What are the differences between adapter tuning and prefix turing?

I am trying to understand the concept of adapter-tuning, prompt-tuning, and prefix-tuning in the context of few-shot learning.