Mechinterp-dataset-2digits
收藏资源简介:
该数据集是一个包含训练集和验证集的结构化文本数据集,每个分片各有15000个样本。数据集包含六个字符串类型的字段:唯一标识符(_id)、语言标识(language)、英文问题(eng_question)、英文答案(eng_answer)、模型响应(response)以及提示词(prompt)。数据规模方面,训练集约1.09MB,验证集约1.06MB,总数据集大小约2.15MB。从字段命名推断,数据可能涉及多语言或跨语言场景下的问答或指令遵循任务,包含原始问题、参考答案以及语言模型生成的响应,适用于自然语言处理中的模型训练、评估或分析任务。
This is a structured text dataset comprising a training split and a validation split, with 15,000 samples per split. The dataset includes six string-type fields: unique identifier (_id), language indicator (language), English question (eng_question), English reference answer (eng_answer), model response (response), and prompt. Regarding data size, the training split is approximately 1.09 MB, the validation split is around 1.06 MB, and the overall dataset size totals roughly 2.15 MB. Based on the field naming conventions, this dataset is likely intended for question answering or instruction following tasks in multilingual or cross-lingual scenarios. It contains original questions, reference answers, and responses generated by large language models, and is applicable to model training, evaluation, and analysis tasks in the field of natural language processing.
- 数据集名称:Mechinterp-dataset-2digits
- 数据集地址:https://huggingface.co/datasets/Manoj-kanur/Mechinterp-dataset-2digits
- 数据集规模:训练集15,000条样本,验证集15,000条样本,总共30,000条样本,下载大小约690.7 KB,数据集总大小约2.1 MB。
- 数据集特征:包含6个字段,分别是
_id(字符串)、language(字符串)、eng_question(字符串)、eng_answer(字符串)、response(字符串)、prompt(字符串)。 - 数据划分:分为训练集(train)和验证集(validation)两个部分,训练集数据文件路径为
data/train-*,验证集数据文件路径为data/validation-*。 - 配置信息:提供默认配置(config_name为default),使用该配置即可加载数据。




