怪谈
卡迪夫大学新研究：AI 尚无法像人类教师一样可靠地批改作业_我的网站
一 | Saiyaara actors Ahaan Panday and Aneet Padda have surpassed big names such as Aamir Khan and Rishab Shetty, securing the top two positions on IMDb's Most Popular Indian Stars of 2025 list, after the huge success of their romantic drama. IMDb recently announced the most popular Indian stars of 2025, determined by page views from over 250 million monthly visitors globally and Ahaan and Aneet secured the top two positions. In the list, Aamir Khan is positioned at number three, followed by Ishaan Khatter and Lakshya in the fourth and fifth spots, respectively. Rashmika Mandanna holds the sixth position, with Kalyani Priyadarshan at seventh. Triptii Dimri, Rukmini Vasanth, and Rishab Shetty complete the list, occupying the eighth, ninth, and tenth places. Ahaan and Aneet are delighted to have secured the top positions and expressed their gratitude to Saiyaara director Mohit Suri for his confidence in their abilities. Ahaan shared, “This is deeply humbling for me. To be ranked No. 1 on the list with my first film is nothing short of a dream come true and a full circle Paulo Coelho moment. But truly, if anything, this recognition makes me a little more aware of the responsibility I have towards my craft, and at the same time, it excites me for what’s to come.” He added, “I would like to thank my director Mohit Suri for allowing me to add to his canvas. Whatever I did on screen was a reflection of his guidance and his genius… I look forward to returning soon with my second film, and this honour is motivation enough to try and deliver a performance that will resonate deeply.” Aneet expressed her gratitude, saying, “Saiyaara changed my life in ways I am only beginning to understand, and to know that people across countries and languages connected with my work means more to me than I can explain.” She added, “I am deeply grateful to everyone who found something to love in Vaani. Your warmth has made this moment possible for me. I want to thank Mohit Suri for seeing something in me before I saw it in myself. The faith in his eyes on set gave me a strange, steady courage. Saiyaara made me braver as an actor.” Also Read: Photos: Ahaan Panday and Aneet Padda’s Sweet Fan Moment Goes Viral。 8 月 24 日消息，据外媒 Phys.org 今天（24 日）报道，卡迪夫大学和墨尔本大学的一项新研究发现，生成式人工智能（GenAI）目前还无法像人类教师一样可靠地评判学生的书面作业。
二 | 研究人员使用 ChatGPT 的两个版本，对 50 篇生物科学专业本科生论文进行评分，以测试大模型评估学生作业的能力。ChatGPT 需要按照 7 项标准批改这些论文，研究人员还采用了 4 种不同的提示方式，随后将大模型与人工评分的平均分和分数波动情况进行比较。
三 | 图源：Pexels卡迪夫大学生物科学学院威廉 · 凯博士指出：“研究结果显示，模型给出的分数波动很大，无法准确预测人工评分。
四 | 生成式 AI 与人类批改同一批论文时存在明显差异。只看论文总分，两者给出的结果相对接近；一旦具体到每项评分标准和每一篇论文，大模型与人工评分之间的差距就相当明显。
五 | ”除一种情况外，AI 模型给出的平均分普遍高于人工评分。平均分方面，AI 模型与人工评分的最大差距达到 16.1 分；具体到单篇论文，最大差距达到 40 分。获悉，威廉 · 凯还发现，AI 往往会压低高分论文的成绩，又把低分作业的成绩抬高，最终使分数系统性地向中间集中。研究结果说明，至少现阶段，即使经过大量训练，AI 仍无法可靠地对主观性较强的书面作业给出与人类相当的分数。“我们测试的大模型目前并不适合取代教师，为学生预测作业成绩。”研究人员指出，高等教育领域确实很关注大模型能否利用模式识别能力，让学生作业评分更加客观，同时提高批改效率、减轻教职人员压力。
六 | 但这项研究表明，目前并不适合这么做。此外，未经学生明确同意便把作业提交给 AI 工具，本身还涉及伦理问题；大模型也不能、也不应该被依赖来给学生的长篇书面作业评分。“随着大模型继续发展，未来模仿人类判断的能力可能会提高。但从这项研究来看，要让 AI 给出的分数与人工评分真正保持一致，恐怕并不容易。
Current article:http://www.chenzhuaigecoupuhongyuntuizei.sbs/8aah/osehe.docx
Published on:15:33:07
