Increasing Trust in Language Models through the Reuse of Verified Circuits

Language Models (LMs) are increasingly used for a wide range of prediction tasks, but their training can often neglect rare edge cases, reducing their reliability. Here, we define a stringent standard of trustworthiness whereby the task algorithm and circuit implementation must be verified, accounti...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Quirke, Philip, Neo, Clement, Barez, Fazl
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Quirke, Philip
Neo, Clement
Barez, Fazl
description Language Models (LMs) are increasingly used for a wide range of prediction tasks, but their training can often neglect rare edge cases, reducing their reliability. Here, we define a stringent standard of trustworthiness whereby the task algorithm and circuit implementation must be verified, accounting for edge cases, with no known failure modes. We show that a model can be trained to meet this standard if built using mathematically and logically specified frameworks. In this paper, we fully verify an auto-regressive transformer model for n-digit integer addition. To exhibit the reusability of verified modules, we insert the trained integer addition model into a larger untrained model and train the combined model to perform both addition and subtraction. We find extensive reuse of the addition circuits for both tasks, easing verification of the more complex subtractor model. We discuss how inserting verified task modules into LMs can leverage model reuse to improve verifiability and trustworthiness of language models built using them. The reuse of verified circuits reduces the effort to verify more complex composite models which we believe to be a significant step towards safety of language models.
doi_str_mv 10.48550/arxiv.2402.02619
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2402_02619</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2402_02619</sourcerecordid><originalsourceid>FETCH-LOGICAL-a679-712b09cc0bfa3056df300ddafc2fa2784aaa75063bf2b43c0ffde9fb7a8a25c33</originalsourceid><addsrcrecordid>eNotz8tOhDAUgOFuXJgZH8CVfQHw0FIKywnxMoaJiSFuyenlME1GMC0YfXt1dPXv_uRj7LqAvKyVgluMn-EjFyWIHERVNJfsaT_Z6DGFaeR9XNPCw8Q7nMYVR88Ps_OnxJdjnNfx-FPPX_yaPJ-Jv_oYKHjH2xDtGpa0ZReEp-Sv_rth_f1d3z5m3fPDvt11GVa6yXQhDDTWgiGUoCpHEsA5JCsIha5LRNQKKmlImFJaIHK-IaOxRqGslBt287c9Y4b3GN4wfg2_qOGMkt9kjUhM</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Increasing Trust in Language Models through the Reuse of Verified Circuits</title><source>arXiv.org</source><creator>Quirke, Philip ; Neo, Clement ; Barez, Fazl</creator><creatorcontrib>Quirke, Philip ; Neo, Clement ; Barez, Fazl</creatorcontrib><description>Language Models (LMs) are increasingly used for a wide range of prediction tasks, but their training can often neglect rare edge cases, reducing their reliability. Here, we define a stringent standard of trustworthiness whereby the task algorithm and circuit implementation must be verified, accounting for edge cases, with no known failure modes. We show that a model can be trained to meet this standard if built using mathematically and logically specified frameworks. In this paper, we fully verify an auto-regressive transformer model for n-digit integer addition. To exhibit the reusability of verified modules, we insert the trained integer addition model into a larger untrained model and train the combined model to perform both addition and subtraction. We find extensive reuse of the addition circuits for both tasks, easing verification of the more complex subtractor model. We discuss how inserting verified task modules into LMs can leverage model reuse to improve verifiability and trustworthiness of language models built using them. The reuse of verified circuits reduces the effort to verify more complex composite models which we believe to be a significant step towards safety of language models.</description><identifier>DOI: 10.48550/arxiv.2402.02619</identifier><language>eng</language><subject>Computer Science - Computation and Language ; Computer Science - Learning</subject><creationdate>2024-02</creationdate><rights>http://creativecommons.org/licenses/by-sa/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2402.02619$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2402.02619$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Quirke, Philip</creatorcontrib><creatorcontrib>Neo, Clement</creatorcontrib><creatorcontrib>Barez, Fazl</creatorcontrib><title>Increasing Trust in Language Models through the Reuse of Verified Circuits</title><description>Language Models (LMs) are increasingly used for a wide range of prediction tasks, but their training can often neglect rare edge cases, reducing their reliability. Here, we define a stringent standard of trustworthiness whereby the task algorithm and circuit implementation must be verified, accounting for edge cases, with no known failure modes. We show that a model can be trained to meet this standard if built using mathematically and logically specified frameworks. In this paper, we fully verify an auto-regressive transformer model for n-digit integer addition. To exhibit the reusability of verified modules, we insert the trained integer addition model into a larger untrained model and train the combined model to perform both addition and subtraction. We find extensive reuse of the addition circuits for both tasks, easing verification of the more complex subtractor model. We discuss how inserting verified task modules into LMs can leverage model reuse to improve verifiability and trustworthiness of language models built using them. The reuse of verified circuits reduces the effort to verify more complex composite models which we believe to be a significant step towards safety of language models.</description><subject>Computer Science - Computation and Language</subject><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz8tOhDAUgOFuXJgZH8CVfQHw0FIKywnxMoaJiSFuyenlME1GMC0YfXt1dPXv_uRj7LqAvKyVgluMn-EjFyWIHERVNJfsaT_Z6DGFaeR9XNPCw8Q7nMYVR88Ps_OnxJdjnNfx-FPPX_yaPJ-Jv_oYKHjH2xDtGpa0ZReEp-Sv_rth_f1d3z5m3fPDvt11GVa6yXQhDDTWgiGUoCpHEsA5JCsIha5LRNQKKmlImFJaIHK-IaOxRqGslBt287c9Y4b3GN4wfg2_qOGMkt9kjUhM</recordid><startdate>20240204</startdate><enddate>20240204</enddate><creator>Quirke, Philip</creator><creator>Neo, Clement</creator><creator>Barez, Fazl</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20240204</creationdate><title>Increasing Trust in Language Models through the Reuse of Verified Circuits</title><author>Quirke, Philip ; Neo, Clement ; Barez, Fazl</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a679-712b09cc0bfa3056df300ddafc2fa2784aaa75063bf2b43c0ffde9fb7a8a25c33</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Computer Science - Computation and Language</topic><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Quirke, Philip</creatorcontrib><creatorcontrib>Neo, Clement</creatorcontrib><creatorcontrib>Barez, Fazl</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Quirke, Philip</au><au>Neo, Clement</au><au>Barez, Fazl</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Increasing Trust in Language Models through the Reuse of Verified Circuits</atitle><date>2024-02-04</date><risdate>2024</risdate><abstract>Language Models (LMs) are increasingly used for a wide range of prediction tasks, but their training can often neglect rare edge cases, reducing their reliability. Here, we define a stringent standard of trustworthiness whereby the task algorithm and circuit implementation must be verified, accounting for edge cases, with no known failure modes. We show that a model can be trained to meet this standard if built using mathematically and logically specified frameworks. In this paper, we fully verify an auto-regressive transformer model for n-digit integer addition. To exhibit the reusability of verified modules, we insert the trained integer addition model into a larger untrained model and train the combined model to perform both addition and subtraction. We find extensive reuse of the addition circuits for both tasks, easing verification of the more complex subtractor model. We discuss how inserting verified task modules into LMs can leverage model reuse to improve verifiability and trustworthiness of language models built using them. The reuse of verified circuits reduces the effort to verify more complex composite models which we believe to be a significant step towards safety of language models.</abstract><doi>10.48550/arxiv.2402.02619</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2402.02619
ispartof
issn
language eng
recordid cdi_arxiv_primary_2402_02619
source arXiv.org
subjects Computer Science - Computation and Language
Computer Science - Learning
title Increasing Trust in Language Models through the Reuse of Verified Circuits
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-03T07%3A51%3A25IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Increasing%20Trust%20in%20Language%20Models%20through%20the%20Reuse%20of%20Verified%20Circuits&rft.au=Quirke,%20Philip&rft.date=2024-02-04&rft_id=info:doi/10.48550/arxiv.2402.02619&rft_dat=%3Carxiv_GOX%3E2402_02619%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true