Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning

A growing body of literature has focused on detailing the linguistic knowledge embedded in large, pretrained language models. Existing work has shown that non-linguistic biases in models can drive model behavior away from linguistic generalizations. We hypothesized that competing linguistic processe...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2021-06
Hauptverfasser: Davis, rest, Marten van Schijndel
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 Davis, rest
Marten van Schijndel
description A growing body of literature has focused on detailing the linguistic knowledge embedded in large, pretrained language models. Existing work has shown that non-linguistic biases in models can drive model behavior away from linguistic generalizations. We hypothesized that competing linguistic processes within a language, rather than just non-linguistic model biases, could obscure underlying linguistic knowledge. We tested this claim by exploring a single phenomenon in four languages: English, Chinese, Spanish, and Italian. While human behavior has been found to be similar across languages, we find cross-linguistic variation in model behavior. We show that competing processes in a language act as constraints on model behavior and demonstrate that targeted fine-tuning can re-weight the learned constraints, uncovering otherwise dormant linguistic knowledge in models. Our results suggest that models need to learn both the linguistic constraints in a language and their relative ranking, with mismatches in either producing non-human-like behavior.
format Article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2536671408</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2536671408</sourcerecordid><originalsourceid>FETCH-proquest_journals_25366714083</originalsourceid><addsrcrecordid>eNqNjL0OgjAYABsTE4nyDk2cm5SWAjNE4iITzqSRTywhX7U_PL8MPoDTDXe5HUmElBmrciEOJPV-5pyLohRKyYR0d3zYFZzBiTYWfXDaYGC19jDSGl56NdZRg7SD6PRCb3aExdPVaNprN0HYstYgsD7i9jiR_VMvHtIfj-TcXvrmyt7OfiL4MMw2OtzUIJQsijLLeSX_q763nz3e</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2536671408</pqid></control><display><type>article</type><title>Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning</title><source>Free E- Journals</source><creator>Davis, rest ; Marten van Schijndel</creator><creatorcontrib>Davis, rest ; Marten van Schijndel</creatorcontrib><description>A growing body of literature has focused on detailing the linguistic knowledge embedded in large, pretrained language models. Existing work has shown that non-linguistic biases in models can drive model behavior away from linguistic generalizations. We hypothesized that competing linguistic processes within a language, rather than just non-linguistic model biases, could obscure underlying linguistic knowledge. We tested this claim by exploring a single phenomenon in four languages: English, Chinese, Spanish, and Italian. While human behavior has been found to be similar across languages, we find cross-linguistic variation in model behavior. We show that competing processes in a language act as constraints on model behavior and demonstrate that targeted fine-tuning can re-weight the learned constraints, uncovering otherwise dormant linguistic knowledge in models. Our results suggest that models need to learn both the linguistic constraints in a language and their relative ranking, with mismatches in either producing non-human-like behavior.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Constraint modelling ; Human behavior ; Languages ; Linguistics</subject><ispartof>arXiv.org, 2021-06</ispartof><rights>2021. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (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>Davis, rest</creatorcontrib><creatorcontrib>Marten van Schijndel</creatorcontrib><title>Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning</title><title>arXiv.org</title><description>A growing body of literature has focused on detailing the linguistic knowledge embedded in large, pretrained language models. Existing work has shown that non-linguistic biases in models can drive model behavior away from linguistic generalizations. We hypothesized that competing linguistic processes within a language, rather than just non-linguistic model biases, could obscure underlying linguistic knowledge. We tested this claim by exploring a single phenomenon in four languages: English, Chinese, Spanish, and Italian. While human behavior has been found to be similar across languages, we find cross-linguistic variation in model behavior. We show that competing processes in a language act as constraints on model behavior and demonstrate that targeted fine-tuning can re-weight the learned constraints, uncovering otherwise dormant linguistic knowledge in models. Our results suggest that models need to learn both the linguistic constraints in a language and their relative ranking, with mismatches in either producing non-human-like behavior.</description><subject>Constraint modelling</subject><subject>Human behavior</subject><subject>Languages</subject><subject>Linguistics</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><recordid>eNqNjL0OgjAYABsTE4nyDk2cm5SWAjNE4iITzqSRTywhX7U_PL8MPoDTDXe5HUmElBmrciEOJPV-5pyLohRKyYR0d3zYFZzBiTYWfXDaYGC19jDSGl56NdZRg7SD6PRCb3aExdPVaNprN0HYstYgsD7i9jiR_VMvHtIfj-TcXvrmyt7OfiL4MMw2OtzUIJQsijLLeSX_q763nz3e</recordid><startdate>20210602</startdate><enddate>20210602</enddate><creator>Davis, rest</creator><creator>Marten van Schijndel</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>20210602</creationdate><title>Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning</title><author>Davis, rest ; Marten van Schijndel</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_25366714083</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Constraint modelling</topic><topic>Human behavior</topic><topic>Languages</topic><topic>Linguistics</topic><toplevel>online_resources</toplevel><creatorcontrib>Davis, rest</creatorcontrib><creatorcontrib>Marten van Schijndel</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>Publicly Available Content Database</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>Davis, rest</au><au>Marten van Schijndel</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning</atitle><jtitle>arXiv.org</jtitle><date>2021-06-02</date><risdate>2021</risdate><eissn>2331-8422</eissn><abstract>A growing body of literature has focused on detailing the linguistic knowledge embedded in large, pretrained language models. Existing work has shown that non-linguistic biases in models can drive model behavior away from linguistic generalizations. We hypothesized that competing linguistic processes within a language, rather than just non-linguistic model biases, could obscure underlying linguistic knowledge. We tested this claim by exploring a single phenomenon in four languages: English, Chinese, Spanish, and Italian. While human behavior has been found to be similar across languages, we find cross-linguistic variation in model behavior. We show that competing processes in a language act as constraints on model behavior and demonstrate that targeted fine-tuning can re-weight the learned constraints, uncovering otherwise dormant linguistic knowledge in models. Our results suggest that models need to learn both the linguistic constraints in a language and their relative ranking, with mismatches in either producing non-human-like behavior.</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, 2021-06
issn 2331-8422
language eng
recordid cdi_proquest_journals_2536671408
source Free E- Journals
subjects Constraint modelling
Human behavior
Languages
Linguistics
title Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-05T18%3A18%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=Uncovering%20Constraint-Based%20Behavior%20in%20Neural%20Models%20via%20Targeted%20Fine-Tuning&rft.jtitle=arXiv.org&rft.au=Davis,%20rest&rft.date=2021-06-02&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2536671408%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2536671408&rft_id=info:pmid/&rfr_iscdi=true