In the function 'test' of test_few.py(as well train_few.py), the current code computes img-AUROC of the dataset Brain (as well as Liver, Retina_RESC) using the seg_patch_tokens, instead of det_patch_tokens, which is different with the paper. Will it make differences in testing result?