mlfoundations-dev/instruction_filtering_gemini_length_math

该数据集包含了指令种子（instruction_seed）、来源（source）、Gemini模型响应（gemini_response）、原始行索引（__original_row_idx）、长度（length）、r1_distill_70b模型响应（r1_distill_70b_response）和对话（conversations）等字段。对话字段是一个列表，包含了对话的发起者和对话内容。数据集仅包含一个训练集split，共有2500个样本，总大小为约118MB。

The dataset includes fields such as instruction_seed, source, gemini_response, __original_row_idx, length, r1_distill_70b_response, and conversations. The conversations field is a list containing the initiator and content of the conversation. The dataset contains only one training set split with a total of 2500 samples and a total size of approximately 118MB.