@inproceedings{schoch-etal-2023-data,
title = "Data Selection for Fine-tuning Large Language Models Using Transferred Shapley Values",
author = "Schoch, Stephanie and
Mishra, Ritwick and
Ji, Yangfeng",
editor = "Padmakumar, Vishakh and
Vallejo, Gisela and
Fu, Yao",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-srw.37",
doi = "10.18653/v1/2023.acl-srw.37",
pages = "266--275",
abstract = "Although Shapley values have been shown to be highly effective for identifying harmful training instances, dataset size and model complexity constraints limit the ability to apply Shapley-based data valuation to fine-tuning large pre-trained language models. To address this, we propose TS-DShapley, an algorithm that reduces computational cost of Shapley-based data valuation through: 1) an efficient sampling-based method that aggregates Shapley values computed from subsets for valuation of the entire training set, and 2) a value transfer method that leverages value information extracted from a simple classifier trained using representations from the target language model. Our experiments applying TS-DShapley to select data for fine-tuning BERT-based language models on benchmark natural language understanding (NLU) datasets show that TS-DShapley outperforms existing data selection methods. Further, TS-DShapley can filter fine-tuning data to increase language model performance compared to training with the full fine-tuning dataset.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="schoch-etal-2023-data">
<titleInfo>
<title>Data Selection for Fine-tuning Large Language Models Using Transferred Shapley Values</title>
</titleInfo>
<name type="personal">
<namePart type="given">Stephanie</namePart>
<namePart type="family">Schoch</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ritwick</namePart>
<namePart type="family">Mishra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yangfeng</namePart>
<namePart type="family">Ji</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 61st Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Vishakh</namePart>
<namePart type="family">Padmakumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gisela</namePart>
<namePart type="family">Vallejo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yao</namePart>
<namePart type="family">Fu</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>Although Shapley values have been shown to be highly effective for identifying harmful training instances, dataset size and model complexity constraints limit the ability to apply Shapley-based data valuation to fine-tuning large pre-trained language models. To address this, we propose TS-DShapley, an algorithm that reduces computational cost of Shapley-based data valuation through: 1) an efficient sampling-based method that aggregates Shapley values computed from subsets for valuation of the entire training set, and 2) a value transfer method that leverages value information extracted from a simple classifier trained using representations from the target language model. Our experiments applying TS-DShapley to select data for fine-tuning BERT-based language models on benchmark natural language understanding (NLU) datasets show that TS-DShapley outperforms existing data selection methods. Further, TS-DShapley can filter fine-tuning data to increase language model performance compared to training with the full fine-tuning dataset.</abstract>
<identifier type="citekey">schoch-etal-2023-data</identifier>
<identifier type="doi">10.18653/v1/2023.acl-srw.37</identifier>
<location>
<url>https://aclanthology.org/2023.acl-srw.37</url>
</location>
<part>
<date>2023-07</date>
<extent unit="page">
<start>266</start>
<end>275</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Data Selection for Fine-tuning Large Language Models Using Transferred Shapley Values
%A Schoch, Stephanie
%A Mishra, Ritwick
%A Ji, Yangfeng
%Y Padmakumar, Vishakh
%Y Vallejo, Gisela
%Y Fu, Yao
%S Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 4: Student Research Workshop)
%D 2023
%8 July
%I Association for Computational Linguistics
%C Toronto, Canada
%F schoch-etal-2023-data
%X Although Shapley values have been shown to be highly effective for identifying harmful training instances, dataset size and model complexity constraints limit the ability to apply Shapley-based data valuation to fine-tuning large pre-trained language models. To address this, we propose TS-DShapley, an algorithm that reduces computational cost of Shapley-based data valuation through: 1) an efficient sampling-based method that aggregates Shapley values computed from subsets for valuation of the entire training set, and 2) a value transfer method that leverages value information extracted from a simple classifier trained using representations from the target language model. Our experiments applying TS-DShapley to select data for fine-tuning BERT-based language models on benchmark natural language understanding (NLU) datasets show that TS-DShapley outperforms existing data selection methods. Further, TS-DShapley can filter fine-tuning data to increase language model performance compared to training with the full fine-tuning dataset.
%R 10.18653/v1/2023.acl-srw.37
%U https://aclanthology.org/2023.acl-srw.37
%U https://doi.org/10.18653/v1/2023.acl-srw.37
%P 266-275
Markdown (Informal)
[Data Selection for Fine-tuning Large Language Models Using Transferred Shapley Values](https://aclanthology.org/2023.acl-srw.37) (Schoch et al., ACL 2023)
ACL