mteb/MAUDLegalBenchClassification

MAUDLegalBenchClassification数据集是由专家注释的，包含152个并购协议中的47,000多个标签，用于识别每个协议中的92个问题。数据集被格式化为一系列多项选择题，模型需要根据并购协议的片段和一个交易点问题选择最能表征协议的回答。数据集是单语言的，仅包含英语。它包含训练集和测试集，其中训练集有941个样本，测试集有2048个样本。该数据集是MTEB（大规模文本嵌入基准）的一部分，用于评估嵌入模型。

The MAUDLegalBenchClassification dataset is expert-annotated, containing over 47,000 labels across 152 merger agreements to identify 92 questions in each agreement. It is formatted as a series of multiple-choice questions where the model must choose the best answer to characterize the agreement given a segment of the merger agreement and a Deal Point question. The dataset is monolingual, consisting only of English. It includes a training set with 941 samples and a test set with 2048 samples. The dataset is part of the MTEB (Massive Text Embedding Benchmark) for evaluating embedding models.