copenlu/cub-nq
收藏Hugging Face2025-09-23 更新2025-07-05 收录
下载链接：
https://hf-mirror.com/datasets/copenlu/cub-nq
下载链接
链接失效反馈官方服务：
资源简介：
NQ数据集是一个经过筛选和处理的自然问答数据集，它基于原始NQ数据集，包含了能够从原始维基页面中恢复金标段落且短答案长度小于5个单词的样本。数据集分为gpt2-xl和pythia-6.9b两个版本，每个版本都有验证集和测试集，用于自然问答和文本生成任务。
The NQ dataset is a filtered and processed version of the original NQ natural question answering dataset. It includes samples that can recover the gold passage from the original Wikipedia page with short answers of less than 5 words in length. The dataset is divided into two versions, gpt2-xl and pythia-6.9b, each with a validation set and a test set, used for question answering and text generation tasks.
提供机构：
copenlu
