internlm/CapRL-2M
收藏Hugging Face2026-04-16 更新2025-10-18 收录
下载链接：
https://hf-mirror.com/datasets/internlm/CapRL-2M
下载链接
链接失效反馈官方服务：
资源简介：
CapRL-2M数据集包含来自ShareGPT-1M和DenseFusion-1M的两个图像集合，使用CapRL-3B模型重新注释了高质量的字幕，总计200万个样本。该数据集适用于图像字幕任务，特别是CapRL-3B模型能够生成具有良好组织结构、详细描述自然图像的字幕。
The CapRL-2M dataset includes two sets of images from ShareGPT-1M and DenseFusion-1M, with high-quality captions re-annotated using the CapRL-3B model, totaling 2M samples. This dataset is suitable for image captioning tasks, especially as the CapRL-3B model is capable of generating captions with good organizational structure and detailed descriptions of natural images.
提供机构：
internlm
