mlfoundations-dev/instruction_filtering_scale_up_math_base_embedding_filter_mean_16K

该数据集包含多个字段，其中有字符串类型的指令种子(instruction_seed)、来源(source)、推理(reasoning)等，浮点数类型的均值分数(mean_positive_score、mean_negative_score)和差异分数(difference_score)，以及固定长度的浮点数序列(embeddings)。此外，还包括一个对话列表(conversations)，每个对话包含发送者(from)和消息(value)。数据集分为训练集，包含大约16000个示例，文件大小约为1.2GB。

The dataset includes multiple fields such as instruction_seed, source, reasoning of string type, mean scores (mean_positive_score, mean_negative_score) and difference score (difference_score) of float type, as well as a fixed-length float sequence (embeddings). It also contains a list of conversations, each with a sender (from) and a message (value). The dataset is split into a training set with approximately 16,000 examples, and the file size is about 1.2GB.