ParentFull threadisusmelj·Yes, we're just finalizing other case studies with more diverse data. The trend however is similar. Removing semantic redundancy from a dataset helps reduce annotation and storage/compute costs with little loss in accuracy.View on HN