OUTFOX

该数据集名为OUTFOX，旨在评估IPAD检测方法在面对不同大型语言模型（LLMs）时的鲁棒性。它包含了一系列文本，用于评估检测效果的有效性。此外，该数据集还包含了在攻击情况下以及无攻击情况下检测方法的性能评估，为衡量检测模型的泛化能力和鲁棒性提供了基准。在实验中，该数据集包含了1,000个样本，任务是对人类编写文本与大型语言模型生成文本进行检测对比。

This dataset, named OUTFOX, is developed to evaluate the robustness of the IPAD detection method against diverse large language models (LLMs). It includes a set of texts for assessing the effectiveness of detection performance. Furthermore, the dataset incorporates performance evaluations of detection methods under both adversarial attack and attack-free scenarios, serving as a benchmark for gauging the generalization capability and robustness of detection models. Comprising 1,000 samples, the dataset is used for the task of detecting and comparing human-written texts and texts generated by large language models in experimental trials.