If two grader models disagree on 37% of the same patches, there's no actual methodology. It's just a coin flip.
If two grader models disagree on 37% of the same patches, there's no actual methodology. It's just a coin flip.