open-paws/animal-alignment-feedback

这个数据集包含合成的和真实的人类/非人类生成的反馈，专注于动物对齐问题，即个体（人类）和通过大型语言模型（LLMs）模拟的人类/非人类视角如何评价与动物福利和权利相关的内容的伦理考量。数据集旨在探索与动物福利问题相关的各种人类和非人类价值观、情感影响和认知特征。超过95%的数据是由AI模型生成的合成反馈，以模拟关于动物对齐和相关话题的多样化和非人类视角。不到5%的数据来自真实的人类志愿者，而大部分是模拟人类和非人类视角的合成反馈。

This dataset contains a mixture of synthetic and real human/nonhuman-generated feedback, specifically focused on animal alignment — that is, how individuals (humans) and human/nonhuman perspectives simulated through Large Language Models (LLMs) evaluate the ethical considerations of content in relation to animal welfare and rights. The dataset is designed to explore various human and nonhuman values, emotional impacts, and cognitive traits in the context of animal welfare issues. Over 95% of the data consists of synthetic feedback generated through AI models to simulate diverse human and nonhuman perspectives on animal alignment and related topics, while less than 5% comes from real human volunteers.