mlfoundations-dev/instruction_filtering_scale_up_math_base_gemini_length_4K

该数据集包含多个文本字段，如指令种子(instruction_seed)、来源(source)、Gemini模型的响应(gemini_response)等，以及一些数值字段，如原始行索引(__original_row_idx)和长度(length)。还包括推理(reasoning)、DeepSeek解决方案(deepseek_solution)和最终的推理轨迹(final_reasoning_trace)等字段。数据集分为训练集，包含4000个示例，总字节数为290,956,370字节。数据集的下载大小为121,103,252字节，整体大小为290,956,370字节。具体用途和背景未在README中说明。

The dataset includes multiple text fields such as instruction_seed, source, gemini_response, etc., as well as numerical fields like __original_row_idx and length. It also contains fields for reasoning, DeepSeek solution, and final reasoning trace. The dataset is split into a training set with 4000 examples, totaling 290,956,370 bytes. The download size of the dataset is 121,103,252 bytes, and the overall size is 290,956,370 bytes. The specific purpose and background are not described in the README.