mlfoundations-dev/science_verification_scale_up_all

这个数据集包含了问题、推理过程、解决方案等信息的记录。具体字段包括问题描述（problem）、推理（reasoning）、解决方案（deepseek_solution）、原始行索引（__original_row_idx）、最终推理轨迹（final_reasoning_trace）以及对话信息（conversations）。对话信息包含对话来源（from）和对话内容（value）。数据集分为训练集（train），共有48000个示例。

The dataset includes records of problems, reasoning processes, solutions, etc. Specific fields include problem description (problem), reasoning (reasoning), solution (deepseek_solution), original row index (__original_row_idx), final reasoning trace (final_reasoning_trace), and conversation information (conversations). Conversation information includes the source of the conversation (from) and the content of the conversation (value). The dataset is split into a training set (train) with a total of 48000 examples.