FACTTRACK: Time-Aware World State Tracking in Story Outlines

While accurately detecting and correcting factual contradictions in language model outputs has become increasingly important as their capabilities improve, doing so is highly challenging. We propose a novel method, FACTTRACK, for tracking atomic facts and addressing factual contradictions. Crucially...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2024-07
Hauptverfasser: Lyu, Zhiheng, Yang, Kevin, Kong, Lingpeng, Klein, Daniel
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 Lyu, Zhiheng
Yang, Kevin
Kong, Lingpeng
Klein, Daniel
description While accurately detecting and correcting factual contradictions in language model outputs has become increasingly important as their capabilities improve, doing so is highly challenging. We propose a novel method, FACTTRACK, for tracking atomic facts and addressing factual contradictions. Crucially, FACTTRACK also maintains time-aware validity intervals for each fact, allowing for change over time. At a high level, FACTTRACK consists of a four-step pipeline to update a world state data structure for each new event: (1) decompose the event into directional atomic facts; (2) determine the validity interval of each atomic fact using the world state; (3) detect contradictions with existing facts in the world state; and finally (4) add new facts to the world state and update existing atomic facts. When we apply FACTTRACK to contradiction detection on structured story outlines, we find that FACTTRACK using LLaMA2-7B-Chat substantially outperforms a fair baseline using LLaMA2-7B-Chat, and achieves performance comparable to a GPT4 baseline. Moreover, when using GPT4, FACTTRACK significantly outperforms the GPT4 baseline.
format Article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_3084093661</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>3084093661</sourcerecordid><originalsourceid>FETCH-proquest_journals_30840936613</originalsourceid><addsrcrecordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mSwcXN0DgkJcnT2tlIIycxN1XUsTyxKVQjPL8pJUQguSSxJVQgpSkzOzsxLV8jMA4rkF1Uq-JeW5GTmpRbzMLCmJeYUp_JCaW4GZTfXEGcP3YKi_MLS1OKS-Kz80qI8oFS8sYGFiYGlsZmZoTFxqgBUjjXx</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>3084093661</pqid></control><display><type>article</type><title>FACTTRACK: Time-Aware World State Tracking in Story Outlines</title><source>Free E- Journals</source><creator>Lyu, Zhiheng ; Yang, Kevin ; Kong, Lingpeng ; Klein, Daniel</creator><creatorcontrib>Lyu, Zhiheng ; Yang, Kevin ; Kong, Lingpeng ; Klein, Daniel</creatorcontrib><description>While accurately detecting and correcting factual contradictions in language model outputs has become increasingly important as their capabilities improve, doing so is highly challenging. We propose a novel method, FACTTRACK, for tracking atomic facts and addressing factual contradictions. Crucially, FACTTRACK also maintains time-aware validity intervals for each fact, allowing for change over time. At a high level, FACTTRACK consists of a four-step pipeline to update a world state data structure for each new event: (1) decompose the event into directional atomic facts; (2) determine the validity interval of each atomic fact using the world state; (3) detect contradictions with existing facts in the world state; and finally (4) add new facts to the world state and update existing atomic facts. When we apply FACTTRACK to contradiction detection on structured story outlines, we find that FACTTRACK using LLaMA2-7B-Chat substantially outperforms a fair baseline using LLaMA2-7B-Chat, and achieves performance comparable to a GPT4 baseline. Moreover, when using GPT4, FACTTRACK significantly outperforms the GPT4 baseline.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Atomic structure ; Data structures ; Tracking</subject><ispartof>arXiv.org, 2024-07</ispartof><rights>2024. This work is published under http://creativecommons.org/publicdomain/zero/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>776,780</link.rule.ids></links><search><creatorcontrib>Lyu, Zhiheng</creatorcontrib><creatorcontrib>Yang, Kevin</creatorcontrib><creatorcontrib>Kong, Lingpeng</creatorcontrib><creatorcontrib>Klein, Daniel</creatorcontrib><title>FACTTRACK: Time-Aware World State Tracking in Story Outlines</title><title>arXiv.org</title><description>While accurately detecting and correcting factual contradictions in language model outputs has become increasingly important as their capabilities improve, doing so is highly challenging. We propose a novel method, FACTTRACK, for tracking atomic facts and addressing factual contradictions. Crucially, FACTTRACK also maintains time-aware validity intervals for each fact, allowing for change over time. At a high level, FACTTRACK consists of a four-step pipeline to update a world state data structure for each new event: (1) decompose the event into directional atomic facts; (2) determine the validity interval of each atomic fact using the world state; (3) detect contradictions with existing facts in the world state; and finally (4) add new facts to the world state and update existing atomic facts. When we apply FACTTRACK to contradiction detection on structured story outlines, we find that FACTTRACK using LLaMA2-7B-Chat substantially outperforms a fair baseline using LLaMA2-7B-Chat, and achieves performance comparable to a GPT4 baseline. Moreover, when using GPT4, FACTTRACK significantly outperforms the GPT4 baseline.</description><subject>Atomic structure</subject><subject>Data structures</subject><subject>Tracking</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>BENPR</sourceid><recordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mSwcXN0DgkJcnT2tlIIycxN1XUsTyxKVQjPL8pJUQguSSxJVQgpSkzOzsxLV8jMA4rkF1Uq-JeW5GTmpRbzMLCmJeYUp_JCaW4GZTfXEGcP3YKi_MLS1OKS-Kz80qI8oFS8sYGFiYGlsZmZoTFxqgBUjjXx</recordid><startdate>20240723</startdate><enddate>20240723</enddate><creator>Lyu, Zhiheng</creator><creator>Yang, Kevin</creator><creator>Kong, Lingpeng</creator><creator>Klein, Daniel</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>PHGZM</scope><scope>PHGZT</scope><scope>PIMPY</scope><scope>PKEHL</scope><scope>PQEST</scope><scope>PQGLB</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20240723</creationdate><title>FACTTRACK: Time-Aware World State Tracking in Story Outlines</title><author>Lyu, Zhiheng ; Yang, Kevin ; Kong, Lingpeng ; Klein, Daniel</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_30840936613</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Atomic structure</topic><topic>Data structures</topic><topic>Tracking</topic><toplevel>online_resources</toplevel><creatorcontrib>Lyu, Zhiheng</creatorcontrib><creatorcontrib>Yang, Kevin</creatorcontrib><creatorcontrib>Kong, Lingpeng</creatorcontrib><creatorcontrib>Klein, Daniel</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>ProQuest Central (New)</collection><collection>ProQuest One Academic (New)</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Middle East (New)</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Applied &amp; Life Sciences</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>Lyu, Zhiheng</au><au>Yang, Kevin</au><au>Kong, Lingpeng</au><au>Klein, Daniel</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>FACTTRACK: Time-Aware World State Tracking in Story Outlines</atitle><jtitle>arXiv.org</jtitle><date>2024-07-23</date><risdate>2024</risdate><eissn>2331-8422</eissn><abstract>While accurately detecting and correcting factual contradictions in language model outputs has become increasingly important as their capabilities improve, doing so is highly challenging. We propose a novel method, FACTTRACK, for tracking atomic facts and addressing factual contradictions. Crucially, FACTTRACK also maintains time-aware validity intervals for each fact, allowing for change over time. At a high level, FACTTRACK consists of a four-step pipeline to update a world state data structure for each new event: (1) decompose the event into directional atomic facts; (2) determine the validity interval of each atomic fact using the world state; (3) detect contradictions with existing facts in the world state; and finally (4) add new facts to the world state and update existing atomic facts. When we apply FACTTRACK to contradiction detection on structured story outlines, we find that FACTTRACK using LLaMA2-7B-Chat substantially outperforms a fair baseline using LLaMA2-7B-Chat, and achieves performance comparable to a GPT4 baseline. Moreover, when using GPT4, FACTTRACK significantly outperforms the GPT4 baseline.</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, 2024-07
issn 2331-8422
language eng
recordid cdi_proquest_journals_3084093661
source Free E- Journals
subjects Atomic structure
Data structures
Tracking
title FACTTRACK: Time-Aware World State Tracking in Story Outlines
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-19T16%3A26%3A38IST&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=FACTTRACK:%20Time-Aware%20World%20State%20Tracking%20in%20Story%20Outlines&rft.jtitle=arXiv.org&rft.au=Lyu,%20Zhiheng&rft.date=2024-07-23&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E3084093661%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=3084093661&rft_id=info:pmid/&rfr_iscdi=true