efederici/mc-translation

该数据集包含来自OpenAI的MMMLU数据集的专业人工翻译，用于训练翻译模型，以帮助翻译未来的评估数据集。翻译评估基准是一个关键但具有挑战性的任务。虽然自动翻译可能会引入错误或偏见，但专业人工翻译既昂贵又耗时。该数据集利用现有的专业翻译（MMMLU）来训练专门的翻译模型，以协助翻译未来的评估集。

This dataset contains professional human translations from OpenAIs MMMLU dataset, repurposed to train translation models that can help translate future evaluation datasets. The dataset supports multiple languages including English, Swahili, Spanish, German, Chinese, Bengali, Italian, Hindi, Japanese, Korean, Portuguese, Arabic, and Indonesian. The dataset features include prompt, output, input language, output language, and conversations. The dataset is divided into a training set containing 209,115 samples. The goal of the dataset is to address the challenges of translating evaluation benchmarks, leveraging professional human translations to train specialized translation models to assist in future translation tasks.