chamber111/VPPO_ViRL39K_train
收藏Hugging Face2025-10-16 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/chamber111/VPPO_ViRL39K_train
下载链接
链接失效反馈官方服务：
资源简介：
这是一个包含39,000个多模态推理问题的数据集，覆盖数学、几何和逻辑等多个领域。该数据集旨在测试和提升大型视觉-语言模型的复杂推理能力。这是用于微调VPPO-7B和VPPO-32B模型的官方训练集，直接复制自TIGER-Lab/ViRL39K数据集。
This dataset contains 39,000 multimodal reasoning problems spanning various domains such as mathematics, geometry, and logic. It is designed to test and improve the complex reasoning capabilities of Large Vision-Language Models. This is the official training split used for fine-tuning the VPPO-7B and VPPO-32B models, directly copied from the TIGER-Lab/ViRL39K dataset.
提供机构：
chamber111
