envyr/natural-questions-hard-negatives_gemma_MarginMSE_instructions_STATISTICAL
收藏数据链接：
官方服务：
资源简介：
该数据集是一个用于训练任务的数据集，包含三元组文本数据，每个样本由anchor（锚点文本）、positive（正例文本）和negative（负例文本）组成，并附带一个浮点型标签（label）。数据集仅包含训练分割，共有484,310个样本，总大小约为678.7 MB，适用于对比学习或相似性学习等NLP任务。
This dataset is designed for training tasks and consists of triplet text data, where each sample includes an anchor text, a positive text, and a negative text, along with a float label. It contains only a training split with 484,310 examples and a total size of approximately 678.7 MB, suitable for NLP tasks such as contrastive learning or similarity learning.
提供机构：
envyr
