P2TPO: Structure-Aligned Page-to-Token Credit Assignment for Full-Page Document Parsing Paper • 2610.21979 • Published Oct 15 • 7
view article Article Cosmopedia:如何为大语言模型预训练构建大规模合成数据集 +1 loubnabnl, anton-l, davanstrien • Mar 20, 2024 • 4