Efficient and Long-Tailed Generalization for Pre-trained Vision-Language Model | Read Paper on Bytez