maartenbuyl/alignment-discretion

该数据集与AI对齐原则相关，包含两个配置的数据文件：hh-rlhf和PKU-SafeRLHF。数据集基于一系列预定义的原则，这些原则旨在确保AI行为的安全性和伦理性。这些原则涵盖了从避免有害语言到尊重人权和平等等各种伦理和社会关注点。数据集中的偏好映射到这些原则上，用于评估AI响应的合适性。

This dataset is related to AI Alignment principles, containing data files for two configurations: hh-rlhf and PKU-SafeRLHF. The dataset is based on a set of predefined principles aimed at ensuring the safety and ethics of AI behavior. These principles cover a range of ethical and societal concerns from avoiding harmful language to respecting human rights and equality. The preferences in the dataset are mapped to these principles to evaluate the appropriateness of AI responses.