SPoT: Better Frozen Model Adaptation through Soft Prompt Transfer
Tu Vu1,2★ Brian Lester1 Noah Constant1 Rami Al-Rfou1 Daniel Cer1 Google Research1 University of Massachusetts Amherst2
Abstract
There has been growing interest in parameter-efficient methods to apply pre-trained language models to downstream tasks. Building on the PromptTuning approach of Lester et al. (2021), which learns task-specific soft prompts to condition a frozen pre-trained model to perform different tasks, we propose a novel prompt-based transfer learning approach called SPoT: Soft Prompt Transfer. SPoT first learns a prompt on one or more source tasks and then uses it to initialize the prompt for a target task. We show that SPoT significantly boosts the performance of PromptTuning across many tasks. More remarkably, across all model sizes, SPoT matches or outperforms standard ModelTuning (which fine-tunes all model parameters) on the SuperGLUE benchmark, while using up to 27,000 $\times$ fewer task-specific parameters. To understand where SPoT is most effective, we conduct a large-scale study on task transferability with 26 NLP tasks in 160 combinations, and demonstrate that many tasks can benefit each other via prompt transfer. Finally, we propose an efficient retrieval approach that interprets task prompts as task embeddings to identify similar tasks and predict the most transferable source tas
中文速览
大规模预训练语言模型针对每个下游任务都做全量微调代价极高，软提示调优（Prompt Tuning）虽然只训练少量可学习的提示词就能冻结模型权重、完成任务适配，但在中小模型上性能与全参数微调仍有明显差距。为此，研究者提出了 SPoT（Soft Prompt Transfer），核心思路是先在一个或多个源任务上训练好软提示，再把它作为初始化参数迁移到目标任务的提示调优中，相当于给提示词一个"热身起点"。实验表明，SPoT 在 SuperGLUE 基准上全面追平乃至超越全参数微调，所用任务专属参数却少了约 27,000 倍，在 T5 Base 模型上更带来高达 10.1 分的平均精度提升。研究者还系统分析了 26 个 NLP 任务在 160 种源-目标组合下的迁移规律，并提出将任务提示词作为"任务嵌入"（task embedding）来度量任务相似度，从而高效预测哪些源任务最适合迁移到新目标任务，为低资源场景下的参数高效学习提供了切实可行的方法论。
原文 arXiv:2110.07904;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2110.07904v2