GPTZero
PulseAugur coverage of GPTZero — every cluster mentioning GPTZero across labs, papers, and developer communities, ranked by signal.
2 天有情绪数据
-
安永因 AI 生成的虚假来源撤回网络安全报告
据报道，安永 (Ernst & Young, EY) 在 GPTZero 发现大量虚假来源后撤回了一份网络安全报告。此事件凸显了治理问题，而非 AI 本身的固有缺陷，强调了在使用生成模型进行内容创作时，严格的来源验证和审查流程的必要性。
-
新研究表明，基础AI模型可逃避检测
一项新的研究论文揭示，与经过指令微调的模型不同，基础AI模型常被GPTZero和Pangram等流行的AI文本检测器误判为人类。该研究提出了一种名为“迭代释义人性化”（Humanization by Iterative Paraphrasing, HIP）的方法，用于微调基础模型使其成为释义器，然后迭代地优化生成的文本以逃避检测。该技术在不同尺寸的Llama-3和Qwen-3模型上进行了测试，证明了其在保留语义意义的同时提高了检测逃避…
-
EY's AI-generated report riddled with hallucinations
A report by consulting firm EY on cybersecurity threats in loyalty systems has been found to contain numerous AI-generated hallucinations, citing nonexistent sources. This discovery raises concerns about the reliability…
-
AI Doesn’t Have Perspective (And Your Readers Can Tell Before Any Detector Can) Detector tools didn't catch my AI proposal. My client did. Here's what your read
A consultant discovered that using AI to write a client proposal, despite having done the underlying work, resulted in the client perceiving it as low-effort and insincere. This experience highlights that readers can de…