permitt/superglue-sr

BalkanBench SuperGLUE - Serbian是BalkanBench项目的一部分，旨在为塞尔维亚语（BCMS语言之一）提供语言模型评估基准。该数据集包含7个任务配置，涵盖了文本分类、多项选择、问答等多种任务类型，是通过将原始SuperGLUE英语数据集翻译成塞尔维亚语并经过人工验证而创建的。数据集分为训练集、验证集和测试集，其中测试集仅包含在私有仓库中以保持排行榜的完整性。总规模为65,853项，分布在不同的任务和分割中。数据集的使用需要引用相关论文和发布说明，并采用CC-BY-4.0许可证。

BalkanBench SuperGLUE - Serbian is part of the BalkanBench project, designed to provide an evaluation benchmark for language models in Serbian (one of the BCMS languages). The dataset includes 7 task configurations covering various task types such as text classification, multiple-choice, and question-answering. It was created by translating the original SuperGLUE English dataset into Serbian and undergoing human verification. The dataset is divided into training, validation, and test sets, with the test set only available in a private repository to maintain leaderboard integrity. The total size is 65,853 items distributed across different tasks and splits. Use of the dataset requires citation of relevant papers and release notes, and it is licensed under CC-BY-4.0.