题：要在每个领域确定一个千分之一的标记率，需要每个领域数千篇控制论文，而2023年前的arXiv数量不具备。","低分可以表示低采纳率或检测器盲点。数学是最明显的案例。数学论文以符号和定理-证明结构为主，一旦删除方程和引用，其余的文字稀疏且与检测器训练的科学英语不同。一篇在大量模型辅助下撰写的数学论文可能得分低，因为其文字不在检测器的分布范围内，因此数学中的低分并不能强有力地证明该论文由人类撰写。结果可以用两个完全不同的解释来理解：采纳率较低，或者在该领域检测器敏感度降低，而这些数据无法区分它们。那些符合分布假设最强的、以文字为主的领域也是增长最多的领域，因此这种混杂因素无法解释整体趋势。但在低分领域中，排名应被视为采用率的下限。","检测器覆盖范围。检测器对某些生成器比其他生成器更敏感，而我们无法针对作者实际使用的确切、私有的模型和提示组合进行评估。不完整的覆盖会降低标记率，因此报告的普及率是下限：真实比例至少与我们测得的相同。检测器说明：/blog/how-our-ai-text-detector-works 报告了按生成器划分的性能。","一个标记并不等同于作者身份。检测器会以已校准的概率估计文本是否像机器生成的，并且具有已知的错误率。它无法区分经过轻度编辑的文档与完全生成的文档，而且单一评分永远不能作为指控某个特定人的依据。我们报告机器化写作的普遍性，其中包括大量 AI 辅助编辑的内容。","该检测器运行成本低，我们并不从中盈利。你可以在这里免费尝试任意 arXiv 论文：/check，也可以在这里尝试你自己的文本：/app。","在你自己的文本上运行 unslop →：/app"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"研究对12,750篇arXiv论文的第一版全文进行评分。按其校准阈值，截至2026年7月，最近完整季度约32%的新投稿超过检测阈值，2026年初一度接近39%，但这表示文本特征与机器撰写一致，并非直接证明论文由AI生成。","background":"样本覆盖十个学科组、2023年1月至2026年7月，并以2021至2022年的论文作为前大模型时期对照。检测器在设定的0.4%假阳性率下，可识别85%的AI学术文本；研究使用第一版PDF全文，并为结果提供自助法95%置信区间。","viewpoint":"Aioga判断，这项研究的价值在于将前大模型论文作为内置对照，并主动披露假阳性率，较单纯公布“AI占比”更具解释力。但超过阈值仍只是检测结果，不能据此认定具体作者使用了AI，也不能等同于学术不端。","implications":"计算机科学近12个月的标记比例约65%，数学约0.7%，但跨学科排名不能直接解释为真实采用率差异。数学论文的符号、定理与证明结构可能偏离检测器训练分布，因此低分既可能代表较少采用，也可能来自检测盲区。","nextStep":"值得关注的是扩大各学科的前大模型对照样本，并用不同检测方法、人工评审及已知写作来源的数据进行交叉验证。后续解读还应持续区分“超过检测阈值”“使用模型辅助”和“由AI撰写”，避免把相关文本特征升级为作者行为事实。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-07-22T21:51:30.841Z","sourceHash":"631c5250f2d01699","review":{"approved":true,"groundedness":94,"clarity":95,"duplicationRisk":18,"blockingIssues":[],"notes":["“扩大各学科的前大模型对照样本”属于编辑建议而非来源结论；来源还指出，要精确估计各学科极低的对照标记率，需要每个学科数千篇论文，而当时的arXiv论文量不足。可补充这一现实限制，避免让读者误以为简单扩大样本即可解决问题。","“不能等同于学术不端”是合理的审慎判断，但并非来源材料直接讨论的结论；当前已明确置于观点段，不构成阻断问题。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["论文研究","Hacker News 热门（buzzing.cc 中文翻译）"],"translations":{"zh-CN":{"title":"ArXiv上超30%新投稿文本特征与AI撰写一致","summary":"一项对12，750篇ArXiv论文全文的检测显示，截至2026年7月，约32%的新投稿文本特征与AI撰写一致，该比例在2026年初峰值接近39%。计算机科学领域最高（65%），数学领域最低（0.7%）。检测器在0.4%假阳性率下可识别85%的AI学术文本。","category":"论文研究","source":"unslop.run","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"ArXiv上超30%新投稿文本特征与AI撰写一致 - Aioga AI资讯","description":"一项对12，750篇ArXiv论文全文的检测显示，截至2026年7月，约32%的新投稿文本特征与AI撰写一致，该比例在2026年初峰值接近39%。计算机科学领域最高（65%），数学领域最低（0.7%）。检测器在0.4%假阳性率下可识别85%的AI学术文本。","url":"https://www.aioga.com/news/cmrtkxcfk49mmbitl5rnaga7n/"},"en":{"title":"Over 30% of new submissions on ArXiv show textual features consistent with AI authorship","summary":"An analysis of the full text of 12,750 ArXiv papers shows that, as of July 2026, about 32% of newly submitted texts exhibit features consistent with AI authorship, with the proportion peaking near 39% in early 2026. The highest rate is in the field of computer science (65%) and the lowest in mathematics (0.7%). Detectors can identify 85% of AI-generated academic texts at a 0.4% false positive rate.","category":"Research","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Over 30% of new submissions on ArXiv show textual features consistent with AI authorship - Aioga AI News","description":"An analysis of the full text of 12,750 ArXiv papers shows that, as of July 2026, about 32% of newly submitted texts exhibit features consistent with AI authorship, with the proport...","url":"https://www.aioga.com/en/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:06:28.096Z"},"ja":{"title":"ArXiv上の30%以上の新規投稿のテキスト特徴がAIによる作成と一致","summary":"12,750本のArXiv論文全文に対する調査によると、2026年7月時点で約32%の新規投稿テキストの特徴がAIによる執筆と一致しており、この割合は2026年初めにピークで39%近くに達していた。コンピューターサイエンス分野が最も高く（65%）、数学分野が最も低い（0.7%）。検出器は0.4%の偽陽性率で85%のAI学術テキストを識別できる。","category":"論文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"ArXiv上の30%以上の新規投稿のテキスト特徴がAIによる作成と一致 - Aioga AIニュース","description":"12,750本のArXiv論文全文に対する調査によると、2026年7月時点で約32%の新規投稿テキストの特徴がAIによる執筆と一致しており、この割合は2026年初めにピークで39%近くに達していた。コンピューターサイエンス分野が最も高く（65%）、数学分野が最も低い（0.7%）。検出器は0.4%の偽陽性率で85%のAI学術テキストを識別できる。","url":"https://www.aioga.com/ja/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:06:44.866Z"},"ko":{"title":"ArXiv에서 새로 제출된 글의 30% 이상이 AI 작성과 일치하는 텍스트 특징을 보임","summary":"12,750편의 ArXiv 논문 전체를 대상으로 한 한 번의 검출 결과, 2026년 7월 기준으로 약 32%의 신규 제출 텍스트 특성이 AI 작성과 일치했으며, 이 비율은 2026년 초에 최고 39%에 달했다. 컴퓨터 과학 분야가 가장 높았으며(65%), 수학 분야가 가장 낮았다(0.7%). 탐지기는 0.4%의 오탐률에서 85%의 AI 학술 텍스트를 식별할 수 있다.","category":"연구","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"ArXiv에서 새로 제출된 글의 30% 이상이 AI 작성과 일치하는 텍스트 특징을 보임 - Aioga AI 뉴스","description":"12,750편의 ArXiv 논문 전체를 대상으로 한 한 번의 검출 결과, 2026년 7월 기준으로 약 32%의 신규 제출 텍스트 특성이 AI 작성과 일치했으며, 이 비율은 2026년 초에 최고 39%에 달했다. 컴퓨터 과학 분야가 가장 높았으며(65%), 수학 분야가 가장 낮았다(0.7%). 탐지기는 0.4%의 오탐률에서...","url":"https://www.aioga.com/ko/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:07:34.910Z"},"es":{"title":"Más del 30% de los nuevos envíos en ArXiv tienen características de texto consistentes con la escritura de IA","summary":"Un análisis de texto completo de 12.750 artículos de ArXiv mostró que, hasta julio de 2026, aproximadamente el 32% de los textos de nuevas presentaciones presentaban características consistentes con la redacción por IA, un porcentaje que alcanzó un máximo cercano al 39% a principios de 2026. El campo de la informática tiene el porcentaje más alto (65%) y el de matemáticas el más bajo (0,7%). Los detectores pueden identificar el 85% de los textos académicos generados por IA con una tasa de falsos positivos del 0,4%.","category":"Investigación","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Más del 30% de los nuevos envíos en ArXiv tienen características de texto consistentes con la escritura de IA - Aioga Noticias de IA","description":"Un análisis de texto completo de 12.750 artículos de ArXiv mostró que, hasta julio de 2026, aproximadamente el 32% de los textos de nuevas presentaciones presentaban característica...","url":"https://www.aioga.com/es/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:07:32.980Z"},"fr":{"title":"Plus de 30 % des nouveaux textes soumis sur ArXiv présentent des caractéristiques cohérentes avec la rédaction par IA","summary":"Une analyse de 12 750 articles complets sur ArXiv a révélé qu'en juillet 2026, environ 32 % des nouveaux textes soumis présentaient des caractéristiques cohérentes avec une rédaction par IA, ce pourcentage ayant atteint un pic proche de 39 % au début de 2026. Le domaine de l'informatique est le plus élevé (65 %), tandis que celui des mathématiques est le plus bas (0,7 %). Le détecteur peut identifier 85 % des textes académiques générés par IA avec un taux de faux positifs de 0,4 %.","category":"Recherche","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Plus de 30 % des nouveaux textes soumis sur ArXiv présentent des caractéristiques cohérentes avec la rédaction par IA - Aioga Actualités IA","description":"Une analyse de 12 750 articles complets sur ArXiv a révélé qu'en juillet 2026, environ 32 % des nouveaux textes soumis présentaient des caractéristiques cohérentes avec une rédacti...","url":"https://www.aioga.com/fr/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:08:17.474Z"},"de":{"title":"Über 30 % der neuen Einreichungen auf ArXiv weisen Textmerkmale auf, die mit AI-Verfassungen übereinstimmen","summary":"Eine Untersuchung von 12.750 vollständigen ArXiv-Artikeln zeigt, dass bis Juli 2026 etwa 32% der neuen Einreichungstexte Merkmale aufweisen, die mit KI-Verfassung übereinstimmen, wobei dieser Anteil Anfang 2026 mit fast 39% seinen Höhepunkt erreichte. Im Bereich der Informatik ist er am höchsten (65%), im Bereich der Mathematik am niedrigsten (0,7%). Detektoren können 85% der KI-Wissenschaftstexte bei einer Fehlpositivrate von 0,4% erkennen.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Über 30 % der neuen Einreichungen auf ArXiv weisen Textmerkmale auf, die mit AI-Verfassungen übereinstimmen - Aioga KI-News","description":"Eine Untersuchung von 12.750 vollständigen ArXiv-Artikeln zeigt, dass bis Juli 2026 etwa 32% der neuen Einreichungstexte Merkmale aufweisen, die mit KI-Verfassung übereinstimmen, w...","url":"https://www.aioga.com/de/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:08:21.025Z"},"pt-BR":{"title":"Mais de 30% dos novos envios no ArXiv apresentam características de texto consistentes com redação por IA","summary":"Uma análise de 12.750 artigos completos do ArXiv mostrou que, até julho de 2026, cerca de 32% dos novos textos submetidos apresentavam características consistentes com escrita por IA, com o pico chegando a quase 39% no início de 2026. A área de ciência da computação teve a maior proporção (65%), e a de matemática, a menor (0,7%). O detector é capaz de identificar 85% dos textos acadêmicos gerados por IA com uma taxa de falso positivo de 0,4%.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Mais de 30% dos novos envios no ArXiv apresentam características de texto consistentes com redação por IA - Aioga Notícias de IA","description":"Uma análise de 12.750 artigos completos do ArXiv mostrou que, até julho de 2026, cerca de 32% dos novos textos submetidos apresentavam características consistentes com escrita por...","url":"https://www.aioga.com/pt-BR/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:09:04.380Z"},"ru":{"title":"Более 30% новых текстов на ArXiv соответствуют особенностям написания ИИ","summary":"Проверка полного текста 12 750 статей на ArXiv показала, что к июлю 2026 года около 32% новых поданных текстов соответствуют характеристикам написанных ИИ, при этом в начале 2026 года этот показатель достиг пика в 39%. Наибольший показатель в области компьютерных наук (65%), наименьший в области математики (0,7%). Детектор может выявлять 85% академических текстов, созданных ИИ, при ложноположительном показателе 0,4%.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Более 30% новых текстов на ArXiv соответствуют особенностям написания ИИ - Aioga Новости ИИ","description":"Проверка полного текста 12 750 статей на ArXiv показала, что к июлю 2026 года около 32% новых поданных текстов соответствуют характеристикам написанных ИИ, при этом в начале 2026 г...","url":"https://www.aioga.com/ru/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:09:07.225Z"},"ar":{"title":"أكثر من 30٪ من النصوص الجديدة المرسلة على ArXiv تتوافق خصائصها مع كتابة الذكاء الاصطناعي","summary":"أظهر فحص كامل النصوص لـ 12,750 ورقة على ArXiv أنه حتى يوليو 2026، حوالي 32% من النصوص المقدمة حديثًا تتوافق خصائصها مع الكتابة بواسطة الذكاء الاصطناعي، بينما بلغ هذا النسبة ذروتها في أوائل 2026 بما يقارب 39%. كان أعلى معدل في مجال علوم الكمبيوتر (65%)، وأدنى معدل في مجال الرياضيات (0.7%). يمكن للكاشف التعرف على 85% من النصوص الأكاديمية المكتوبة بالذكاء الاصطناعي بمعدل إيجابيات كاذبة 0.4%.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"أكثر من 30٪ من النصوص الجديدة المرسلة على ArXiv تتوافق خصائصها مع كتابة الذكاء الاصطناعي - Aioga أخبار الذكاء الاصطناعي","description":"أظهر فحص كامل النصوص لـ 12,750 ورقة على ArXiv أنه حتى يوليو 2026، حوالي 32% من النصوص المقدمة حديثًا تتوافق خصائصها مع الكتابة بواسطة الذكاء الاصطناعي، بينما بلغ هذا النسبة ذروتها...","url":"https://www.aioga.com/ar/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:09:53.063Z"},"hi":{"title":"ArXiv पर 30% से अधिक नए सबमिशन पाठ की विशेषताएँ AI द्वारा लिखे गए समान हैं","summary":"对12,750篇ArXiv论文全文的一项检测显示，截至2026年7月，约32%的新提交文本特征与AI撰写一致，该比例在2026年初峰值接近39%。计算机科学领域最高（65%），数学领域最低（0.7%）。检测器在0.4%的假阳性率下可识别85%的AI学术文本。","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"ArXiv पर 30% से अधिक नए सबमिशन पाठ की विशेषताएँ AI द्वारा लिखे गए समान हैं - Aioga AI समाचार","description":"对12,750篇ArXiv论文全文的一项检测显示，截至2026年7月，约32%的新提交文本特征与AI撰写一致，该比例在2026年初峰值接近39%。计算机科学领域最高（65%），数学领域最低（0.7%）。检测器在0.4%的假阳性率下可识别85%的AI学术文本。","url":"https://www.aioga.com/hi/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:09:56.952Z"},"it":{"title":"Oltre il 30% dei nuovi articoli su ArXiv presenta caratteristiche del testo coerenti con la scrittura AI","summary":"Un'analisi di 12.750 articoli completi su ArXiv ha mostrato che, a luglio 2026, circa il 32% dei nuovi testi presentava caratteristiche coerenti con la scrittura AI, una percentuale che ha raggiunto un picco vicino al 39% all'inizio del 2026. Il tasso più alto si riscontra nel campo dell'informatica (65%), il più basso in matematica (0,7%). Il rilevatore può identificare l'85% dei testi accademici generati dall'AI con un tasso di falsi positivi dello 0,4%.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Oltre il 30% dei nuovi articoli su ArXiv presenta caratteristiche del testo coerenti con la scrittura AI - Aioga Notizie IA","description":"Un'analisi di 12.750 articoli completi su ArXiv ha mostrato che, a luglio 2026, circa il 32% dei nuovi testi presentava caratteristiche coerenti con la scrittura AI, una percentual...","url":"https://www.aioga.com/it/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:10:48.783Z"},"nl":{"title":"Meer dan 30% van de nieuwe ingediende teksten op ArXiv komt overeen met AI-schrijven","summary":"Een analyse van de volledige teksten van 12.750 ArXiv-papers toont aan dat tot juli 2026 ongeveer 32% van de nieuw ingediende teksten kenmerken van AI-schrijven vertoonde, met een piek van bijna 39% aan het begin van 2026. Het hoogste percentage wordt gevonden in de informatica (65%) en het laagste in de wiskunde (0,7%). Detectoren kunnen 85% van de AI-academische teksten identificeren bij een vals-positieve rate van 0,4%.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Meer dan 30% van de nieuwe ingediende teksten op ArXiv komt overeen met AI-schrijven - Aioga AI-nieuws","description":"Een analyse van de volledige teksten van 12.750 ArXiv-papers toont aan dat tot juli 2026 ongeveer 32% van de nieuw ingediende teksten kenmerken van AI-schrijven vertoonde, met een...","url":"https://www.aioga.com/nl/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:10:40.645Z"},"tr":{"title":"ArXiv'de yeni gönderilen metinlerin %30'undan fazlasının özellikleri AI tarafından yazılmışlarla tutarlı","summary":"12.750 ArXiv makalesinin tamamına yapılan bir tespit, Temmuz 2026 itibarıyla yaklaşık %32'sinin yeni gönderilen metin özelliklerinin AI tarafından yazılmış ile uyumlu olduğunu gösteriyor; bu oran 2026'nın başında zirve yaparak yaklaşık %39'a ulaştı. Bilgisayar bilimi alanı en yüksek (%65), matematik alanı en düşük (%0,7) seviyedeydi. Tespit aracı, %0,4 sahte pozitif oranında AI akademik metinlerin %85'ini tanıyabiliyor.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"ArXiv'de yeni gönderilen metinlerin %30'undan fazlasının özellikleri AI tarafından yazılmışlarla tutarlı - Aioga AI Haberleri","description":"12.750 ArXiv makalesinin tamamına yapılan bir tespit, Temmuz 2026 itibarıyla yaklaşık %32'sinin yeni gönderilen metin özelliklerinin AI tarafından yazılmış ile uyumlu olduğunu göst...","url":"https://www.aioga.com/tr/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:11:29.693Z"},"vi":{"title":"Hơn 30% các bản thảo mới trên ArXiv có đặc điểm văn bản giống với văn bản do AI viết","summary":"Một cuộc kiểm tra toàn văn đối với 12.750 bài báo ArXiv cho thấy, tính đến tháng 7 năm 2026, khoảng 32% các bài nộp mới có đặc trưng văn bản phù hợp với việc được AI viết, tỷ lệ này đạt đỉnh vào đầu năm 2026 gần 39%. Lĩnh vực khoa học máy tính cao nhất (65%), lĩnh vực toán học thấp nhất (0,7%). Bộ phát hiện có thể nhận dạng 85% văn bản học thuật do AI viết với tỷ lệ dương tính giả 0,4%.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Hơn 30% các bản thảo mới trên ArXiv có đặc điểm văn bản giống với văn bản do AI viết - Tin tức AI Aioga","description":"Một cuộc kiểm tra toàn văn đối với 12.750 bài báo ArXiv cho thấy, tính đến tháng 7 năm 2026, khoảng 32% các bài nộp mới có đặc trưng văn bản phù hợp với việc được AI viết, tỷ lệ nà...","url":"https://www.aioga.com/vi/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:11:34.830Z"},"id":{"title":"Lebih dari 30% teks makalah baru di ArXiv sesuai dengan penulisan AI","summary":"Sebuah deteksi terhadap 12.750 makalah ArXiv lengkap menunjukkan bahwa hingga Juli 2026, sekitar 32% dari teks kiriman baru memiliki karakteristik yang konsisten dengan penulisan AI, dengan puncak sekitar 39% pada awal 2026. Bidang ilmu komputer tertinggi (65%), bidang matematika terendah (0,7%). Detektor dapat mengidentifikasi 85% teks akademik AI dengan tingkat positif palsu 0,4%.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Lebih dari 30% teks makalah baru di ArXiv sesuai dengan penulisan AI - Berita AI Aioga","description":"Sebuah deteksi terhadap 12.750 makalah ArXiv lengkap menunjukkan bahwa hingga Juli 2026, sekitar 32% dari teks kiriman baru memiliki karakteristik yang konsisten dengan penulisan A...","url":"https://www.aioga.com/id/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:12:15.284Z"},"th":{"title":"บน ArXiv มากกว่า 30% ของข้อความการส่งใหม่ตรงกับลักษณะการเขียนของ AI","summary":"การตรวจสอบบทความ ArXiv ทั้งหมด 12,750 บทความแสดงให้เห็นว่า จนถึงเดือนกรกฎาคม 2026 ประมาณ 32% ของข้อความที่ส่งใหม่มีลักษณะคล้ายกับการเขียนโดย AI อัตรานี้เคยพุ่งสูงสุดเกือบ 39% ในช่วงต้นปี 2026 สาขาวิทยาการคอมพิวเตอร์มีอัตราสูงสุด (65%) ในขณะที่สาขาคณิตศาสตร์ต่ำสุด (0.7%) ตัวตรวจจับสามารถระบุข้อความทางวิชาการที่เขียนโดย AI ได้ 85% ที่อัตราเกิดผลบวกเทียม 0.4%","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"บน ArXiv มากกว่า 30% ของข้อความการส่งใหม่ตรงกับลักษณะการเขียนของ AI - ข่าว AI Aioga","description":"การตรวจสอบบทความ ArXiv ทั้งหมด 12,750 บทความแสดงให้เห็นว่า จนถึงเดือนกรกฎาคม 2026 ประมาณ 32% ของข้อความที่ส่งใหม่มีลักษณะคล้ายกับการเขียนโดย AI อัตรานี้เคยพุ่งสูงสุดเกือบ 39% ในช่ว...","url":"https://www.aioga.com/th/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:12:24.596Z"},"pl":{"title":"Ponad 30% nowych zgłoszeń na ArXiv ma cechy tekstu zgodne z twórczością AI","summary":"Badanie całego tekstu 12 750 artykułów z ArXiv wykazało, że do lipca 2026 roku około 32% nowych zgłoszeń miało cechy tekstu tworzonego przez AI, a wskaźnik ten osiągnął szczyt na początku 2026 roku, bliski 39%. Najwyższy odsetek zanotowano w dziedzinie informatyki (65%), najniższy w matematyce (0,7%). Detektor jest w stanie rozpoznać 85% tekstów akademickich stworzonych przez AI przy 0,4% wskaźniku fałszywie pozytywnych wyników.","category":"论文研究","source":"Hacker News 热门（buzzing.cc 中文翻译）","aggregationSource":"Hacker News 热门（buzzing.cc 中文翻译）","pageTitle":"Ponad 30% nowych zgłoszeń na ArXiv ma cechy tekstu zgodne z twórczością AI - Aioga Wiadomości AI","description":"Badanie całego tekstu 12 750 artykułów z ArXiv wykazało, że do lipca 2026 roku około 32% nowych zgłoszeń miało cechy tekstu tworzonego przez AI, a wskaźnik ten osiągnął szczyt na p...","url":"https://www.aioga.com/pl/news/cmrtkxcfk49mmbitl5rnaga7n/","contentTranslated":true,"sourceHash":"a8205a5766962663","translatedAt":"2026-07-22T17:13:13.091Z"}}}}