qihoo360/Light-R1-SFTData
收藏Hugging Face2025-03-17 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/qihoo360/Light-R1-SFTData
下载链接
链接失效反馈官方服务：
资源简介：
本文未直接给出数据集的中文描述，但根据内容可以推断，该数据集包含了用于训练Light-R1-32B模型的两个阶段的数据，分别是76k的训练数据和3k的更困难的数据集。这些数据来源于公开的数学数据集，并通过课程学习策略和DPO方法进行了处理。
The README does not provide a direct description of the dataset in English. However, it can be inferred from the content that the dataset consists of two stages of data used to train the Light-R1-32B model, which are 76k training data and 3k more challenging dataset. These data are sourced from public math datasets and have been processed using a curriculum learning strategy and DPO method.
提供机构：
qihoo360
