medix-rl-data
收藏资源简介：
MediX-R1是一个开放式的医学强化学习数据集，由Mohamed Bin Zayed人工智能大学（MBZUAI）开发。该数据集专为图像文本到文本的任务设计，包含医学和生物学领域的英文内容。数据集包含51,335个训练样本和2,451个测试样本，每个样本包括id、图像（RGBA模式）、来源、问题和解决方案等字段。数据集总大小为24.1GB，适用于医学图像理解和文本生成等研究任务。数据集采用CC-BY-NC-SA 4.0许可证，仅限研究使用，明确禁止临床或商业应用。使用时应谨慎，建议在现实医疗场景中咨询专业医疗人员验证模型输出。
MediX-R1 is an open medical reinforcement learning dataset developed by Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI). It is specifically designed for image-text-to-text tasks and contains English content from the medical and biological fields. The dataset comprises 51,335 training samples and 2,451 test samples, with each sample including fields such as id, image (RGBA mode), source, question, and solution. The total size of the dataset is 24.1 GB, and it is suitable for research tasks including medical image understanding and text generation. The dataset is licensed under CC-BY-NC-SA 4.0, for research use only, and explicitly prohibits clinical or commercial applications. Caution should be exercised when using the dataset, and it is recommended to consult professional medical personnel to verify model outputs in real medical scenarios.
数据集概述
基本信息
- 数据集名称: MediX-R1: Open-Ended Medical Reinforcement Learning
- 发布机构: Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI), UAE
- 许可证: CC-BY-NC-SA 4.0 (仅限研究用途)
- 主要语言: 英语 (en)
- 任务类别: 图像-文本到文本 (image-text-to-text)
- 领域标签: 医学 (medical)、生物学 (biology)
- 规模分类: 10K < n < 100K
数据集结构与内容
数据特征
id: 字符串类型，标识符。image: 图像列表，格式为RGBA模式。source: 字符串类型，数据来源。problem: 字符串类型，问题描述。solution: 字符串类型，解决方案。
数据划分
- 训练集 (train):
- 样本数量: 51,335
- 数据大小: 22,901,597,169 字节
- 测试集 (test):
- 样本数量: 2,451
- 数据大小: 1,226,261,011 字节
存储信息
- 下载大小: 47,058,109,860 字节
- 数据集大小: 24,127,858,180 字节
相关资源
模型
- MediX-R1-2B: https://huggingface.co/MBZUAI/MediX-R1-2B
- MediX-R1-8B: https://huggingface.co/MBZUAI/MediX-R1-8B
- MediX-R1-30B: https://huggingface.co/MBZUAI/MediX-R1-30B
外部链接
- 项目网站: https://medix.cvmbzuai.com
- 论文: https://arxiv.org/pdf/2602.23363
- HuggingFace 集合: https://huggingface.co/collections/MBZUAI/medix-r1
- 排行榜: https://medix.cvmbzuai.com/leaderboard
引用信息
如果使用本数据集，请引用以下文献： bibtex @misc{mullappilly2026medixr1openendedmedical, title={MediX-R1: Open Ended Medical Reinforcement Learning}, author={Sahal Shaji Mullappilly and Mohammed Irfan Kurpath and Omair Mohamed and Mohamed Zidan and Fahad Khan and Salman Khan and Rao Anwer and Hisham Cholakkal}, year={2026}, eprint={2602.23363}, archivePrefix={arXiv}, primaryClass={cs.CV}, url={https://arxiv.org/abs/2602.23363}, }
重要声明
- 本数据集仅限研究用途，不适用于临床或商业用途。
- 在现实医疗场景中应用模型输出时，用户必须负责任地使用，并务必咨询合格的医疗专业人员以核实建议，不得依赖其进行医疗诊断或治疗决策。
