Robustness in Both Domains: CLIP Needs a Robust Text Encoder | Read Paper on Bytez