Towards Foundation Models for Scientific Machine Learning: Characterizing Scaling and Transfer Behavior

Pre-trained machine learning (ML) models have shown great performance for a wide range of applications, in particular in natural language processing (NLP) and computer vision (CV).