JoAPR: Cleaning the Lens of Prompt Learning for Vision-Language Models | Read Paper on Bytez