sapienzanlp/hellaswag_italian

HellaSwag数据集的意大利语翻译版本是一个大规模的常识推理数据集，要求通过阅读理解和常识推理来预测句子的正确结尾。数据集包含上下文和四个可能答案的多项选择题，任务是从中选择正确的句子结尾。数据集分为训练集和验证集，并且提供了全数据集、WikiHow领域和ActivityNet领域的独立分割。数据集是英语和意大利语完全并行的，翻译过程使用了开源的LLM工具OBenTO-LLM。数据集格式包括唯一ID、任务类型、原始英语句子、意大利语翻译、原始英语选项、意大利语选项翻译以及正确答案的索引。

This dataset is an Italian translation of the original HellaSwag dataset, which is a large-scale commonsense reasoning dataset. It requires reading comprehension and commonsense reasoning to predict the correct ending of a sentence. The dataset includes instances with a context and a multiple-choice question with four possible answers. It is fully parallel between English and Italian, allowing for comparable evaluation setups and results across the two languages. The dataset is split into three configurations: All, WikiHow, and ActivityNet, each with different domains and number of instances. The translation process was carried out using an open-source tool called OBenTO-LLM, which encourages free, open, reproducible, and transparent research in LLM evaluation.