aracape/cai-education-single-turn

该数据集提供了旨在帮助助教在教学方面提高的偏好对。这些偏好对是使用Llama-3.1-8B-Instruct模型以宪法式方法生成的，并包含了最终的偏好对以及批评-修订历史。宪法原则旨在鼓励批判性思维和苏格拉底式提问。修订的目的是在对话过程中给出提示和指导，而不是直接给出完整的答案，以防止作弊。This dataset provides preference pairs aimed at helping assistants improve at teaching. Generated using a constitutional approach with the Llama-3.1-8B-Instruct model, the dataset includes the final preference pairs along with the critique-revision history. The constitutional principles are designed to encourage critical thinking and Socratic questioning. The revisions aim to provide hints and guidance in the course of a dialogue, rather than giving away full answers that could facilitate cheating.