Global Prompt Refinement with Non-Interfering Attention Masking for One-Shot Federated Learning | Read Paper on Bytez