RussianGLUE

RussianGLUE是一个针对俄语的高级通用语言理解评估基准，由Sberbank和国家研究大学高等经济学院等机构共同开发。该数据集包含九个任务，覆盖了自然语言推理、常识推理、机器阅读和世界知识等多个领域。数据集的创建过程涉及从零开始收集和组织数据，以及提供基线和人类水平评估。RussianGLUE旨在解决俄语环境下语言模型的广泛诊断和测试问题，推动俄语语言模型的进一步发展。

RussianGLUE is a general language understanding evaluation benchmark tailored for the Russian language, co-developed by institutions including Sberbank and the National Research University Higher School of Economics. The dataset comprises nine tasks spanning diverse domains such as natural language inference, commonsense reasoning, machine reading comprehension, and world knowledge. The development of RussianGLUE involves collecting and curating data from scratch, as well as providing baseline performance metrics and conducting human-level evaluations. RussianGLUE aims to enable comprehensive diagnostic and testing of language models in the Russian linguistic context, and advance the development of Russian language models.