{"id":"https://openalex.org/W2888169584","doi":"https://doi.org/10.18653/v1/w18-5426","title":"Under the Hood: Using Diagnostic Classifiers to Investigate and Improve how Language Models Track Agreement Information","display_name":"Under the Hood: Using Diagnostic Classifiers to Investigate and Improve how Language Models Track Agreement Information","publication_year":2018,"publication_date":"2018-01-01","ids":{"openalex":"https://openalex.org/W2888169584","doi":"https://doi.org/10.18653/v1/w18-5426","mag":"2888169584"},"language":"en","primary_location":{"id":"doi:10.18653/v1/w18-5426","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/w18-5426","pdf_url":"https://www.aclweb.org/anthology/W18-5426.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP","raw_type":"proceedings-article"},"type":"preprint","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://www.aclweb.org/anthology/W18-5426.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5013040086","display_name":"Mario Giulianelli","orcid":"https://orcid.org/0009-0004-1281-9686"},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]}],"countries":["NL"],"is_corresponding":true,"raw_author_name":"Mario Giulianelli","raw_affiliation_strings":["University of Amsterdam"],"affiliations":[{"raw_affiliation_string":"University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060073175","display_name":"Jacqueline Harding","orcid":"https://orcid.org/0000-0002-2239-7817"},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]},{"id":"https://openalex.org/I97018004","display_name":"Stanford University","ror":"https://ror.org/00f54p054","country_code":"US","type":"education","lineage":["https://openalex.org/I97018004"]}],"countries":["NL","US"],"is_corresponding":false,"raw_author_name":"Jack Harding","raw_affiliation_strings":["University of Amsterdam","Stanford University ()"],"affiliations":[{"raw_affiliation_string":"University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]},{"raw_affiliation_string":"Stanford University ()","institution_ids":["https://openalex.org/I97018004"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047891775","display_name":"Florian Mohnert","orcid":null},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Florian Mohnert","raw_affiliation_strings":["University of Amsterdam"],"affiliations":[{"raw_affiliation_string":"University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018069499","display_name":"Dieuwke Hupkes","orcid":null},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Dieuwke Hupkes","raw_affiliation_strings":["ILLC, University of Amsterdam"],"affiliations":[{"raw_affiliation_string":"ILLC, University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5007928903","display_name":"Willem Zuidema","orcid":"https://orcid.org/0000-0002-2362-5447"},"institutions":[{"id":"https://openalex.org/I887064364","display_name":"University of Amsterdam","ror":"https://ror.org/04dkp9463","country_code":"NL","type":"education","lineage":["https://openalex.org/I887064364"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Willem Zuidema","raw_affiliation_strings":["ILLC, University of Amsterdam","University of Amsterdam"],"affiliations":[{"raw_affiliation_string":"ILLC, University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]},{"raw_affiliation_string":"University of Amsterdam","institution_ids":["https://openalex.org/I887064364"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5013040086"],"corresponding_institution_ids":["https://openalex.org/I887064364"],"apc_list":null,"apc_paid":null,"fwci":0.5077,"has_fulltext":true,"cited_by_count":5,"citation_normalized_percentile":{"value":0.74038519,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"240","last_page":"248"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9998999834060669,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9998999834060669,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9998000264167786,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T13629","display_name":"Text Readability and Simplification","score":0.9994000196456909,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/agreement","display_name":"Agreement","score":0.7950094938278198},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.704582929611206},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.6390352845191956},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.6320253610610962},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.5976505279541016},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.5637481212615967},{"id":"https://openalex.org/keywords/track","display_name":"Track (disk drive)","score":0.5274113416671753},{"id":"https://openalex.org/keywords/subject","display_name":"Subject (documents)","score":0.4286738634109497},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.3705483078956604},{"id":"https://openalex.org/keywords/linguistics","display_name":"Linguistics","score":0.27043092250823975}],"concepts":[{"id":"https://openalex.org/C2776818064","wikidata":"https://www.wikidata.org/wiki/Q829903","display_name":"Agreement","level":2,"score":0.7950094938278198},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.704582929611206},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6390352845191956},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.6320253610610962},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.5976505279541016},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5637481212615967},{"id":"https://openalex.org/C89992363","wikidata":"https://www.wikidata.org/wiki/Q5961558","display_name":"Track (disk drive)","level":2,"score":0.5274113416671753},{"id":"https://openalex.org/C2777855551","wikidata":"https://www.wikidata.org/wiki/Q12310021","display_name":"Subject (documents)","level":2,"score":0.4286738634109497},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3705483078956604},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.27043092250823975},{"id":"https://openalex.org/C199539241","wikidata":"https://www.wikidata.org/wiki/Q7748","display_name":"Law","level":1,"score":0.0},{"id":"https://openalex.org/C94625758","wikidata":"https://www.wikidata.org/wiki/Q7163","display_name":"Politics","level":2,"score":0.0},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0},{"id":"https://openalex.org/C161191863","wikidata":"https://www.wikidata.org/wiki/Q199655","display_name":"Library science","level":1,"score":0.0},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.0},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0}],"mesh":[],"locations_count":8,"locations":[{"id":"doi:10.18653/v1/w18-5426","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/w18-5426","pdf_url":"https://www.aclweb.org/anthology/W18-5426.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:1808.08079","is_oa":true,"landing_page_url":"http://arxiv.org/abs/1808.08079","pdf_url":"https://arxiv.org/pdf/1808.08079","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"pmh:oai:dare.uva.nl:publications/289ae0a6-cc49-4968-940a-438d7d6a3315","is_oa":true,"landing_page_url":"https://handle.uba.uva.nl/personal/pure/en/publications/under-the-hood-using-diagnostic-classifiers-to-investigate-and-improve-how-language-models-track-agreement-information(289ae0a6-cc49-4968-940a-438d7d6a3315).html","pdf_url":"https://pure.uva.nl/ws/files/40247163/W18_5426.pdf","source":{"id":"https://openalex.org/S4306400088","display_name":"UvA-DARE (University of Amsterdam)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I887064364","host_organization_name":"University of Amsterdam","host_organization_lineage":["https://openalex.org/I887064364"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Giulianelli, M, Harding, J, Mohnert, F, Hupkes, D & Zuidema, W 2018, Under the Hood: Using Diagnostic Classifiers to Investigate and Improve how Language Models Track Agreement Information. in T Linzen, G Chrupa\u0142a & A Alishahi (eds), The 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP : EMNLP 2018 : proceedings of the First Workshop : November 1, 2018, Brussels, Belgium. Stroudsburg, PA, pp. 240\u2013248, 2018 EMNLP Workshop BlackboxNLP, Brussels, Belgium, 1/11/18. https://doi.org/10.18653/v1/W18-5426","raw_type":"info:eu-repo/semantics/publishedVersion"},{"id":"pmh:oai:dare.uva.nl:openaire_cris_publications/289ae0a6-cc49-4968-940a-438d7d6a3315","is_oa":true,"landing_page_url":"https://hdl.handle.net/11245.1/289ae0a6-cc49-4968-940a-438d7d6a3315","pdf_url":null,"source":{"id":"https://openalex.org/S4306400088","display_name":"UvA-DARE (University of Amsterdam)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I887064364","host_organization_name":"University of Amsterdam","host_organization_lineage":["https://openalex.org/I887064364"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Giulianelli , M , Harding , J , Mohnert , F , Hupkes , D & Zuidema , W 2018 , Under the Hood: Using Diagnostic Classifiers to Investigate and Improve how Language Models Track Agreement Information . in T Linzen , G Chrupa\u0142a & A Alishahi (eds) , The 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP : EMNLP 2018 : proceedings of the First Workshop : November 1, 2018, Brussels, Belgium . Stroudsburg, PA , pp. 240\u2013248 , 2018 EMNLP Workshop BlackboxNLP , Brussels , Belgium , 1/11/18 . https://doi.org/10.18653/v1/W18-5426","raw_type":"contributionToPeriodical"},{"id":"pmh:oai:eprints.ucl.ac.uk.OAI2:10216470","is_oa":false,"landing_page_url":"https://discovery.ucl.ac.uk/id/eprint/10216470/","pdf_url":null,"source":{"id":"https://openalex.org/S4306400024","display_name":"UCL Discovery (University College London)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I45129253","host_organization_name":"University College London","host_organization_lineage":["https://openalex.org/I45129253"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"In: Linzen, Tal and Chrupa\u0142a, Grzegorz and Alishahi, Afra, (eds.) Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP.  (pp. pp. 240-248).  Association for Computational Linguistics: Brussels, Belgium. (2018)","raw_type":"Proceedings paper"},{"id":"pmh:uvapub:oai:dare.uva.nl:publications/289ae0a6-cc49-4968-940a-438d7d6a3315","is_oa":true,"landing_page_url":"https://dare.uva.nl/personal/pure/en/publications/under-the-hood-using-diagnostic-classifiers-to-investigate-and-improve-how-language-models-track-agreement-information(289ae0a6-cc49-4968-940a-438d7d6a3315).html","pdf_url":null,"source":{"id":"https://openalex.org/S4306401843","display_name":"Data Archiving and Networked Services (DANS)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1322597698","host_organization_name":"Royal Netherlands Academy of Arts and Sciences","host_organization_lineage":["https://openalex.org/I1322597698"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"The 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP: EMNLP 2018 : proceedings of the First Workshop : November 1, 2018, Brussels, Belgium","raw_type":"info:eu-repo/semantics/conferencepaper"},{"id":"doi:10.48550/arxiv.1808.08079","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.1808.08079","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"},{"id":"mag:2888169584","is_oa":false,"landing_page_url":null,"pdf_url":null,"source":null,"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":null}],"best_oa_location":{"id":"doi:10.18653/v1/w18-5426","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/w18-5426","pdf_url":"https://www.aclweb.org/anthology/W18-5426.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","score":0.8199999928474426,"display_name":"Quality Education"}],"awards":[{"id":"https://openalex.org/G2084304408","display_name":null,"funder_award_id":"024.001.006","funder_id":"https://openalex.org/F4320321800","funder_display_name":"Nederlandse Organisatie voor Wetenschappelijk Onderzoek"},{"id":"https://openalex.org/G5365867299","display_name":null,"funder_award_id":"Grant","funder_id":"https://openalex.org/F4320321800","funder_display_name":"Nederlandse Organisatie voor Wetenschappelijk Onderzoek"},{"id":"https://openalex.org/G629491556","display_name":null,"funder_award_id":"(NWO)","funder_id":"https://openalex.org/F4320321800","funder_display_name":"Nederlandse Organisatie voor Wetenschappelijk Onderzoek"},{"id":"https://openalex.org/G6728881174","display_name":null,"funder_award_id":"Gravitation Grant 024.001.006","funder_id":"https://openalex.org/F4320321800","funder_display_name":"Nederlandse Organisatie voor Wetenschappelijk Onderzoek"},{"id":"https://openalex.org/G6840858781","display_name":null,"funder_award_id":"24.001.006","funder_id":"https://openalex.org/F4320321800","funder_display_name":"Nederlandse Organisatie voor Wetenschappelijk Onderzoek"}],"funders":[{"id":"https://openalex.org/F4320321800","display_name":"Nederlandse Organisatie voor Wetenschappelijk Onderzoek","ror":"https://ror.org/04jsz6e67"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W2888169584.pdf","grobid_xml":"https://content.openalex.org/works/W2888169584.grobid-xml"},"referenced_works_count":12,"referenced_works":["https://openalex.org/W1539312853","https://openalex.org/W2037504148","https://openalex.org/W2064675550","https://openalex.org/W2292919134","https://openalex.org/W2549835527","https://openalex.org/W2557263767","https://openalex.org/W2792376130","https://openalex.org/W2799124508","https://openalex.org/W2962776659","https://openalex.org/W2963430224","https://openalex.org/W2963751529","https://openalex.org/W2963951265"],"related_works":["https://openalex.org/W2964222268","https://openalex.org/W2952718809","https://openalex.org/W3014761147","https://openalex.org/W3154493564","https://openalex.org/W2784977129","https://openalex.org/W3102226577","https://openalex.org/W3176834550","https://openalex.org/W3007100556","https://openalex.org/W2972433973","https://openalex.org/W3093264153","https://openalex.org/W3105511442","https://openalex.org/W2475227985","https://openalex.org/W2252108951","https://openalex.org/W3116152597","https://openalex.org/W2158468113","https://openalex.org/W3028787912","https://openalex.org/W3046003314","https://openalex.org/W3125115207","https://openalex.org/W2889506631","https://openalex.org/W2895975470"],"abstract_inverted_index":{"How":[0],"do":[1],"neural":[2,131],"language":[3,29,61,108],"models":[4],"keep":[5],"track":[6],"of":[7,27,35,88,94,127],"number":[8,22,53],"agreement":[9,66,82],"between":[10],"subject":[11],"and":[12,38,51,133],"verb?":[13],"We":[14],"show":[15,114],"that":[16,115,135],"'diagnostic":[17],"classifiers',":[18],"trained":[19],"to":[20,84,141],"predict":[21],"from":[23,98],"the":[24,60,70,75,86,89,92,107,125],"internal":[25],"states":[26],"a":[28,32,103],"model,":[30],"provide":[31],"detailed":[33,122],"understanding":[34],"how,":[36],"when,":[37],"where":[39,52,59],"this":[40,136],"information":[41,54,83,129],"is":[42,55],"represented.":[43],"Moreover,":[44],"they":[45],"give":[46,118],"us":[47,119],"insight":[48],"into":[49,124],"when":[50],"corrupted":[56],"in":[57,106,130],"cases":[58],"model":[62],"ends":[63],"up":[64],"making":[65],"errors.":[67],"To":[68],"demonstrate":[69,134],"causal":[71],"role":[72],"played":[73],"by":[74],"representations":[76],"we":[77,79],"find,":[78],"then":[80],"use":[81],"influence":[85],"course":[87],"LSTM":[90],"during":[91],"processing":[93],"difficult":[95],"sentences.":[96],"Results":[97],"such":[99],"an":[100,120],"intervention":[101],"reveal":[102],"large":[104],"increase":[105],"model's":[109],"accuracy.":[110],"Together,":[111],"these":[112],"results":[113],"diagnostic":[116],"classifiers":[117],"unrivalled":[121],"look":[123],"representation":[126],"linguistic":[128],"models,":[132],"knowledge":[137],"can":[138],"be":[139],"used":[140],"improve":[142],"their":[143],"performance.":[144]},"counts_by_year":[{"year":2023,"cited_by_count":2},{"year":2020,"cited_by_count":3}],"updated_date":"2026-04-10T15:06:20.359241","created_date":"2025-10-10T00:00:00"}