PPT: Pre-trained Prompt Tuning for Few-shot Learning
Yuxian Gu1,3∗, Xu Han2,3∗, Zhiyuan Liu2,3,4, Minlie Huang1,3,4† 1The CoAI group, Tsinghua University, Beijing, China 2The THUNLP group, Tsinghua University, Beijing, China 3Institute for Artificial Intelligence, State Key Lab of Intelligent Technology and Systems, Beijing National Research Center for Information Science and Technology, Department of Computer Science and Technology, Tsinghua University, Beijing, China 4 Beijing Academy of Artificial Intelligence, BAAI, Beijing, China
Abstract
Prompts for pre-trained language models (PLMs) have shown remarkable performance by bridging the gap between pre-training tasks and various downstream tasks. Among these methods, prompt tuning, which freezes PLMs and only tunes soft prompts, provides an efficient and effective solution for adapting large-scale PLMs to downstream tasks. However, prompt tuning is yet to be fully explored. In our pilot experiments, we find that prompt tuning performs comparably with conventional full-model tuning when downstream data are sufficient, whereas it is much worse under few-shot learning settings, which may hinder the application of prompt tuning. We attribute this low performance to the manner of initializing soft prompts. Therefore, in this work, we propose to pre-train prompts by adding soft prompts into the pre-training stage to obtain a better initialization. We name this Pre-trained Prompt Tuning framework “PPT”. To ensure the generalization of PPT, we formulate similar classification tasks into a unified task form and pre-train soft prompts for this unified task. Extensive experiments show that tuning pre-trained prompts for downstream tasks can reach or even outperform full-model fin
中文速览
软提示微调（prompt tuning, PT）是一种高效适配大规模预训练语言模型的方法——它冻结模型全部参数、只优化一小段可学习的"软提示"向量，避免了为每个下游任务保存一份完整模型的巨大开销。然而研究者发现，当下游标注数据极少时，PT的效果远不如对模型全参数微调（full-model tuning），瓶颈在于软提示的初始化质量太差。为此，本文提出PPT（Pre-trained Prompt Tuning）框架：在正式接触下游任务之前，先把软提示插入预训练阶段，用海量无标注语料上精心设计的自监督任务对其进行预训练，使其获得一个高质量的初始化；为了让预训练提示具有跨任务的泛化能力，作者将常见分类任务归纳为句对分类、多项选择和单句分类三种统一格式，分别设计对应的预训练目标，并进一步发现多项选择格式最为通用、可覆盖全部分类场景。在三个110亿参数级别的大模型（T5-XXL、mT5-XXL、CPM-2）上的实验表明，PPT在少样本和全量数据两种设置下均能达到甚至超过全参数微调的性能，同时大幅降低少样本学习的结果方差，对在实际中高效部署超大规模语言模型具有重要意义。
原文 arXiv:2109.04332;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2109.04332v3