ChatGPT's ability to generate realistic experimental images poses a new challenge to academic integrity.
ChatGPT生成逼真實驗圖像的能力對學術誠信提出了新挑戰。
J Hematol Oncol 2024-05-01
AdversaFlow: Visual Red Teaming for Large Language Models with Multi-Level Adversarial Flow.
AdversaFlow:針對大型語言模型的多層對抗流可視化紅隊測試。
IEEE Trans Vis Comput Graph 2024-09-16
Investigating generative AI models and detection techniques: impacts of tokenization and dataset size on identification of AI-generated text.
探討生成式 AI 模型及檢測技術:標記化和數據集大小對識別 AI 生成文本的影響。
Front Artif Intell 2024-12-04
這項研究探討了生成式 AI 模型(如 ChatGPT、Gemini 和 Claude)在 K-12 教育中的應用,強調其在各科目的優勢,並討論學術不誠實的倫理問題。研究使用傳統機器學習模型和大型語言模型來檢測高風險寫作評估中的 AI 生成內容,並評估檢測方法的效果,考慮改寫工具的影響。研究還引入新方法,利用同義詞資訊識別人性化的 AI 文字,並探討數據集大小對模型表現的影響,以指導未來的數據收集。
PubMedDOI
Improving authenticity and provenance in digital biomarkers: the case for digital watermarking.
提升數位生物標記的真實性和來源:數位水印的案例。
NPJ Digit Med 2025-01-15