mlfoundations-dev/instruction_filtering_embedding_filter_code_mean

该数据集包含了一系列的特征字段，包括文本指令种子、数据来源、浮点数类型的嵌入向量、正面和负面评分的平均值、评分差值、以及文本类型的模型响应。此外，每个样本还包含了原始行索引和会话信息，会话信息中又包括了消息发送者和消息内容。数据集划分为训练集，可用于模型训练等任务。

The dataset includes a set of feature fields such as text instruction seeds, data sources, float-type embedding vectors, average positive and negative scores, score differences, and text-type model responses. Each sample also contains the original row index and conversation information, which includes the message sender and message content. The dataset is split into a training set, which can be used for tasks like model training.