LangAGI-Lab/llama-8b-instruct-7k-dpo-preference-set
收藏Hugging Face2025-02-11 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/LangAGI-Lab/llama-8b-instruct-7k-dpo-preference-set
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了指令（instruction）、答案（answer）、选中的响应（chosen）以及被拒绝的响应（rejected）等信息。选中的响应和被拒绝的响应都包含了是否正确（is_correct）、响应内容（response）和token长度（token_length）。此外，数据集还包含了预思考token的数量（o1-mini_thought_tokens）和数据来源（source）。数据集分为训练集，其大小为27056412字节，共有7152个示例。
The dataset includes instruction, answer, chosen response, and rejected response information. Both chosen and rejected responses contain correctness (is_correct), response content (response), and token length (token_length). Additionally, the dataset has the number of pre-thought tokens (o1-mini_thought_tokens) and the source of the data. The dataset is split into a training set, which is 27056412 bytes in size and contains 7152 examples.
提供机构：
LangAGI-Lab
