We propose prototypical fine-tuning, a novel framework for fine-tuning pretrained language models that maintains robust performance across varying data sizes.
Jan 1, 2023
