ryota39/dpo-ja-45k
收藏Hugging Face2024-06-15 更新2024-06-12 收录
下载链接:
https://hf-mirror.com/datasets/ryota39/dpo-ja-45k
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含四个子数据集,分别是llm-jp/hh-rlhf-12k-ja、cyberagent/chatbot-arena-ja-calm2-7b-chat-experimental、ryota39/boolq-3k-ja和ryota39/truthy-dpo-ja。这些数据集主要用于自然语言处理任务,如对话生成和问答系统。数据生成方法包括使用Anthropic社的Claude-52B模型生成用户输入和响应,使用chatbot-arena的用户输入进行翻译和响应生成,以及使用Google搜索日志和Wikipedia文章进行问答生成。数据格式为JSON,包含prompt、chosen、rejected和dataset_name字段。
This dataset includes four sub-datasets: llm-jp/hh-rlhf-12k-ja, cyberagent/chatbot-arena-ja-calm2-7b-chat-experimental, ryota39/boolq-3k-ja, and ryota39/truthy-dpo-ja. These datasets are primarily used for natural language processing tasks such as dialogue generation and question-answering systems. The data generation methods involve using Anthropics Claude-52B model to generate user inputs and responses, translating and generating responses using chatbot-arenas user inputs, and generating Q&A using Google search logs and Wikipedia articles. The data format is JSON, containing fields for prompt, chosen, rejected, and dataset_name.
提供机构:
ryota39
原始信息汇总
数据集许可证信息
- 许可证类型:Creative Commons Attribution-ShareAlike 3.0 (CC-BY-SA-3.0)



