InstructAlign: High-and-Low Resource Language Alignment via Continual Crosslingual Instruction Tuning

Large language models (LLMs) that are tuned with instructions have demonstrated remarkable capabilities in various tasks and languages. However, their ability to generalize to underrepresented languages is limited due to the scarcity of available data. Additionally, directly adapting new languages t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2023-10
Hauptverfasser: Cahyawijaya, Samuel, Holy Lovenia, Yu, Tiezheng, Chung, Willy, Fung, Pascale
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Cahyawijaya, Samuel
Holy Lovenia
Yu, Tiezheng
Chung, Willy
Fung, Pascale
description Large language models (LLMs) that are tuned with instructions have demonstrated remarkable capabilities in various tasks and languages. However, their ability to generalize to underrepresented languages is limited due to the scarcity of available data. Additionally, directly adapting new languages to instruction-tuned LLMs can result in catastrophic forgetting, which leads to the loss of multitasking ability. To address this issue, we propose InstructAlign which uses continual crosslingual instruction tuning to enable LLMs to align new unseen languages with previously learned high-resource languages. Our results demonstrate the effectiveness of InstructAlign in enabling the model to understand low-resource languages with limited parallel data while preventing catastrophic forgetting. Our work contributes to the advancement of language adaptation methods, particularly for adapting instruction-tuned LLMs to underrepresented languages. Our code is released on https://github.com/HLTCHKUST/InstructAlign
format Article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2818533590</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2818533590</sourcerecordid><originalsourceid>FETCH-proquest_journals_28185335903</originalsourceid><addsrcrecordid>eNqNjcsKwjAURIMgWLT_cMF1oU2sVndSlApdSfcl1BhT4o3mob9vFd27GjhzmBmRiDKWJcWC0gmJnevTNKXLFc1zFhFxQOdt6PxWK4kbqJS8JBxPSW2ecBTOBNsJqDnKwKWAj3UV6OGhOJQGvcLANZTWOKfV29Lwm1QGoQk40BkZn7l2Iv7mlMz3u6askps19yCcb_vhCIeqpUVW5Izl65T9Z70AiKNHJw</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2818533590</pqid></control><display><type>article</type><title>InstructAlign: High-and-Low Resource Language Alignment via Continual Crosslingual Instruction Tuning</title><source>Free E- Journals</source><creator>Cahyawijaya, Samuel ; Holy Lovenia ; Yu, Tiezheng ; Chung, Willy ; Fung, Pascale</creator><creatorcontrib>Cahyawijaya, Samuel ; Holy Lovenia ; Yu, Tiezheng ; Chung, Willy ; Fung, Pascale</creatorcontrib><description>Large language models (LLMs) that are tuned with instructions have demonstrated remarkable capabilities in various tasks and languages. However, their ability to generalize to underrepresented languages is limited due to the scarcity of available data. Additionally, directly adapting new languages to instruction-tuned LLMs can result in catastrophic forgetting, which leads to the loss of multitasking ability. To address this issue, we propose InstructAlign which uses continual crosslingual instruction tuning to enable LLMs to align new unseen languages with previously learned high-resource languages. Our results demonstrate the effectiveness of InstructAlign in enabling the model to understand low-resource languages with limited parallel data while preventing catastrophic forgetting. Our work contributes to the advancement of language adaptation methods, particularly for adapting instruction-tuned LLMs to underrepresented languages. Our code is released on https://github.com/HLTCHKUST/InstructAlign</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Alignment ; Languages ; Large language models ; Multitasking ; Teaching methods ; Tuning</subject><ispartof>arXiv.org, 2023-10</ispartof><rights>2023. This work is published under http://creativecommons.org/licenses/by-sa/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>780,784</link.rule.ids></links><search><creatorcontrib>Cahyawijaya, Samuel</creatorcontrib><creatorcontrib>Holy Lovenia</creatorcontrib><creatorcontrib>Yu, Tiezheng</creatorcontrib><creatorcontrib>Chung, Willy</creatorcontrib><creatorcontrib>Fung, Pascale</creatorcontrib><title>InstructAlign: High-and-Low Resource Language Alignment via Continual Crosslingual Instruction Tuning</title><title>arXiv.org</title><description>Large language models (LLMs) that are tuned with instructions have demonstrated remarkable capabilities in various tasks and languages. However, their ability to generalize to underrepresented languages is limited due to the scarcity of available data. Additionally, directly adapting new languages to instruction-tuned LLMs can result in catastrophic forgetting, which leads to the loss of multitasking ability. To address this issue, we propose InstructAlign which uses continual crosslingual instruction tuning to enable LLMs to align new unseen languages with previously learned high-resource languages. Our results demonstrate the effectiveness of InstructAlign in enabling the model to understand low-resource languages with limited parallel data while preventing catastrophic forgetting. Our work contributes to the advancement of language adaptation methods, particularly for adapting instruction-tuned LLMs to underrepresented languages. Our code is released on https://github.com/HLTCHKUST/InstructAlign</description><subject>Alignment</subject><subject>Languages</subject><subject>Large language models</subject><subject>Multitasking</subject><subject>Teaching methods</subject><subject>Tuning</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><recordid>eNqNjcsKwjAURIMgWLT_cMF1oU2sVndSlApdSfcl1BhT4o3mob9vFd27GjhzmBmRiDKWJcWC0gmJnevTNKXLFc1zFhFxQOdt6PxWK4kbqJS8JBxPSW2ecBTOBNsJqDnKwKWAj3UV6OGhOJQGvcLANZTWOKfV29Lwm1QGoQk40BkZn7l2Iv7mlMz3u6askps19yCcb_vhCIeqpUVW5Izl65T9Z70AiKNHJw</recordid><startdate>20231024</startdate><enddate>20231024</enddate><creator>Cahyawijaya, Samuel</creator><creator>Holy Lovenia</creator><creator>Yu, Tiezheng</creator><creator>Chung, Willy</creator><creator>Fung, Pascale</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20231024</creationdate><title>InstructAlign: High-and-Low Resource Language Alignment via Continual Crosslingual Instruction Tuning</title><author>Cahyawijaya, Samuel ; Holy Lovenia ; Yu, Tiezheng ; Chung, Willy ; Fung, Pascale</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_28185335903</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Alignment</topic><topic>Languages</topic><topic>Large language models</topic><topic>Multitasking</topic><topic>Teaching methods</topic><topic>Tuning</topic><toplevel>online_resources</toplevel><creatorcontrib>Cahyawijaya, Samuel</creatorcontrib><creatorcontrib>Holy Lovenia</creatorcontrib><creatorcontrib>Yu, Tiezheng</creatorcontrib><creatorcontrib>Chung, Willy</creatorcontrib><creatorcontrib>Fung, Pascale</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Access via ProQuest (Open Access)</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Cahyawijaya, Samuel</au><au>Holy Lovenia</au><au>Yu, Tiezheng</au><au>Chung, Willy</au><au>Fung, Pascale</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>InstructAlign: High-and-Low Resource Language Alignment via Continual Crosslingual Instruction Tuning</atitle><jtitle>arXiv.org</jtitle><date>2023-10-24</date><risdate>2023</risdate><eissn>2331-8422</eissn><abstract>Large language models (LLMs) that are tuned with instructions have demonstrated remarkable capabilities in various tasks and languages. However, their ability to generalize to underrepresented languages is limited due to the scarcity of available data. Additionally, directly adapting new languages to instruction-tuned LLMs can result in catastrophic forgetting, which leads to the loss of multitasking ability. To address this issue, we propose InstructAlign which uses continual crosslingual instruction tuning to enable LLMs to align new unseen languages with previously learned high-resource languages. Our results demonstrate the effectiveness of InstructAlign in enabling the model to understand low-resource languages with limited parallel data while preventing catastrophic forgetting. Our work contributes to the advancement of language adaptation methods, particularly for adapting instruction-tuned LLMs to underrepresented languages. Our code is released on https://github.com/HLTCHKUST/InstructAlign</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2023-10
issn 2331-8422
language eng
recordid cdi_proquest_journals_2818533590
source Free E- Journals
subjects Alignment
Languages
Large language models
Multitasking
Teaching methods
Tuning
title InstructAlign: High-and-Low Resource Language Alignment via Continual Crosslingual Instruction Tuning
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-20T10%3A19%3A00IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=InstructAlign:%20High-and-Low%20Resource%20Language%20Alignment%20via%20Continual%20Crosslingual%20Instruction%20Tuning&rft.jtitle=arXiv.org&rft.au=Cahyawijaya,%20Samuel&rft.date=2023-10-24&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2818533590%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2818533590&rft_id=info:pmid/&rfr_iscdi=true