LLMTeamAkiyama/cleand_moremilk_CoT_Reasoning_Scientific_Discovery_and_Research

这是一个包含科学发现和研究推理任务的数据集，共有3733条数据，平均每个条目包含1193个トークン，最多包含2489个トークン，总トーク数达到4,453,517个。数据集以JSONL格式存储，经过了一系列的数据清洗和加工过程，包括解析元数据列、筛选难度、去除重复表达、计算トーク数以及按照特定条件进行最终筛选。数据集适用于问答和推理任务。

This dataset contains scientific discovery and research reasoning tasks, with a total of 3,733 entries, an average of 1,193 tokens per entry, a maximum of 2,489 tokens, and a total of 4,453,517 tokens. The dataset is stored in JSONL format and has undergone a series of data cleaning and processing steps, including parsing metadata columns, filtering by difficulty, removing duplicate expressions, calculating token counts, and conducting final filtering based on specific conditions. The dataset is suitable for question-answering and reasoning tasks.