Learning with Logical Constraints but without Shortcut Satisfaction

Open in new window