Text this: MDAPT: Multi-Modal Depth Adversarial Prompt Tuning to Enhance the Adversarial Robustness of Visual Language Models