Detailed Information

Cited 0 time in webofscience Cited 0 time in scopus
Metadata Downloads

Enhancing Out-of-Distribution Detection in Natural Language Understanding via Implicit Layer Ensemble

Full metadata record
DC Field Value Language
dc.contributor.authorCho, Hyunsoo-
dc.contributor.authorPark, Choonghyun-
dc.contributor.authorKang, Jaewook-
dc.contributor.authorYoo, Kang Min-
dc.contributor.authorKim, Tae Uk-
dc.contributor.authorLee, Sang-goo-
dc.date.accessioned2023-08-01T07:08:12Z-
dc.date.available2023-08-01T07:08:12Z-
dc.date.created2023-07-21-
dc.date.issued2022-12-
dc.identifier.urihttps://scholarworks.bwise.kr/hanyang/handle/2021.sw.hanyang/188683-
dc.description.abstractOut-of-distribution (OOD) detection aims todiscern outliers from the intended data distribution, which is crucial to maintaining high reliability and a good user experience. Most recentstudies in OOD detection utilize the information from a single representation that residesin the penultimate layer to determine whetherthe input is anomalous or not. Although sucha method is straightforward, the potential ofdiverse information in the intermediate layersis overlooked. In this paper, we propose a novelframework based on contrastive learning thatencourages intermediate features to learn layerspecialized representations and assembles themimplicitly into a single representation to absorbrich information in the pre-trained languagemodel. Extensive experiments in various intentclassification and OOD datasets demonstratethat our approach is significantly more effective than other works. The source code for ourmodel is available online.-
dc.language영어-
dc.language.isoen-
dc.publisherAssociation for Computational Linguistics-
dc.titleEnhancing Out-of-Distribution Detection in Natural Language Understanding via Implicit Layer Ensemble-
dc.typeArticle-
dc.contributor.affiliatedAuthorKim, Tae Uk-
dc.identifier.bibliographicCitationEmpirical Methods in Natural Language Processing, pp.783 - 798-
dc.relation.isPartOfEmpirical Methods in Natural Language Processing-
dc.citation.titleEmpirical Methods in Natural Language Processing-
dc.citation.startPage783-
dc.citation.endPage798-
dc.type.rimsART-
dc.type.docTypeProceeding-
dc.description.journalClass1-
dc.description.isOpenAccessN-
dc.description.journalRegisteredClassother-
dc.identifier.urlhttps://aclanthology.org/2022.findings-emnlp.55.pdf-
Files in This Item
Go to Link
Appears in
Collections
서울 공과대학 > 서울 컴퓨터소프트웨어학부 > 1. Journal Articles

qrcode

Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.

Related Researcher

Researcher Kim, Taeuk photo

Kim, Taeuk
COLLEGE OF ENGINEERING (SCHOOL OF COMPUTER SCIENCE)
Read more

Altmetrics

Total Views & Downloads

BROWSE