SCIMAT: Science and Mathematics Dataset

Abstract
In this work, we announce a comprehensive well curated and opensource dataset with millions of samples for pre-college and college level problems in mathematicsand science. A preliminary set of results using transformer architecture with character to character encoding is shown. The dataset identifies some challenging problem and invites research on better architecture search
View on arXivComments on this paper