MMODELYST
Papers/Prefix-Tuning: Optimizing Continuous Prompts for Generation
PAP

Prefix-Tuning: Optimizing Continuous Prompts for Generation

Jan 1, 2021

arXiv
292 citations
Abstract

Fine-tuning is the de facto way to leverage large pretrained language models to perform downstream tasks. However, it modifies all the language model parameters and therefore necessitates storing a full copy for each task. In this paper, we propose prefix-tuning, a lightweight alternative to fine-tuning for natural language generation tasks, which keeps language model parameters frozen, but optimizes a small continuous task-specific vector (called the prefix). Prefix-tuning draws inspiration from prompting, allowing subsequent tokens to attend to this prefix as if it were "virtual tokens". We apply prefix-tuning to GPT-2 for table-to-text generation and to BART for summarization. We find that by learning only 0.1\% of the parameters, prefix-tuning obtains comparable performance in the full data setting, outperforms fine-tuning in low-data settings, and extrapolates better to examples with topics unseen during training.

Select text to highlight · click a highlight to remove · saved in this browser only
Authors
Xiang Lisa Li, Percy Liang
Your notes (browser-local)
saved
arXiv:2101.00190