rohanpaul_ai· @rohanpaul_ai · X·· 2 天前AI 评分
Pangram 漏检 79.8% 由 Meta Muse-Glimmer 改写的学术摘要
东京 Metropolitan University 的一篇新论文发现，AI 文本检测器 Pangram 漏检了 79.8% 由 Meta 的 Muse-Glimmer 改写的科学摘要，同时在 5,000 篇人类摘要中仅误报 1 篇。论文指出，Pangram 对 GPT-5 改写文本的检出率为 93.5%，漏检率主要取决于改写所用的模型版本。
Pangram, the AI-text detector, missed 79.8% of scientific abstracts rewritten by Meta's Muse-Glimmer, while flagging just 1 of 5,000 human abstracts.
In a new paper from Tokyo Metropolitan University, reseaerchers find the share of AI-rewritten abstracts that Pangram misses depends strongly on the LLM version
Shows that it caught 93.5% of GPT-5 rewrites but missed 79.8% from another new model. Its miss rate depended mostly on which model did the rewriting.
来源：rohanpaul_ai · x.com