发表机构
Pangram Labs; University of Maryland（Pangram实验室; 马里兰大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
Pangram Labs 推出的 Pangram 4 是基于深度学习的 AI 文本分类模型，其在 AUROC、误判率等指标上优于前代模型，具备更强的泛化与鲁棒性，在 AI 文本检测任务中达到当前最优性能。
AI 中文摘要
我们推出了 Pangram 4，这是 Pangram Labs 最新的基于深度学习的 AI 文本分类模型。该模型的 AUROC 达到 0.9916，假阳性率为 0.0041%，假阴性率为 0.3396%。与 Pangram 3 相比，Pangram 4 不仅整体准确率更高，还展现出更优的分布外泛化能力和对抗攻击鲁棒性。Pangram 4 的另一项创新贡献是其区分细粒度编辑内容以及 AI 与人类合著混合文本的能力有所提升，我们在边界检测任务和交错式 AI 辅助文本检测上均验证了其性能改进。最后，我们在标准 AI 检测基准上报告了相关指标，结果显示 Pangram 4 在多种设置和领域的 AI 文本检测任务中达到了当前最优性能。
英文摘要
We present Pangram 4, the latest deep-learning-based AI-text classification model from Pangram Labs. We achieve an AUROC of 0.9916 with a false positive rate of 0.0041% and a false negative rate of 0.3396%. In addition to its increased overall accuracy compared with Pangram 3, Pangram 4 exhibits superior out-of-distribution generalization and robustness to adversarial attacks. Another novel contribution of Pangram 4 is its improved ability to distinguish fine-grained edits and mixed AI-human co-authored text. We demonstrate improvements to both boundary detection tasks and the detection of interleaved AI assistance. Finally, we report metrics on standard AI detection benchmarks showing that Pangram 4 achieves state-of-the-art performance on the AI text detection task across a wide variety of settings and domains.
