mlfoundations-dev/instruction_filtering_scale_up_math_base_embedding_filter_mean_per_domain_8K

该数据集包含多个字段，如指令种子(instruction_seed)、来源(source)、嵌入(embeddings)等，其中包含了浮点数类型的评分字段，以及文本类型的推理字段。数据集分为训练集(train)等部分，提供了每个部分的字节数和示例数量。具体应用场景和详细描述未在README中提供。

The dataset includes multiple fields such as instruction_seed, source, embeddings, etc., which contain floating-point score fields and text-type reasoning fields. The dataset is split into parts such as the training set (train), with the number of bytes and examples for each part provided. The specific application scenario and detailed description are not provided in the README.