What is the difference between prompt tuning and finetuning?

Prompt tuning involves optimizing a small set of parameters to tailor the model's output based on a prompt without altering the core model weights. This technique is generally lightweight and more efficient as it requires less data and computational resources. In contrast, finetuning is a comprehensive process that updates the entire model weights using a larger dataset, which can lead to improved performance on specific tasks but is more resource-intensive and time-consuming.