mlfoundations-dev/instruction_filtering_scale_up_math_base_gemini_length_2K

该数据集包含了指令种子、来源、Gemini响应等文本信息，以及一个包含对话发起者和内容的对话列表。数据集被划分为训练集，其中包含2000个示例。此外，数据集还包含了推理过程、DeepSeek解决方案和最终推理轨迹等字段，用于记录和分析对话中的推理过程。

The dataset includes text information such as instruction seed, source, Gemini response, and a conversation list containing the initiator and content of the conversation. The dataset is split into a training set, which contains 2000 examples. Additionally, the dataset features fields for reasoning process, DeepSeek solution, and final reasoning trace, used for recording and analyzing the reasoning process in conversations.