recogna-nlp/dpo-mmlu-ptbr

这是一个葡萄牙语医学遗传学问题回答数据集，它是从Allen Institute for AI的Natural Instructions项目中改编而来的。数据集经过了完全翻译，并添加了新的样本以增加多样性。此外，数据集已被转换为DPO格式，适合进行微调和偏好学习。还进行了术语审查以确保医学词汇的准确性。数据集可以用于训练模型以回答医学遗传学领域的问题，也可以用于通过DPO进行微调以优化正确答案，或作为葡萄牙语医学问答的基准测试。

This is a Portuguese medical genetics question answering dataset adapted from the Allen Institute for AIs Natural Instructions project. The dataset has been fully translated and expanded with new samples for greater diversity. It has also been converted to the DPO format, suitable for fine-tuning and preference learning. A terminological review has been conducted to ensure the accuracy of medical vocabulary. The dataset can be used to train models to answer questions in the field of medical genetics, fine-tune through DPO to optimize correct answers, or serve as a benchmark for Portuguese medical QA.