@inproceedings{obadinma-etal-2023-effectiveness,
title = "Effectiveness of Data Augmentation for Parameter Efficient Tuning with Limited Data",
author = "Obadinma, Stephen and
Guo, Hongyu and
Zhu, Xiaodan",
editor = "Can, Burcu and
Mozes, Maximilian and
Cahyawijaya, Samuel and
Saphra, Naomi and
Kassner, Nora and
Ravfogel, Shauli and
Ravichander, Abhilasha and
Zhao, Chen and
Augenstein, Isabelle and
Rogers, Anna and
Cho, Kyunghyun and
Grefenstette, Edward and
Voita, Lena",
booktitle = "Proceedings of the 8th Workshop on Representation Learning for NLP (RepL4NLP 2023)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.repl4nlp-1.19",
doi = "10.18653/v1/2023.repl4nlp-1.19",
pages = "226--237",
abstract = "Recent work has demonstrated that using parameter efficient tuning techniques such as prefix tuning (or P-tuning) on pretrained language models can yield performance that is comparable or superior to fine-tuning while dramatically reducing trainable parameters. Nevertheless, the effectiveness of such methods under the context of data augmentation, a common strategy to improve learning under low data regimes, has not been fully explored. In this paper, we examine the effectiveness of several popular task-agnostic data augmentation techniques, i.e., EDA, Back Translation, and Mixup, when using two general parameter efficient tuning methods, P-tuning v2 and LoRA, under data scarcity. We show that data augmentation can be used to boost the performance of P-tuning and LoRA models, but the effectiveness of each technique varies and certain methods can lead to a notable degradation in performance, particularly when using larger models and on harder tasks. We further analyze the sentence representations of P-tuning compared to fine-tuning to help understand the above behaviour, and reveal how P-tuning generally presents a more limited ability to separate the sentence embeddings from different classes of augmented data. In addition, it displays poorer performance on heavily altered data. However, we demonstrate that by adding a simple contrastive loss function it can help mitigate such issues for prefix tuning, resulting in sizable improvements to augmented data performance.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="obadinma-etal-2023-effectiveness">
<titleInfo>
<title>Effectiveness of Data Augmentation for Parameter Efficient Tuning with Limited Data</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stephen</namePart>
<namePart type="family">Obadinma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hongyu</namePart>
<namePart type="family">Guo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaodan</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 8th Workshop on Representation Learning for NLP (RepL4NLP 2023)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Burcu</namePart>
<namePart type="family">Can</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maximilian</namePart>
<namePart type="family">Mozes</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Samuel</namePart>
<namePart type="family">Cahyawijaya</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Naomi</namePart>
<namePart type="family">Saphra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nora</namePart>
<namePart type="family">Kassner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shauli</namePart>
<namePart type="family">Ravfogel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Abhilasha</namePart>
<namePart type="family">Ravichander</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chen</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Isabelle</namePart>
<namePart type="family">Augenstein</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kyunghyun</namePart>
<namePart type="family">Cho</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Edward</namePart>
<namePart type="family">Grefenstette</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lena</namePart>
<namePart type="family">Voita</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Toronto, Canada</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Recent work has demonstrated that using parameter efficient tuning techniques such as prefix tuning (or P-tuning) on pretrained language models can yield performance that is comparable or superior to fine-tuning while dramatically reducing trainable parameters. Nevertheless, the effectiveness of such methods under the context of data augmentation, a common strategy to improve learning under low data regimes, has not been fully explored. In this paper, we examine the effectiveness of several popular task-agnostic data augmentation techniques, i.e., EDA, Back Translation, and Mixup, when using two general parameter efficient tuning methods, P-tuning v2 and LoRA, under data scarcity. We show that data augmentation can be used to boost the performance of P-tuning and LoRA models, but the effectiveness of each technique varies and certain methods can lead to a notable degradation in performance, particularly when using larger models and on harder tasks. We further analyze the sentence representations of P-tuning compared to fine-tuning to help understand the above behaviour, and reveal how P-tuning generally presents a more limited ability to separate the sentence embeddings from different classes of augmented data. In addition, it displays poorer performance on heavily altered data. However, we demonstrate that by adding a simple contrastive loss function it can help mitigate such issues for prefix tuning, resulting in sizable improvements to augmented data performance.</abstract>
<identifier type="citekey">obadinma-etal-2023-effectiveness</identifier>
<identifier type="doi">10.18653/v1/2023.repl4nlp-1.19</identifier>
<location>
<url>https://aclanthology.org/2023.repl4nlp-1.19</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>226</start>
<end>237</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Effectiveness of Data Augmentation for Parameter Efficient Tuning with Limited Data
%A Obadinma, Stephen
%A Guo, Hongyu
%A Zhu, Xiaodan
%Y Can, Burcu
%Y Mozes, Maximilian
%Y Cahyawijaya, Samuel
%Y Saphra, Naomi
%Y Kassner, Nora
%Y Ravfogel, Shauli
%Y Ravichander, Abhilasha
%Y Zhao, Chen
%Y Augenstein, Isabelle
%Y Rogers, Anna
%Y Cho, Kyunghyun
%Y Grefenstette, Edward
%Y Voita, Lena
%S Proceedings of the 8th Workshop on Representation Learning for NLP (RepL4NLP 2023)
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F obadinma-etal-2023-effectiveness
%X Recent work has demonstrated that using parameter efficient tuning techniques such as prefix tuning (or P-tuning) on pretrained language models can yield performance that is comparable or superior to fine-tuning while dramatically reducing trainable parameters. Nevertheless, the effectiveness of such methods under the context of data augmentation, a common strategy to improve learning under low data regimes, has not been fully explored. In this paper, we examine the effectiveness of several popular task-agnostic data augmentation techniques, i.e., EDA, Back Translation, and Mixup, when using two general parameter efficient tuning methods, P-tuning v2 and LoRA, under data scarcity. We show that data augmentation can be used to boost the performance of P-tuning and LoRA models, but the effectiveness of each technique varies and certain methods can lead to a notable degradation in performance, particularly when using larger models and on harder tasks. We further analyze the sentence representations of P-tuning compared to fine-tuning to help understand the above behaviour, and reveal how P-tuning generally presents a more limited ability to separate the sentence embeddings from different classes of augmented data. In addition, it displays poorer performance on heavily altered data. However, we demonstrate that by adding a simple contrastive loss function it can help mitigate such issues for prefix tuning, resulting in sizable improvements to augmented data performance.
%R 10.18653/v1/2023.repl4nlp-1.19
%U https://aclanthology.org/2023.repl4nlp-1.19
%U https://doi.org/10.18653/v1/2023.repl4nlp-1.19
%P 226-237
Markdown (Informal)
[Effectiveness of Data Augmentation for Parameter Efficient Tuning with Limited Data](https://aclanthology.org/2023.repl4nlp-1.19) (Obadinma et al., RepL4NLP 2023)
ACL