dyyyyyyyy/FAPO-Reasoning-Dataset
收藏Hugging Face2025-10-28 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/dyyyyyyyy/FAPO-Reasoning-Dataset
下载链接
链接失效反馈官方服务：
资源简介：
FAPO数据集用于训练FAPO-32B模型，包含train.parquet和test.parquet两个文件。train.parquet基于DAPO-Math-17K数据集，使用特定的指令格式，并复制了20次。test.parquet是由AIME24、AIME25和GPQA-Diamond三个数据集混合而成，分别复制了32次、32次和4次。
The FAPO dataset is used for training the FAPO-32B model, consisting of two files: train.parquet and test.parquet. The train.parquet is based on the DAPO-Math-17K dataset, using a specific instruction format and duplicated 20 times. The test.parquet is a mixture of AIME24, AIME25, and GPQA-Diamond datasets, duplicated 32 times, 32 times, and 4 times respectively.
提供机构：
dyyyyyyyy
