WANDS
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/wayfair/wands
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为WANDS，包含了23.4万个标注好的查询与产品配对，为了涵盖所有独特的查询与产品组合，这一数量被扩展至2060万对。该数据集用于评估模型在语义搜索泛化能力方面的表现，并包含了基于查询和产品在训练数据中是否存在而设立的不同分析桶。其规模从23.4万对标注数据扩展至2060万对，任务专注于语义产品搜索。
This dataset, named WANDS, is scaled from 234,000 initial labeled query-product pairs to 20.6 million pairs, with the expansion aimed at covering all unique query-product combinations. It is developed to evaluate models' performance on semantic search generalization, and incorporates distinct analysis buckets defined based on whether queries and products exist in the training data. The core task of this dataset is semantic product search.
