Parameter-Efficient Learning

The Power of Scale for Parameter-Efficient Prompt Tuning

Prompt tuning improves substantially with model scale.