{"id":"https://openalex.org/W4385573668","doi":"https://doi.org/10.18653/v1/2022.findings-emnlp.180","title":"Towards Tracing Knowledge in Language Models Back to the Training Data","display_name":"Towards Tracing Knowledge in Language Models Back to the Training Data","publication_year":2022,"publication_date":"2022-01-01","ids":{"openalex":"https://openalex.org/W4385573668","doi":"https://doi.org/10.18653/v1/2022.findings-emnlp.180"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2022.findings-emnlp.180","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2022.findings-emnlp.180","pdf_url":"https://aclanthology.org/2022.findings-emnlp.180.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":"Findings of the Association for Computational Linguistics: EMNLP 2022","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2022.findings-emnlp.180.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5071734030","display_name":"Ekin Aky\u00fcrek","orcid":"https://orcid.org/0000-0002-5166-4689"},"institutions":[{"id":"https://openalex.org/I4405258931","display_name":"MIT Computer Science and Artificial Intelligence Laboratory","ror":"https://ror.org/01e85r662","country_code":"US","type":"facility","lineage":["https://openalex.org/I4405258931","https://openalex.org/I63966007"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ekin Akyurek","raw_affiliation_strings":["MIT CSAIL"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIT CSAIL","institution_ids":["https://openalex.org/I4405258931"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082178510","display_name":"Tolga Bolukbasi","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tolga Bolukbasi","raw_affiliation_strings":["Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5033750922","display_name":"Frederick Liu","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Frederick Liu","raw_affiliation_strings":["Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048733127","display_name":"Binbin Xiong","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Binbin Xiong","raw_affiliation_strings":["Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082431329","display_name":"Ian Tenney","orcid":"https://orcid.org/0000-0002-6225-9283"},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ian Tenney","raw_affiliation_strings":["Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070829652","display_name":"Jacob Andreas","orcid":"https://orcid.org/0000-0002-3141-5845"},"institutions":[{"id":"https://openalex.org/I4405258931","display_name":"MIT Computer Science and Artificial Intelligence Laboratory","ror":"https://ror.org/01e85r662","country_code":"US","type":"facility","lineage":["https://openalex.org/I4405258931","https://openalex.org/I63966007"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jacob Andreas","raw_affiliation_strings":["MIT CSAIL"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIT CSAIL","institution_ids":["https://openalex.org/I4405258931"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5076669462","display_name":"Kelvin Guu","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Kelvin Guu","raw_affiliation_strings":["Google Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":28,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"2429","last_page":"2446"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"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"}},"topics":[{"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9984999895095825,"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/T11719","display_name":"Data Quality and Management","score":0.9786999821662903,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8044719099998474},{"id":"https://openalex.org/keywords/assertion","display_name":"Assertion","score":0.7634283304214478},{"id":"https://openalex.org/keywords/tracing","display_name":"Tracing","score":0.6340284943580627},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.5245973467826843},{"id":"https://openalex.org/keywords/memorization","display_name":"Memorization","score":0.4974994957447052},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.4651642143726349},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.4321703314781189},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.41876283288002014},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.41477060317993164},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.34372586011886597},{"id":"https://openalex.org/keywords/information-retrieval","display_name":"Information retrieval","score":0.3326065242290497},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.08379960060119629}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8044719099998474},{"id":"https://openalex.org/C40422974","wikidata":"https://www.wikidata.org/wiki/Q741248","display_name":"Assertion","level":2,"score":0.7634283304214478},{"id":"https://openalex.org/C138673069","wikidata":"https://www.wikidata.org/wiki/Q322229","display_name":"Tracing","level":2,"score":0.6340284943580627},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5245973467826843},{"id":"https://openalex.org/C30038468","wikidata":"https://www.wikidata.org/wiki/Q4354775","display_name":"Memorization","level":2,"score":0.4974994957447052},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.4651642143726349},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.4321703314781189},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.41876283288002014},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.41477060317993164},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.34372586011886597},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.3326065242290497},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.08379960060119629},{"id":"https://openalex.org/C205649164","wikidata":"https://www.wikidata.org/wiki/Q1071","display_name":"Geography","level":0,"score":0.0},{"id":"https://openalex.org/C145420912","wikidata":"https://www.wikidata.org/wiki/Q853077","display_name":"Mathematics education","level":1,"score":0.0},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.0},{"id":"https://openalex.org/C13280743","wikidata":"https://www.wikidata.org/wiki/Q131089","display_name":"Geodesy","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2022.findings-emnlp.180","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2022.findings-emnlp.180","pdf_url":"https://aclanthology.org/2022.findings-emnlp.180.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":"Findings of the Association for Computational Linguistics: EMNLP 2022","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2022.findings-emnlp.180","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2022.findings-emnlp.180","pdf_url":"https://aclanthology.org/2022.findings-emnlp.180.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":"Findings of the Association for Computational Linguistics: EMNLP 2022","raw_type":"proceedings-article"},"sustainable_development_goals":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.8100000023841858}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4385573668.pdf"},"referenced_works_count":34,"referenced_works":["https://openalex.org/W1989898472","https://openalex.org/W2005545412","https://openalex.org/W2577240759","https://openalex.org/W2594633041","https://openalex.org/W2597603852","https://openalex.org/W2785611959","https://openalex.org/W2909737760","https://openalex.org/W2946417913","https://openalex.org/W2949279173","https://openalex.org/W2951534261","https://openalex.org/W2962833140","https://openalex.org/W2962851944","https://openalex.org/W2963341956","https://openalex.org/W2970476646","https://openalex.org/W3027879771","https://openalex.org/W3034475796","https://openalex.org/W3034671305","https://openalex.org/W3037881973","https://openalex.org/W3093205559","https://openalex.org/W3093517588","https://openalex.org/W3099700870","https://openalex.org/W3101656801","https://openalex.org/W3102659883","https://openalex.org/W3118560954","https://openalex.org/W3131061281","https://openalex.org/W3152884768","https://openalex.org/W3169483174","https://openalex.org/W3213681714","https://openalex.org/W4221149101","https://openalex.org/W4283810737","https://openalex.org/W4286897388","https://openalex.org/W4288089799","https://openalex.org/W4288091035","https://openalex.org/W4293569541"],"related_works":["https://openalex.org/W1498103021","https://openalex.org/W2035655557","https://openalex.org/W4230849338","https://openalex.org/W4295166216","https://openalex.org/W2177044681","https://openalex.org/W1968067090","https://openalex.org/W2090313512","https://openalex.org/W2345942070","https://openalex.org/W2141398161","https://openalex.org/W2016410697"],"abstract_inverted_index":{"Language":[0],"models":[1],"(LMs)":[2],"have":[3,115,128],"been":[4],"shown":[5],"to":[6,30,59,89,130],"memorize":[7],"a":[8,61],"great":[9],"deal":[10],"of":[11,49,97,151,162],"factual":[12,63],"knowledge":[13],"contained":[14],"in":[15],"their":[16],"training":[17,54,68],"data.":[18],"But":[19],"when":[20],"an":[21,24,57,120],"LM":[22,58,132],"generates":[23],"assertion,":[25],"it":[26,33,39],"is":[27,40],"often":[28],"difficult":[29],"determine":[31],"where":[32],"learned":[34],"this":[35,43],"information":[36,121],"and":[37,102,105,154],"whether":[38],"true.":[41],"In":[42],"paper,":[44],"we":[45],"propose":[46],"the":[47,85,131,149],"problem":[48,150],"fact":[50,171],"tracing:":[51],"identifying":[52,77],"which":[53],"examples":[55],"taught":[56],"generate":[60],"particular":[62],"assertion.":[64],"Prior":[65],"work":[66],"on":[67],"data":[69],"attribution":[70],"(TDA)":[71],"may":[72,140],"offer":[73],"effective":[74],"tools":[75],"for":[76,143],"such":[78,146],"examples,":[79],"known":[80],"as":[81,147],"\u201cproponents\u201d.":[82],"We":[83,92,135],"present":[84],"first":[86],"quantitative":[87],"benchmark":[88],"evaluate":[90],"this.":[91],"compare":[93],"two":[94],"popular":[95],"families":[96],"TDA":[98,165],"methods":[99,114,166],"\u2014":[100,104],"gradient-based":[101],"embedding-based":[103],"find":[106],"that":[107,125,139],"much":[108],"headroom":[109],"remains.":[110],"For":[111],"example,":[112],"both":[113],"lower":[116],"proponent-retrieval":[117],"precision":[118],"than":[119],"retrieval":[122],"baseline":[123],"(BM25)":[124],"does":[126],"not":[127],"access":[129],"at":[133],"all.":[134],"identify":[136],"key":[137],"challenges":[138],"be":[141],"necessary":[142],"further":[144],"improvement":[145],"overcoming":[148],"gradient":[152],"saturation,":[153],"also":[155],"show":[156],"how":[157],"several":[158],"nuanced":[159],"implementation":[160],"details":[161],"existing":[163],"neural":[164],"can":[167],"significantly":[168],"improve":[169],"overall":[170],"tracing":[172],"performance.":[173]},"counts_by_year":[{"year":2026,"cited_by_count":4},{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":11},{"year":2023,"cited_by_count":9}],"updated_date":"2026-07-16T13:24:37.021932","created_date":"2025-10-10T00:00:00"}