央广网北京7月15日消息（总台中国之声记者任梦岩）据中央广播电视总台中国之声《新闻纵横》 报道，论文"AI味"渐浓，是高校共同面对的新难题。今年的高校毕业生在毕业论文阶段，除了传统的查 重、盲审、答辩，毕业路上又多了一道关卡——AIGC检测，也就是人工智能生成内容检测。 不少高校引入了AI模型来"查文章AI生成内容占比"，而一些平台又推出了"降低现有文章AI生成内 容占比"的大模型，但一些师生在使用过程中发现，这些所谓"降低AI率"的大模型表现实在一般，有时 候越用大模型检测"AI率"越高。让AI查AI，靠谱吗？毕业论文检测，怎样更科学？ 韩同学表示："论文写完之后，我查AI率、查重率都很高，就想着用AI改一下，就在那个软件上花 了128元，希望降重复率和降AI率，两个都降，之后就会出来一个报告，那个报告显示（重复率和AI 率）只有百分之十几。结果我交上去之后，导师返给我说完全不合格。" 河南李同学的遭遇更为蹊跷。她的学校要求使用某"AI率检测平台"进行论文检测，第一次检测合 格，第二次检测，却被告知"AI生成内容占比过高"。 李同学说："我第一次检测的时候，它显示是一个低风险，AI率百分之十五点多，这是在论 ...

以下文章来源于APPSO ，作者发现明日产品的 APPSO . AI 第一新媒体，「超级个体」的灵感指南。 #AIGC #智能设备 #独特应用 #Generative AI 本文来自微信公众号： APPSO （ID：appsolution） ，作者：appso，原文标题：《为了证明没用 AI 我把论文改得面目全非，这届毕业生快被 AI 检测逼疯了｜附降重指南》，题图来自：AI生成 "天临七年"，毕业论文除了查重之外又多了一道关卡——AIGC检测 （人工智能生成内容检测） 。 为了避免学术不端，不少高校开始把AIGC检测结果作为论文能否通过的要求之一。 作为一名刚刚结束答辩的毕业生，前一段时间可是和AIGC检测周旋了许久。经历了"检测—修改— 再检测—再改"的痛苦循环后，终于从61.7%降到0%。 这个过程之所以这么让人崩溃，是因为AIGC检测真的不讲武德： 有些明明是自己一个字一个字敲下的，也会被整段标红，判定为AI生成；在这一个平台上测出的 10%，换个平台又可能测出100%；而哪怕是在同一个平台，同一个段落这次测是0%，下次可能就 变成100%。 这不禁让我深深怀疑，当前的AIGC检测和它最初的目的是否已 ...

Core Insights - The rise of AIGC detection for graduation theses has led to a new industry chain, with universities implementing AI content detection standards for students' papers [1][2] - Many students are using AI tools for various aspects of thesis writing, prompting universities to explore different AIGC detection requirements [2][3] - A gray market has emerged, where individuals and small agencies offer services to reduce AI content in theses, often at low prices [3][4] Group 1: AIGC Detection Implementation - Several universities, including Sichuan University and Beijing Normal University, have set limits on AI-generated content in theses, with humanities and social sciences capped at 20% and sciences at 15% [1] - Institutions like Fudan University require students to disclose the use of AI tools in their thesis commitment forms before defense [1][2] - The AIGC detection services are being offered alongside traditional plagiarism checks by academic websites such as CNKI, Weipu, and PaperPass [2][3] Group 2: Emergence of Gray Market - A significant number of students are turning to "降手" (AI reduction service providers) who charge around 10 yuan per 1000 characters to modify AI-detected content [4][6] - The market for these services has seen a surge, with many new entrants offering low-cost solutions, leading to a chaotic environment [10][11] - Some service providers are using AI tools themselves to modify texts, which raises questions about the effectiveness of these services [11][12] Group 3: Student Experiences and Challenges - Many students report dissatisfaction with the services, experiencing no reduction in AI detection rates or receiving poorly modified texts that distort original meanings [6][7] - The lack of objective metrics for AI content detection complicates the process, leading to confusion and frustration among students [7][8] - Students often face challenges in obtaining refunds or satisfactory service, as many providers operate on a low-cost model with minimal accountability [8][9]

以下文章来源于数字生命卡兹克 ，作者数字生命卡兹克 数字生命卡兹克 . 努力分享一些很新、很酷的AI干货，愿我们永远对世界保持好奇。 我其实一直都很赞成AI的快速发展，很少会看到，让我眉头一皱的AI应用场景。 甚至会有点出离了愤怒。 因为可能本心是好的，但是这个方法，却把好心，变成了一个让我非常痛心却又觉得无奈的事件。 这个事情就是，马上毕业季了，很多的学校， 为了整治学术不端行为，所以对学生们的论文，除了原来的查重检测之外，引入了AIGC检测。 大概就是用一些所谓的AI检测工具，来检测你的论文里，AI生成的含量有多少。 如果你的AIGC检测比例，如果高于一定的指标，就会无法毕业。 起初我以为这种只是一些造谣的图，在我实际查证之后，我发现，是真的。 已经有多所大学，启动了AIGC检测，并且有明确的指标。 本文来自微信公众号： 数字生命卡兹克 ，作者：数字生命卡兹克，原文标题：《看到现在的毕业生被AIGC查重折磨，我有话想说。》，题图来自： AI生成 比如4月9日，四川大学教务处发布的《关于开展2025届本科毕业论文（设计）学术不端行为检测工作的通知》中，就明确提到了，20%和15%这两个 比例。 不止一个，还 ...

Group 1 - The article discusses the implementation of AIGC detection tools in universities to combat academic misconduct, particularly in student thesis submissions [4][9][10] - Several universities have set specific thresholds for AIGC detection, such as 20% and 15%, which can impact students' graduation eligibility [10][6] - The author expresses concern over the effectiveness and reliability of AIGC detection methods, arguing that they may misjudge human-written content as AI-generated [20][22][62] Group 2 - The article critiques the underlying algorithms of AIGC detection tools, categorizing them into three main types: perplexity and entropy analysis, machine learning classifiers, and syntactic and stylistic feature modeling [26][35][39] - The author highlights the absurdity of using AI to judge AI-generated content, emphasizing that the detection systems often fail to account for the nuances of human writing [21][34][66] - There is a significant concern that educational institutions may not fully understand the limitations of these detection tools, leading to unfair consequences for students [55][56][62] Group 3 - The article argues that the reliance on AIGC detection tools reflects a broader trust crisis in the educational system regarding the use of AI [64][67] - The author believes that the current approach to AIGC detection prioritizes algorithmic judgment over human effort and creativity, which could stifle genuine academic expression [70][73] - The piece concludes with a warning about the potential future where individuals may feel compelled to prove their originality through surveillance rather than trust [71][72]

Group 1 - The core issue revolves around the increasing use of AI tools in education, leading to concerns about academic integrity and the effectiveness of AI detection systems [1][2][3] - Reports indicate that classic literary works have high AI-generated content similarity rates, with examples showing 62.88% for "Lotus Pond Moonlight" and 52.88% for "The Wandering Earth" [1] - Educational institutions are responding by implementing regulations on AI usage in assignments, with some universities prohibiting direct use of AI-generated content in theses [2][3] Group 2 - The phenomenon of "using AI to detect AI" raises questions about the accuracy of current detection methods and their ability to distinguish between original and AI-generated content [2][3] - There are emerging tools designed to reduce AI detection rates in written work, indicating a cycle of AI usage and countermeasures in academic settings [3] - The over-reliance on AI tools in education may undermine students' critical thinking and creativity, prompting a need for a balanced approach to AI integration [4] Group 3 - The integration of AI in education is seen as an unstoppable trend, with a focus on fostering AI literacy and talent among students [4][5] - Educational reforms are underway, with initiatives like the Nanjing AI Education Action Plan aiming for 100% AI course coverage in primary and secondary schools by 2027 [5] - There is a call for a comprehensive and scientific approach to AI education, emphasizing the importance of developing a healthy understanding of human-AI collaboration [5]

我其实一直都很赞成AI的快速发展，很少会看到，让我眉头一皱的AI应用场景。 甚至会有点出离了愤怒。 因为可能本心是好的，但是这个方法，却把好心，变成了一个让我非常痛心却又觉得无奈的事件。 这个事情就是，马上毕业季了，很多的学校，为了整治学术不端行为，所以对学生们的论文，除了原来的查重检测之外，引入了AIGC检测。 大概就是用一些所谓的AI检测工具，来检测你的论文里，AI生成的含量有多少。 如果你的AIGC检测比例，如果高于一定的指标，就会无法毕业。 起初我以为这种只是一些造谣的图，在我实际查证之后，我发现，是真的。 已经有多所大学，启动了AIGC检测，并且有明确的指标。 比如4月9日，四川大学教务处发布的《关于开展2025届本科毕业论文（设计）学术不端行为检测工作的通知》中，就明确提到了，20%和15%这两个比 例。 不止一个，还有很多。 如果你去Google上搜一个关键局： 关于2025届本科毕业论文(设计)试行AIGC检测的通知。 你就能看到，大概有哪些学校，在开始实行AIGC检测了。 你如果去小红书在看看，搜索关键词：AIGC论文。 能搜到一大堆的吐槽贴，还有掺杂在里面想捞一笔的所谓几款钱降重的广告。 ...