Abstract: Prompt tuning is a valuable technique for adapting visual language models (VLMs) to different downstream tasks, such as domain generalization and learning from a few examples. Previous ...