deu05232/repro_msmarco-w-instructions_seed42-multipos-subset_add_version-seq

该数据集用于信息检索任务，包含查询和相关的文档段落。每个样本包括查询ID、查询文本、正面相关段落（含文档ID、解释、FollowIR评分、联合ID、文本和标题）、负面不相关段落（含文档ID、文本和标题）、仅指令文本、仅查询文本、是否包含指令的布尔标志，以及新负面段落（含文档ID、解释、FollowIR评分、联合ID、文本和标题）。数据集仅包含训练集，用于训练或评估检索模型，支持相关性判断和解释生成。

This dataset is designed for information retrieval tasks, containing queries and associated document passages. Each sample includes a query ID, query text, positive passages (with document ID, explanation, FollowIR score, joint ID, text, and title), negative passages (with document ID, text, and title), instruction-only text, query-only text, a boolean flag indicating whether instructions are present, and new negative passages (with document ID, explanation, FollowIR score, joint ID, text, and title). The dataset consists only of a training split and is intended for training or evaluating retrieval models, supporting relevance judgment and explanation generation.