deu05232/promptriever-cross_batch_version

该数据集是一个用于信息检索任务的数据集，包含查询和对应的文档数据。每个样本包括查询ID、查询文本、正面文档列表和负面文档列表。正面文档具有详细属性如文档ID、解释、FollowIR分数、联合ID、文本和标题；负面文档则包含文档ID、文本和标题。此外，数据集还提供仅指令、仅查询字段，以及一个布尔值表示是否包含指令。新负面文档列表与正面文档结构类似。数据集用于训练和评估检索模型，支持基于查询和文档的相关性学习。This dataset is intended for information retrieval tasks, consisting of queries and their corresponding document data. Each sample contains a query ID, query text, a list of positive documents and a list of negative documents. Positive documents come with detailed attributes including document ID, explanation, FollowIR score, joint ID, text and title; whereas negative documents only include document ID, text and title. Furthermore, the dataset provides instruction-only and query-only fields, alongside a boolean value indicating whether instructions are contained within the sample. The new negative document list shares a similar structure with the positive document list. This dataset is utilized for training and evaluating retrieval models, enabling relevance learning based on queries and documents.