Systematic Analysis for Pretrained Language Model Priming for Parameter-Efficient Fine-tuning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Huang, Shih-Cheng, Wang, Shih-Heng, Shih, Min-Han, Sahay, Saurav, Lee, Hung-yi
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911892864237568
author Huang, Shih-Cheng
Wang, Shih-Heng
Shih, Min-Han
Sahay, Saurav
Lee, Hung-yi
author_facet Huang, Shih-Cheng
Wang, Shih-Heng
Shih, Min-Han
Sahay, Saurav
Lee, Hung-yi
contents Parameter-efficient (PE) methods (like Prompts or Adapters) for adapting pre-trained language models (PLM) to downstream tasks have been popular recently. However, hindrances still prevent these methods from reaching their full potential. For example, two significant challenges are few-shot adaptation and cross-task generalization. To tackle these issues, we propose a general PE priming framework to enhance and explore the few-shot adaptation and generalization ability of PE methods. In this framework, PLMs are primed with PE methods for rapidly adapting to various target tasks. To evaluate the generalization ability of these PE methods, we conduct experiments on a few-shot cross-domain benchmark containing 160 diverse NLP tasks. Our experiment not only reveals the best priming strategy but also verifies that priming facilitates the adaptation to target tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2212_01032
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle Systematic Analysis for Pretrained Language Model Priming for Parameter-Efficient Fine-tuning
Huang, Shih-Cheng
Wang, Shih-Heng
Shih, Min-Han
Sahay, Saurav
Lee, Hung-yi
Computation and Language
Artificial Intelligence
Parameter-efficient (PE) methods (like Prompts or Adapters) for adapting pre-trained language models (PLM) to downstream tasks have been popular recently. However, hindrances still prevent these methods from reaching their full potential. For example, two significant challenges are few-shot adaptation and cross-task generalization. To tackle these issues, we propose a general PE priming framework to enhance and explore the few-shot adaptation and generalization ability of PE methods. In this framework, PLMs are primed with PE methods for rapidly adapting to various target tasks. To evaluate the generalization ability of these PE methods, we conduct experiments on a few-shot cross-domain benchmark containing 160 diverse NLP tasks. Our experiment not only reveals the best priming strategy but also verifies that priming facilitates the adaptation to target tasks.
title Systematic Analysis for Pretrained Language Model Priming for Parameter-Efficient Fine-tuning
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2212.01032