August4293/Self_Alignment_Preference-Dataset
收藏Hugging Face2024-03-18 更新2024-07-06 收录
下载链接:
https://hf-mirror.com/datasets/August4293/Self_Alignment_Preference-Dataset
下载链接
链接失效反馈官方服务:
资源简介:
Mistral自我对齐偏好数据集是由Mistral 7b模型生成的,使用了Anthropics Red Teaming Prompts数据集作为源数据。该数据集的目的是为了促进自我对齐,主要用于对齐和评估任务。数据集包含提示(prompt)、被拒绝的回答(rejected)和选择的回答(chosen)三个特征,并且只有一个训练集,包含4449个样本。
The Mistral 7b Preference Dataset is a dataset generated by the Mistral 7b model for text generation and question-answering tasks. It includes three main features: prompt, rejected, and chosen, divided into a training set with 4449 samples. The purpose of the dataset is to facilitate self-alignment, generated using the Mistral 7b model from the Anthropics Red Teaming Prompts dataset. The dataset is primarily in English and falls within the size category of 1K to 10K. It is important to note that this dataset contains harmful and offensive content.
提供机构:
August4293
原始信息汇总
Mistral Self-Alignment Preference Dataset
数据集概述
- 名称: Mistral Self-Alignment Preference Dataset
- 大小类别: 1K < n < 10K
- 语言: 英语(en)
- 任务类别:
- 文本生成
- 问答
数据集详情
- 来源: Anthropics Red Teaming Prompts Dataset
- 生成者: Mistral 7b
- 目的: 自我对齐
- 用途: 对齐和评估
数据集结构
- 特征:
prompt: 字符串类型rejected: 字符串类型chosen: 字符串类型
- 分割:
train:- 样本数量: 4449
- 字节数: 8024886
- 下载大小: 4200748 字节
- 数据集大小: 8024886 字节
配置
- 配置名称: default
- 数据文件:
train: data/train-*



