Versatile Medical Image Segmentation Learned from Multi-Source Datasets via Model Self-Disambiguation | Read Paper on Bytez