BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage

We present BlenderBot 3, a 175B parameter dialogue model capable of open-domain conversation with access to the internet and a long-term memory, and having been trained on a large number of user defined tasks. We release both the model weights and code, and have also deployed the model on a public w...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Shuster, Kurt, Xu, Jing, Komeili, Mojtaba, Ju, Da, Smith, Eric Michael, Roller, Stephen, Ung, Megan, Chen, Moya, Arora, Kushal, Lane, Joshua, Behrooz, Morteza, Ngan, William, Poff, Spencer, Goyal, Naman, Szlam, Arthur, Boureau, Y-Lan, Kambadur, Melanie, Weston, Jason
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 Shuster, Kurt
Xu, Jing
Komeili, Mojtaba
Ju, Da
Smith, Eric Michael
Roller, Stephen
Ung, Megan
Chen, Moya
Arora, Kushal
Lane, Joshua
Behrooz, Morteza
Ngan, William
Poff, Spencer
Goyal, Naman
Szlam, Arthur
Boureau, Y-Lan
Kambadur, Melanie
Weston, Jason
description We present BlenderBot 3, a 175B parameter dialogue model capable of open-domain conversation with access to the internet and a long-term memory, and having been trained on a large number of user defined tasks. We release both the model weights and code, and have also deployed the model on a public web page to interact with organic users. This technical report describes how the model was built (architecture, model and training scheme), and details of its deployment, including safety mechanisms. Human evaluations show its superiority to existing open-domain dialogue agents, including its predecessors (Roller et al., 2021; Komeili et al., 2022). Finally, we detail our plan for continual learning using the data collected from deployment, which will also be publicly released. The goal of this research program is thus to enable the community to study ever-improving responsible agents that learn through interaction.
doi_str_mv 10.48550/arxiv.2208.03188
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2208_03188</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2208_03188</sourcerecordid><originalsourceid>FETCH-LOGICAL-a678-eaf2cdccaaf9fc851fad25f43d4ac373078df3be1ea2fb9cf1e8fa8e2099228b3</originalsourceid><addsrcrecordid>eNotj8tOwzAURL1hgQofwAr_QIIfCXHY0YqXVIlNdyyiG_veEsnYkW0q8ve0hdVIM0cjHcZupKgb07biDtLPdKiVEqYWWhpzyT7WHoPDtI6F6wcO3OHs44KO2xgOmDKUKQbwHPYYCi-fUE5LmcI3eL9wj5BC5iXyhHmOIU_jscWwP_JX7ILAZ7z-zxXbPT_tNq_V9v3lbfO4reC-MxUCKeusBaCerGklgVMtNdo1YHWnRWcc6RElgqKxtyTREBhUou-VMqNesdu_27PdMKfpC9IynCyHs6X-BfthUBE</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage</title><source>arXiv.org</source><creator>Shuster, Kurt ; Xu, Jing ; Komeili, Mojtaba ; Ju, Da ; Smith, Eric Michael ; Roller, Stephen ; Ung, Megan ; Chen, Moya ; Arora, Kushal ; Lane, Joshua ; Behrooz, Morteza ; Ngan, William ; Poff, Spencer ; Goyal, Naman ; Szlam, Arthur ; Boureau, Y-Lan ; Kambadur, Melanie ; Weston, Jason</creator><creatorcontrib>Shuster, Kurt ; Xu, Jing ; Komeili, Mojtaba ; Ju, Da ; Smith, Eric Michael ; Roller, Stephen ; Ung, Megan ; Chen, Moya ; Arora, Kushal ; Lane, Joshua ; Behrooz, Morteza ; Ngan, William ; Poff, Spencer ; Goyal, Naman ; Szlam, Arthur ; Boureau, Y-Lan ; Kambadur, Melanie ; Weston, Jason</creatorcontrib><description>We present BlenderBot 3, a 175B parameter dialogue model capable of open-domain conversation with access to the internet and a long-term memory, and having been trained on a large number of user defined tasks. We release both the model weights and code, and have also deployed the model on a public web page to interact with organic users. This technical report describes how the model was built (architecture, model and training scheme), and details of its deployment, including safety mechanisms. Human evaluations show its superiority to existing open-domain dialogue agents, including its predecessors (Roller et al., 2021; Komeili et al., 2022). Finally, we detail our plan for continual learning using the data collected from deployment, which will also be publicly released. The goal of this research program is thus to enable the community to study ever-improving responsible agents that learn through interaction.</description><identifier>DOI: 10.48550/arxiv.2208.03188</identifier><language>eng</language><subject>Computer Science - Artificial Intelligence ; Computer Science - Computation and Language</subject><creationdate>2022-08</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</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/2208.03188$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2208.03188$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Shuster, Kurt</creatorcontrib><creatorcontrib>Xu, Jing</creatorcontrib><creatorcontrib>Komeili, Mojtaba</creatorcontrib><creatorcontrib>Ju, Da</creatorcontrib><creatorcontrib>Smith, Eric Michael</creatorcontrib><creatorcontrib>Roller, Stephen</creatorcontrib><creatorcontrib>Ung, Megan</creatorcontrib><creatorcontrib>Chen, Moya</creatorcontrib><creatorcontrib>Arora, Kushal</creatorcontrib><creatorcontrib>Lane, Joshua</creatorcontrib><creatorcontrib>Behrooz, Morteza</creatorcontrib><creatorcontrib>Ngan, William</creatorcontrib><creatorcontrib>Poff, Spencer</creatorcontrib><creatorcontrib>Goyal, Naman</creatorcontrib><creatorcontrib>Szlam, Arthur</creatorcontrib><creatorcontrib>Boureau, Y-Lan</creatorcontrib><creatorcontrib>Kambadur, Melanie</creatorcontrib><creatorcontrib>Weston, Jason</creatorcontrib><title>BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage</title><description>We present BlenderBot 3, a 175B parameter dialogue model capable of open-domain conversation with access to the internet and a long-term memory, and having been trained on a large number of user defined tasks. We release both the model weights and code, and have also deployed the model on a public web page to interact with organic users. This technical report describes how the model was built (architecture, model and training scheme), and details of its deployment, including safety mechanisms. Human evaluations show its superiority to existing open-domain dialogue agents, including its predecessors (Roller et al., 2021; Komeili et al., 2022). Finally, we detail our plan for continual learning using the data collected from deployment, which will also be publicly released. The goal of this research program is thus to enable the community to study ever-improving responsible agents that learn through interaction.</description><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Computation and Language</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2022</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8tOwzAURL1hgQofwAr_QIIfCXHY0YqXVIlNdyyiG_veEsnYkW0q8ve0hdVIM0cjHcZupKgb07biDtLPdKiVEqYWWhpzyT7WHoPDtI6F6wcO3OHs44KO2xgOmDKUKQbwHPYYCi-fUE5LmcI3eL9wj5BC5iXyhHmOIU_jscWwP_JX7ILAZ7z-zxXbPT_tNq_V9v3lbfO4reC-MxUCKeusBaCerGklgVMtNdo1YHWnRWcc6RElgqKxtyTREBhUou-VMqNesdu_27PdMKfpC9IynCyHs6X-BfthUBE</recordid><startdate>20220805</startdate><enddate>20220805</enddate><creator>Shuster, Kurt</creator><creator>Xu, Jing</creator><creator>Komeili, Mojtaba</creator><creator>Ju, Da</creator><creator>Smith, Eric Michael</creator><creator>Roller, Stephen</creator><creator>Ung, Megan</creator><creator>Chen, Moya</creator><creator>Arora, Kushal</creator><creator>Lane, Joshua</creator><creator>Behrooz, Morteza</creator><creator>Ngan, William</creator><creator>Poff, Spencer</creator><creator>Goyal, Naman</creator><creator>Szlam, Arthur</creator><creator>Boureau, Y-Lan</creator><creator>Kambadur, Melanie</creator><creator>Weston, Jason</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20220805</creationdate><title>BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage</title><author>Shuster, Kurt ; Xu, Jing ; Komeili, Mojtaba ; Ju, Da ; Smith, Eric Michael ; Roller, Stephen ; Ung, Megan ; Chen, Moya ; Arora, Kushal ; Lane, Joshua ; Behrooz, Morteza ; Ngan, William ; Poff, Spencer ; Goyal, Naman ; Szlam, Arthur ; Boureau, Y-Lan ; Kambadur, Melanie ; Weston, Jason</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a678-eaf2cdccaaf9fc851fad25f43d4ac373078df3be1ea2fb9cf1e8fa8e2099228b3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2022</creationdate><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Computation and Language</topic><toplevel>online_resources</toplevel><creatorcontrib>Shuster, Kurt</creatorcontrib><creatorcontrib>Xu, Jing</creatorcontrib><creatorcontrib>Komeili, Mojtaba</creatorcontrib><creatorcontrib>Ju, Da</creatorcontrib><creatorcontrib>Smith, Eric Michael</creatorcontrib><creatorcontrib>Roller, Stephen</creatorcontrib><creatorcontrib>Ung, Megan</creatorcontrib><creatorcontrib>Chen, Moya</creatorcontrib><creatorcontrib>Arora, Kushal</creatorcontrib><creatorcontrib>Lane, Joshua</creatorcontrib><creatorcontrib>Behrooz, Morteza</creatorcontrib><creatorcontrib>Ngan, William</creatorcontrib><creatorcontrib>Poff, Spencer</creatorcontrib><creatorcontrib>Goyal, Naman</creatorcontrib><creatorcontrib>Szlam, Arthur</creatorcontrib><creatorcontrib>Boureau, Y-Lan</creatorcontrib><creatorcontrib>Kambadur, Melanie</creatorcontrib><creatorcontrib>Weston, Jason</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Shuster, Kurt</au><au>Xu, Jing</au><au>Komeili, Mojtaba</au><au>Ju, Da</au><au>Smith, Eric Michael</au><au>Roller, Stephen</au><au>Ung, Megan</au><au>Chen, Moya</au><au>Arora, Kushal</au><au>Lane, Joshua</au><au>Behrooz, Morteza</au><au>Ngan, William</au><au>Poff, Spencer</au><au>Goyal, Naman</au><au>Szlam, Arthur</au><au>Boureau, Y-Lan</au><au>Kambadur, Melanie</au><au>Weston, Jason</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage</atitle><date>2022-08-05</date><risdate>2022</risdate><abstract>We present BlenderBot 3, a 175B parameter dialogue model capable of open-domain conversation with access to the internet and a long-term memory, and having been trained on a large number of user defined tasks. We release both the model weights and code, and have also deployed the model on a public web page to interact with organic users. This technical report describes how the model was built (architecture, model and training scheme), and details of its deployment, including safety mechanisms. Human evaluations show its superiority to existing open-domain dialogue agents, including its predecessors (Roller et al., 2021; Komeili et al., 2022). Finally, we detail our plan for continual learning using the data collected from deployment, which will also be publicly released. The goal of this research program is thus to enable the community to study ever-improving responsible agents that learn through interaction.</abstract><doi>10.48550/arxiv.2208.03188</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2208.03188
ispartof
issn
language eng
recordid cdi_arxiv_primary_2208_03188
source arXiv.org
subjects Computer Science - Artificial Intelligence
Computer Science - Computation and Language
title BlenderBot 3: a deployed conversational agent that continually learns to responsibly engage
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-14T09%3A46%3A52IST&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=BlenderBot%203:%20a%20deployed%20conversational%20agent%20that%20continually%20learns%20to%20responsibly%20engage&rft.au=Shuster,%20Kurt&rft.date=2022-08-05&rft_id=info:doi/10.48550/arxiv.2208.03188&rft_dat=%3Carxiv_GOX%3E2208_03188%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