MMMLU
收藏资源简介:
多语言大规模多任务语言理解(MMMLU)数据集是一个广泛认可的AI模型通用知识基准,涵盖了从基础知识到高级专业学科的57个不同类别。该数据集包括了MMLU测试集的14种语言的专业翻译版本,这些翻译由专业的人类翻译完成,以确保翻译的准确性,特别是对于低资源语言如Yoruba。数据集的发布反映了提高AI模型多语言能力的承诺,确保其在不同语言中的准确表现,特别是对于未被充分代表的社区。
The Multilingual Massive Multitask Language Understanding (MMMLU) dataset is a widely recognized general knowledge benchmark for AI models, covering 57 distinct categories ranging from foundational knowledge to advanced specialized disciplines. This dataset includes professionally translated versions of the MMLU test set across 14 languages, with translations conducted by professional human translators to guarantee translation accuracy, especially for low-resource languages such as Yoruba. The release of this dataset reflects the commitment to improving the multilingual capabilities of AI models and ensuring their accurate performance across different languages, particularly for underrepresented communities.




