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
0
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.