Challenging Set
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/GT-SALT/Guided-Adversarial-Augmentation
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个具有挑战性的测试集，包含了1000个高质量的数据点。这些数据点是通过手动注释CoNLL 2003测试集生成的，并采用了专家指导的增强技术进行了转换。该数据集的创建旨在评估命名实体识别（NER）模型在使用对抗性示例时的泛化能力。规模上，该数据集包含了1000个数据点，针对的任务是命名实体识别。
This dataset is a challenging test set containing 1,000 high-quality data points. These data points are generated from the manually annotated CoNLL 2003 test set and transformed using expert-guided data augmentation techniques. The dataset is developed to evaluate the generalization capability of named entity recognition (NER) models when handling adversarial examples. With 1,000 data points in total, this dataset is tailored for the named entity recognition task.
