trl-lib/rlaif-v
收藏Hugging Face2025-01-08 更新2025-04-12 收录
下载链接:
https://hf-mirror.com/datasets/trl-lib/rlaif-v
下载链接
链接失效反馈官方服务:
资源简介:
RLAIF-V数据集是一个专门为偏好学习任务训练视觉语言模型而设计的数据集。它包含了83132个高质量的图片和文本描述的比较对,每个比较对中包含一个首选文本描述和一个被拒绝的文本描述。这些比较对使得模型能够学习在视觉上下文中的人类偏好,从而提高它们生成和评估图像标题的能力。
The RLAIF-V dataset is designed for training vision-language models for preference learning tasks. It includes 83,132 high-quality comparison pairs consisting of an image and two textual descriptions: one preferred and one rejected. These comparison pairs enable models to learn human preferences in visual contexts, enhancing their ability to generate and evaluate image captions.
提供机构:
trl-lib



