OctoThinker/MegaMath-Web-Pro-Max

OctoThinker数据集是从MegaMath-Web语料库中随机抽取并经过一系列处理（包括注释、训练、筛选和细化）的数学网络数据集。它旨在通过中训练阶段激励强化学习的扩展。数据集的获取和使用受到特定许可的限制。

The OctoThinker dataset is a collection of math web data sampled randomly from the MegaMath-Web corpus and processed through a series of steps including annotation, training, filtering, and refinement. It is designed to incentivize reinforcement learning scaling through mid-training stages. Access and use of the dataset are subject to specific licensing restrictions.