mlfoundations-dev/seed_math_tiger_math

该数据集包含对话信息，每个样本包括指令种子(instruction_seed)、响应种子(response_seed)、实际响应(response)等字段，以及原始行索引(__original_row_idx)。此外，还有一个包含对话发送者(from)和对话内容(value)的列表。数据集目前只有一个训练集(train)，共有50000个样本。

The dataset contains conversational information, with each sample including fields such as instruction seed (instruction_seed), response seed (response_seed), actual response (response), and original row index (__original_row_idx). Additionally, there is a list containing the conversation sender (from) and the content of the conversation (value). The dataset currently has only one training set (train) with a total of 50,000 samples.