arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.10385cs.CLcs.LG
收藏

GPT 也懂理解

GPT Understands, Too

Tsinghua University(清华大学);Massachusetts Institute of Technology(麻省理工学院)

Xiao Liu, Yanan Zheng, Zhengxiao Du, Ming Ding, Yujie Qian, Zhilin Yang, Jie Tang

AI总结:

针对离散提示性能不稳定的问题,提出 P-Tuning 方法,通过结合可训练的连续提示嵌入与离散提示,稳定训练过程并在多种 NLU 任务上大幅提升性能。

摘要:

使用自然语言模式提示预训练语言模型已被证明对自然语言理解(NLU)有效。然而,我们的初步研究表明,人工离散提示通常会导致不稳定的性能——例如,改变提示中的单个单词可能会导致性能大幅下降。我们提出了一种新方法 P-Tuning,该方法采用可训练的连续提示嵌入,并将其与离散提示结合使用。在实验中,P-Tuning 不仅通过最小化各种离散提示之间的差距来稳定训练,而且在包括 LAMA 和 SuperGLUE 在内的广泛 NLU 任务上大幅提升了性能。P-Tuning 在冻结和微调的语言模型中,以及在全监督和少样本设置下均普遍有效。

英文摘要:

Prompting a pretrained language model with natural language patterns has been proved effective for natural language understanding (NLU). However, our preliminary study reveals that manual discrete prompts often lead to unstable performance -- e.g., changing a single word in the prompt might result in substantial performance drop. We propose a novel method P-Tuning that employs trainable continuous prompt embeddings in concatenation with discrete prompts. Empirically, P-Tuning not only stabilizes training by minimizing the gap between various discrete prompts, but also improves performance by a sizeable margin on a wide range of NLU tasks including LAMA and SuperGLUE. P-Tuning is generally effective for both frozen and tuned language models, under both the fully-supervised and few-shot settings.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑