How you eval recall on deepcodebench?

Hi qodo team! I want to know when evaluating recall, do we calculate the recall for each sample separately and then take the average, or calculate the recall of all fact points across the entire test set at once?