Characterizing and addressing the issue of oversmoothing in neural autoregressive sequence modeling

Neural autoregressive sequence models smear the probability among many possible sequences including degenerate ones, such as empty or repetitive sequences. In this work, we tackle one specific case where the model assigns a high probability to unreasonably short sequences. We define the oversmoothin...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kulikov, Ilia, Eremeev, Maksim, Cho, Kyunghyun
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 Kulikov, Ilia
Eremeev, Maksim
Cho, Kyunghyun
description Neural autoregressive sequence models smear the probability among many possible sequences including degenerate ones, such as empty or repetitive sequences. In this work, we tackle one specific case where the model assigns a high probability to unreasonably short sequences. We define the oversmoothing rate to quantify this issue. After confirming the high degree of oversmoothing in neural machine translation, we propose to explicitly minimize the oversmoothing rate during training. We conduct a set of experiments to study the effect of the proposed regularization on both model distribution and decoding performance. We use a neural machine translation task as the testbed and consider three different datasets of varying size. Our experiments reveal three major findings. First, we can control the oversmoothing rate of the model by tuning the strength of the regularization. Second, by enhancing the oversmoothing loss contribution, the probability and the rank of token decrease heavily at positions where it is not supposed to be. Third, the proposed regularization impacts the outcome of beam search especially when a large beam is used. The degradation of translation quality (measured in BLEU) with a large beam significantly lessens with lower oversmoothing rate, but the degradation compared to smaller beam sizes remains to exist. From these observations, we conclude that the high degree of oversmoothing is the main reason behind the degenerate case of overly probable short sequences in a neural autoregressive model.
doi_str_mv 10.48550/arxiv.2112.08914
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2112_08914</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2112_08914</sourcerecordid><originalsourceid>FETCH-LOGICAL-a674-f246aaba137e1701eb3b1ffdb785c872c4a69f558169ba33fe69342dcd6871d83</originalsourceid><addsrcrecordid>eNotj8tOhDAYhbtxYUYfwNX0BUBKS1uWhnhLJnEze_LT_h2aANUWiPr0Cro6OTmX5CPkjhW50FVV3EP89GteMlbmha6ZuCam6SGCmTH6bz9dKEyWgrURU9rs3CP1KS1Ig6NhxZjGEOZ-i_xEJ1wiDBSWOUS87JsVacKPBSeDdAwWh9_qDblyMCS8_dcDOT89npuX7PT2_No8nDKQSmSuFBKgA8YVMlUw7HjHnLOd0pXRqjQCZO2qSjNZd8C5Q1lzUVpjpVbMan4gx7_bnbJ9j36E-NVutO1Oy38AK6ZSVQ</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Characterizing and addressing the issue of oversmoothing in neural autoregressive sequence modeling</title><source>arXiv.org</source><creator>Kulikov, Ilia ; Eremeev, Maksim ; Cho, Kyunghyun</creator><creatorcontrib>Kulikov, Ilia ; Eremeev, Maksim ; Cho, Kyunghyun</creatorcontrib><description>Neural autoregressive sequence models smear the probability among many possible sequences including degenerate ones, such as empty or repetitive sequences. In this work, we tackle one specific case where the model assigns a high probability to unreasonably short sequences. We define the oversmoothing rate to quantify this issue. After confirming the high degree of oversmoothing in neural machine translation, we propose to explicitly minimize the oversmoothing rate during training. We conduct a set of experiments to study the effect of the proposed regularization on both model distribution and decoding performance. We use a neural machine translation task as the testbed and consider three different datasets of varying size. Our experiments reveal three major findings. First, we can control the oversmoothing rate of the model by tuning the strength of the regularization. Second, by enhancing the oversmoothing loss contribution, the probability and the rank of token decrease heavily at positions where it is not supposed to be. Third, the proposed regularization impacts the outcome of beam search especially when a large beam is used. The degradation of translation quality (measured in BLEU) with a large beam significantly lessens with lower oversmoothing rate, but the degradation compared to smaller beam sizes remains to exist. From these observations, we conclude that the high degree of oversmoothing is the main reason behind the degenerate case of overly probable short sequences in a neural autoregressive model.</description><identifier>DOI: 10.48550/arxiv.2112.08914</identifier><language>eng</language><subject>Computer Science - Computation and Language ; Computer Science - Learning</subject><creationdate>2021-12</creationdate><rights>http://creativecommons.org/licenses/by/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/2112.08914$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2112.08914$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Kulikov, Ilia</creatorcontrib><creatorcontrib>Eremeev, Maksim</creatorcontrib><creatorcontrib>Cho, Kyunghyun</creatorcontrib><title>Characterizing and addressing the issue of oversmoothing in neural autoregressive sequence modeling</title><description>Neural autoregressive sequence models smear the probability among many possible sequences including degenerate ones, such as empty or repetitive sequences. In this work, we tackle one specific case where the model assigns a high probability to unreasonably short sequences. We define the oversmoothing rate to quantify this issue. After confirming the high degree of oversmoothing in neural machine translation, we propose to explicitly minimize the oversmoothing rate during training. We conduct a set of experiments to study the effect of the proposed regularization on both model distribution and decoding performance. We use a neural machine translation task as the testbed and consider three different datasets of varying size. Our experiments reveal three major findings. First, we can control the oversmoothing rate of the model by tuning the strength of the regularization. Second, by enhancing the oversmoothing loss contribution, the probability and the rank of token decrease heavily at positions where it is not supposed to be. Third, the proposed regularization impacts the outcome of beam search especially when a large beam is used. The degradation of translation quality (measured in BLEU) with a large beam significantly lessens with lower oversmoothing rate, but the degradation compared to smaller beam sizes remains to exist. From these observations, we conclude that the high degree of oversmoothing is the main reason behind the degenerate case of overly probable short sequences in a neural autoregressive model.</description><subject>Computer Science - Computation and Language</subject><subject>Computer Science - Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj8tOhDAYhbtxYUYfwNX0BUBKS1uWhnhLJnEze_LT_h2aANUWiPr0Cro6OTmX5CPkjhW50FVV3EP89GteMlbmha6ZuCam6SGCmTH6bz9dKEyWgrURU9rs3CP1KS1Ig6NhxZjGEOZ-i_xEJ1wiDBSWOUS87JsVacKPBSeDdAwWh9_qDblyMCS8_dcDOT89npuX7PT2_No8nDKQSmSuFBKgA8YVMlUw7HjHnLOd0pXRqjQCZO2qSjNZd8C5Q1lzUVpjpVbMan4gx7_bnbJ9j36E-NVutO1Oy38AK6ZSVQ</recordid><startdate>20211216</startdate><enddate>20211216</enddate><creator>Kulikov, Ilia</creator><creator>Eremeev, Maksim</creator><creator>Cho, Kyunghyun</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20211216</creationdate><title>Characterizing and addressing the issue of oversmoothing in neural autoregressive sequence modeling</title><author>Kulikov, Ilia ; Eremeev, Maksim ; Cho, Kyunghyun</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a674-f246aaba137e1701eb3b1ffdb785c872c4a69f558169ba33fe69342dcd6871d83</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Computer Science - Computation and Language</topic><topic>Computer Science - Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Kulikov, Ilia</creatorcontrib><creatorcontrib>Eremeev, Maksim</creatorcontrib><creatorcontrib>Cho, Kyunghyun</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Kulikov, Ilia</au><au>Eremeev, Maksim</au><au>Cho, Kyunghyun</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Characterizing and addressing the issue of oversmoothing in neural autoregressive sequence modeling</atitle><date>2021-12-16</date><risdate>2021</risdate><abstract>Neural autoregressive sequence models smear the probability among many possible sequences including degenerate ones, such as empty or repetitive sequences. In this work, we tackle one specific case where the model assigns a high probability to unreasonably short sequences. We define the oversmoothing rate to quantify this issue. After confirming the high degree of oversmoothing in neural machine translation, we propose to explicitly minimize the oversmoothing rate during training. We conduct a set of experiments to study the effect of the proposed regularization on both model distribution and decoding performance. We use a neural machine translation task as the testbed and consider three different datasets of varying size. Our experiments reveal three major findings. First, we can control the oversmoothing rate of the model by tuning the strength of the regularization. Second, by enhancing the oversmoothing loss contribution, the probability and the rank of token decrease heavily at positions where it is not supposed to be. Third, the proposed regularization impacts the outcome of beam search especially when a large beam is used. The degradation of translation quality (measured in BLEU) with a large beam significantly lessens with lower oversmoothing rate, but the degradation compared to smaller beam sizes remains to exist. From these observations, we conclude that the high degree of oversmoothing is the main reason behind the degenerate case of overly probable short sequences in a neural autoregressive model.</abstract><doi>10.48550/arxiv.2112.08914</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2112.08914
ispartof
issn
language eng
recordid cdi_arxiv_primary_2112_08914
source arXiv.org
subjects Computer Science - Computation and Language
Computer Science - Learning
title Characterizing and addressing the issue of oversmoothing in neural autoregressive sequence modeling
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-25T12%3A52%3A20IST&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=Characterizing%20and%20addressing%20the%20issue%20of%20oversmoothing%20in%20neural%20autoregressive%20sequence%20modeling&rft.au=Kulikov,%20Ilia&rft.date=2021-12-16&rft_id=info:doi/10.48550/arxiv.2112.08914&rft_dat=%3Carxiv_GOX%3E2112_08914%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