Hello author, thank you for your work, but I have a question. Why not first calculate the AUC within each sequence and then calculate the average AUC of all sequences as the final result, instead of collecting the corner errors of all sequences first and then calculating AUC?