When Can Models Learn From Explanations? A Formal Framework for Understanding the Roles of Explanation Data

Open in new window