PrimeIntellect/SYNTHETIC-1-SFT-Data
收藏Hugging Face2025-02-21 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/PrimeIntellect/SYNTHETIC-1-SFT-Data
下载链接
链接失效反馈官方服务：
资源简介：
SYNTHETIC-1是一个从Deepseek-R1获得的推理数据集，通过众包计算生成并使用各种验证器（如LLM评委或符号数学验证器）进行注释。该数据集包含数学问题、算法编码问题、现实世界软件工程问题、开放性STEM问题回答和合成代码理解任务等多种类型的问题。具体包括777k个数学问题样本、144k个算法编码问题样本、70k个现实世界软件工程问题样本、313k个开放性STEM问题回答样本和61k个合成代码理解任务样本。
SYNTHETIC-1 is a reasoning dataset obtained from Deepseek-R1, generated with crowdsourced compute and annotated with diverse verifiers such as LLM judges or symbolic mathematics verifiers. The dataset consists of various types of tasks including Mathematics Problems, Algorithmic Coding Problems, Real-World Software Engineering Problems, Open-Ended STEM Question Answering, and Synthetic Code Understanding Tasks, with a total of 777k, 144k, 70k, 313k, and 61k samples respectively.
提供机构：
PrimeIntellect
