Eagle

Eagle数据集是由MBZUAI创建的，旨在从实际的ChatGPT用户交互中提取包含社会偏见、观点偏见、有毒语言和道德问题的实例。该数据集包含2317个实例，每个实例都标记了相应的伦理问题类别。Eagle数据集的创建过程涉及从真实世界对话日志中提取数据，并通过GPT-3.5和GPT-4进行自动分类。该数据集的应用领域主要集中在评估和改进大型语言模型在处理实际用户交互中的伦理挑战，确保模型的安全性和伦理性。

The Eagle Dataset was developed by MBZUAI to extract instances involving social biases, viewpoint biases, toxic language, and ethical issues from real-world ChatGPT user interactions. This dataset contains 2,317 instances, each labeled with its corresponding ethical problem category. The creation process of the Eagle Dataset includes extracting data from real-world conversation logs and performing automatic classification using GPT-3.5 and GPT-4. The primary application areas of this dataset focus on evaluating and improving the ability of large language models to address ethical challenges in real-world user interactions, as well as ensuring the safety and ethical soundness of these models.