mlfoundations-dev/instruction_filtering_gemini_length_code_cheat

该数据集包含多个字符串类型的特征字段，如指令种子(instruction_seed)、来源(source)、Gemini模型响应(gemini_response)等，以及整数类型的原始行索引(__original_row_idx)和长度(length)。此外，还有一个包含对话信息的列表字段，每个对话条目包含发送者(from)和消息内容(value)。数据集目前只有一个训练集(train)，包含2500个示例，总文件大小为87076136字节。具体用途和详细描述未在README中提供。

The dataset includes several string-type feature fields such as instruction_seed, source, gemini_response, etc., as well as integer-type fields for the original row index (__original_row_idx) and length. Additionally, there is a list field containing conversation information, with each conversation entry including a sender (from) and message content (value). The dataset currently has only one training set (train) with 2500 examples, totaling 87076136 bytes in file size. The specific purpose and detailed description are not provided in the README.