mlfoundations-dev/instruction_filtering_scale_up_math_base_askllm_2K

该数据集包含了一系列字段，用于记录指令种子、数据来源、是否使用标记、分类器推理过程、原始数据行索引、推理过程、DeepSeek解决方案、最终推理轨迹以及对话信息。对话信息包括对话的发送者和内容。数据集分为训练集，提供了详细的字段类型和大小信息。

The dataset includes a series of fields for recording instruction seeds, data sources, usage flags, classifier reasoning processes, original data row indices, reasoning processes, DeepSeek solutions, final reasoning traces, and conversation information. Conversation information includes the sender and content of the conversation. The dataset is split into a training set and provides detailed field types and size information.