VLG-CBM: Training Concept Bottleneck Models with Vision-Language Guidance | Read Paper on Bytez