Is there a decent mapping of reasoning levels to problem sets for this model?