McGill-NLP/GlobalNLI
收藏Hugging Face2025-07-26 更新2025-08-09 收录
下载链接:
https://hf-mirror.com/datasets/McGill-NLP/GlobalNLI
下载链接
链接失效反馈官方服务:
资源简介:
Global NLI是一个新的基于文本的基准,它基于现有的自然语言推理数据集的聚合。这个数据集支持59种语言,并且为每种语言提供了验证集和测试集。每个数据点包括前提、假设和标签三个字段。标签字段是一个分类标签,包括蕴含(0)、中性(1)和矛盾(2)三种可能的值。
Global NLI is a new text-based benchmark based on the aggregation of existing NLI datasets that are publicly available. It supports 59 languages and provides validation and test sets for each language. Each data point includes three fields: premise, hypothesis, and label. The label field is a classification label with possible values including entailment (0), neutral (1), and contradiction (2).
提供机构:
McGill-NLP



