@inproceedings{wang-etal-2024-lora,
title = "{L}o{RA} Meets Dropout under a Unified Framework",
author = "Wang, Sheng and
Chen, Liheng and
Jiang, Jiyue and
Xue, Boyang and
Kong, Lingpeng and
Wu, Chuan",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2024",
month = aug,
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-acl.119",
doi = "10.18653/v1/2024.findings-acl.119",
pages = "1995--2008",
abstract = "With the remarkable capabilities, large language models (LLMs) have emergedas essential elements in numerous NLP applications, while parameter-efficientfinetuning, especially LoRA, has gained popularity as a lightweight approachfor model customization. Meanwhile, various dropout methods, initially designedfor full finetuning with all the parameters updated, alleviates overfittingassociated with excessive parameter redundancy. Hence, a possible contradictionarises from negligible trainable parameters of LoRA and the effectiveness ofprevious dropout methods, which has been largely overlooked. To fill this gap,we first confirm that parameter-efficient LoRA is also overfitting-prone. Wethen revisit transformer-specific dropout methods, and establish theirequivalence and distinctions mathematically and empirically. Building upon thiscomparative analysis, we introduce a unified framework for a comprehensiveinvestigation, which instantiates these methods based on dropping position,structural pattern and compensation measure. Through this framework, we revealthe new preferences and performance comparisons of them when involved withlimited trainable parameters. This framework also allows us to amalgamate themost favorable aspects into a novel dropout method named HiddenKey. Extensiveexperiments verify the remarkable superiority and sufficiency of HiddenKeyacross multiple models and tasks, which highlights it as the preferred approachfor high-performance and parameter-efficient finetuning of LLMs.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2024-lora">
<titleInfo>
<title>LoRA Meets Dropout under a Unified Framework</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sheng</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liheng</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiyue</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Boyang</namePart>
<namePart type="family">Xue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lingpeng</namePart>
<namePart type="family">Kong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chuan</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lun-Wei</namePart>
<namePart type="family">Ku</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andre</namePart>
<namePart type="family">Martins</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Srikumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Bangkok, Thailand</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>With the remarkable capabilities, large language models (LLMs) have emergedas essential elements in numerous NLP applications, while parameter-efficientfinetuning, especially LoRA, has gained popularity as a lightweight approachfor model customization. Meanwhile, various dropout methods, initially designedfor full finetuning with all the parameters updated, alleviates overfittingassociated with excessive parameter redundancy. Hence, a possible contradictionarises from negligible trainable parameters of LoRA and the effectiveness ofprevious dropout methods, which has been largely overlooked. To fill this gap,we first confirm that parameter-efficient LoRA is also overfitting-prone. Wethen revisit transformer-specific dropout methods, and establish theirequivalence and distinctions mathematically and empirically. Building upon thiscomparative analysis, we introduce a unified framework for a comprehensiveinvestigation, which instantiates these methods based on dropping position,structural pattern and compensation measure. Through this framework, we revealthe new preferences and performance comparisons of them when involved withlimited trainable parameters. This framework also allows us to amalgamate themost favorable aspects into a novel dropout method named HiddenKey. Extensiveexperiments verify the remarkable superiority and sufficiency of HiddenKeyacross multiple models and tasks, which highlights it as the preferred approachfor high-performance and parameter-efficient finetuning of LLMs.</abstract>
<identifier type="citekey">wang-etal-2024-lora</identifier>
<identifier type="doi">10.18653/v1/2024.findings-acl.119</identifier>
<location>
<url>https://aclanthology.org/2024.findings-acl.119</url>
</location>
<part>
<date>2024-08</date>
<extent unit="page">
<start>1995</start>
<end>2008</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T LoRA Meets Dropout under a Unified Framework
%A Wang, Sheng
%A Chen, Liheng
%A Jiang, Jiyue
%A Xue, Boyang
%A Kong, Lingpeng
%A Wu, Chuan
%Y Ku, Lun-Wei
%Y Martins, Andre
%Y Srikumar, Vivek
%S Findings of the Association for Computational Linguistics: ACL 2024
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand
%F wang-etal-2024-lora
%X With the remarkable capabilities, large language models (LLMs) have emergedas essential elements in numerous NLP applications, while parameter-efficientfinetuning, especially LoRA, has gained popularity as a lightweight approachfor model customization. Meanwhile, various dropout methods, initially designedfor full finetuning with all the parameters updated, alleviates overfittingassociated with excessive parameter redundancy. Hence, a possible contradictionarises from negligible trainable parameters of LoRA and the effectiveness ofprevious dropout methods, which has been largely overlooked. To fill this gap,we first confirm that parameter-efficient LoRA is also overfitting-prone. Wethen revisit transformer-specific dropout methods, and establish theirequivalence and distinctions mathematically and empirically. Building upon thiscomparative analysis, we introduce a unified framework for a comprehensiveinvestigation, which instantiates these methods based on dropping position,structural pattern and compensation measure. Through this framework, we revealthe new preferences and performance comparisons of them when involved withlimited trainable parameters. This framework also allows us to amalgamate themost favorable aspects into a novel dropout method named HiddenKey. Extensiveexperiments verify the remarkable superiority and sufficiency of HiddenKeyacross multiple models and tasks, which highlights it as the preferred approachfor high-performance and parameter-efficient finetuning of LLMs.
%R 10.18653/v1/2024.findings-acl.119
%U https://aclanthology.org/2024.findings-acl.119
%U https://doi.org/10.18653/v1/2024.findings-acl.119
%P 1995-2008
Markdown (Informal)
[LoRA Meets Dropout under a Unified Framework](https://aclanthology.org/2024.findings-acl.119) (Wang et al., Findings 2024)
ACL
- Sheng Wang, Liheng Chen, Jiyue Jiang, Boyang Xue, Lingpeng Kong, and Chuan Wu. 2024. LoRA Meets Dropout under a Unified Framework. In Findings of the Association for Computational Linguistics: ACL 2024, pages 1995–2008, Bangkok, Thailand. Association for Computational Linguistics.