lihaoxin2020/ki-qwen-instruct-synthetic_1_qwen_math_only-sft-on-mmlu_pro-0shot_cot-scillm-914848b0f2

--- dataset_info: - config_name: biology features: - name: data list: - name: content dtype: string - name: role dtype: string - name: raw_response struct: - name: text dtype: string - name: raw_text dtype: string - name: knowledge_pieces list: string - name: doc_id dtype: int64 - name: native_id dtype: int64 splits: - name: train num_bytes: 3400919 num_examples: 200 download_size: 3299032 dataset_size: 3400919 - config_name: chemistry features: - name: data list: - name: content dtype: string - name: role dtype: string - name: raw_response struct: - name: text dtype: string - name: raw_text dtype: string - name: knowledge_pieces list: string - name: doc_id dtype: int64 - name: native_id dtype: int64 splits: - name: train num_bytes: 5274026 num_examples: 200 download_size: 5152922 dataset_size: 5274026 - config_name: computer science features: - name: data list: - name: content dtype: string - name: role dtype: string - name: raw_response struct: - name: text dtype: string - name: raw_text dtype: string - name: knowledge_pieces list: string - name: doc_id dtype: int64 - name: native_id dtype: int64 splits: - name: train num_bytes: 4285931 num_examples: 200 download_size: 4181815 dataset_size: 4285931 - config_name: engineering features: - name: data list: - name: content dtype: string - name: role dtype: string - name: raw_response struct: - name: text dtype: string - name: raw_text dtype: string - name: knowledge_pieces list: string - name: doc_id dtype: int64 - name: native_id dtype: int64 splits: - name: train num_bytes: 8334129 num_examples: 200 download_size: 8186903 dataset_size: 8334129 - config_name: health features: - name: data list: - name: content dtype: string - name: role dtype: string - name: raw_response struct: - name: text dtype: string - name: raw_text dtype: string - name: knowledge_pieces list: string - name: doc_id dtype: int64 - name: native_id dtype: int64 splits: - name: train num_bytes: 3768026 num_examples: 200 download_size: 3662559 dataset_size: 3768026 - config_name: math features: - name: data list: - name: content dtype: string - name: role dtype: string - name: raw_response struct: - name: text dtype: string - name: raw_text dtype: string - name: knowledge_pieces list: string - name: doc_id dtype: int64 - name: native_id dtype: int64 splits: - name: train num_bytes: 3559686 num_examples: 200 download_size: 3465625 dataset_size: 3559686 - config_name: physics features: - name: data list: - name: content dtype: string - name: role dtype: string - name: raw_response struct: - name: text dtype: string - name: raw_text dtype: string - name: knowledge_pieces list: string - name: doc_id dtype: int64 - name: native_id dtype: int64 splits: - name: train num_bytes: 5720930 num_examples: 200 download_size: 5603845 dataset_size: 5720930 configs: - config_name: biology data_files: - split: train path: biology/train-* - config_name: chemistry data_files: - split: train path: chemistry/train-* - config_name: computer science data_files: - split: train path: computer science/train-* - config_name: engineering data_files: - split: train path: engineering/train-* - config_name: health data_files: - split: train path: health/train-* - config_name: math data_files: - split: train path: math/train-* - config_name: physics data_files: - split: train path: physics/train-* ---数据集信息： 共包含7个领域配置，分别为生物学（biology）、化学（chemistry）、计算机科学（computer science）、工程学（engineering）、健康科学（health）、数学（math）以及物理学（physics）。 各配置的特征结构完全一致，具体如下： 1. `data`：列表（list）类型，包含两个子字段： - `content`：内容字段，数据类型（dtype）为字符串（string） - `role`：角色字段，数据类型（dtype）为字符串（string） 2. `raw_response`：结构体（struct）类型，包含子字段`text`，其数据类型（dtype）为字符串（string） 3. `raw_text`：原始文本字段，数据类型（dtype）为字符串（string） 4. `knowledge_pieces`：字符串列表类型 5. `doc_id`：文档编号字段，数据类型（dtype）为64位整型（int64） 6. `native_id`：本地编号字段，数据类型（dtype）为64位整型（int64） 各配置的划分集仅包含训练集（train），各训练集的详细信息如下： - 生物学配置：训练集字节数3400919，样本量200；下载大小3299032，数据集占用大小3400919 - 化学配置：训练集字节数5274026，样本量200；下载大小5152922，数据集占用大小5274026 - 计算机科学配置：训练集字节数4285931，样本量200；下载大小4181815，数据集占用大小4285931 - 工程学配置：训练集字节数8334129，样本量200；下载大小8186903，数据集占用大小8334129 - 健康科学配置：训练集字节数3768026，样本量200；下载大小3662559，数据集占用大小3768026 - 数学配置：训练集字节数3559686，样本量200；下载大小3465625，数据集占用大小3559686 - 物理学配置：训练集字节数5720930，样本量200；下载大小5603845，数据集占用大小5720930 配置项说明： 所有配置均对应训练集划分的数据文件，数据文件路径格式为`[配置名称]/train-*`，例如生物学配置的数据文件路径为`biology/train-*`，其余配置路径可依此类推。