{"id":"https://openalex.org/W4416035110","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.1789","title":"Retracing the Past: LLMs Emit Training Data When They Get Lost","display_name":"Retracing the Past: LLMs Emit Training Data When They Get Lost","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416035110","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.1789"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.1789","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1789","pdf_url":"https://aclanthology.org/2025.emnlp-main.1789.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.emnlp-main.1789.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5120270399","display_name":"Myeongseob Ko","orcid":null},"institutions":[{"id":"https://openalex.org/I4210137942","display_name":"D-Tech (United States)","ror":"https://ror.org/03tw6b878","country_code":"US","type":"company","lineage":["https://openalex.org/I4210137942"]},{"id":"https://openalex.org/I859038795","display_name":"Virginia Tech","ror":"https://ror.org/02smfhw86","country_code":"US","type":"education","lineage":["https://openalex.org/I859038795"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Myeongseob Ko","raw_affiliation_strings":["Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech"],"affiliations":[{"raw_affiliation_string":"Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech","institution_ids":["https://openalex.org/I4210137942","https://openalex.org/I859038795"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120308755","display_name":"Nikhil Reddy Billa","orcid":null},"institutions":[{"id":"https://openalex.org/I4210137942","display_name":"D-Tech (United States)","ror":"https://ror.org/03tw6b878","country_code":"US","type":"company","lineage":["https://openalex.org/I4210137942"]},{"id":"https://openalex.org/I859038795","display_name":"Virginia Tech","ror":"https://ror.org/02smfhw86","country_code":"US","type":"education","lineage":["https://openalex.org/I859038795"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Nikhil Reddy Billa","raw_affiliation_strings":["Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech"],"affiliations":[{"raw_affiliation_string":"Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech","institution_ids":["https://openalex.org/I4210137942","https://openalex.org/I859038795"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066681619","display_name":"Anh Quoc Nguyen","orcid":"https://orcid.org/0000-0001-7298-8060"},"institutions":[{"id":"https://openalex.org/I4210137942","display_name":"D-Tech (United States)","ror":"https://ror.org/03tw6b878","country_code":"US","type":"company","lineage":["https://openalex.org/I4210137942"]},{"id":"https://openalex.org/I859038795","display_name":"Virginia Tech","ror":"https://ror.org/02smfhw86","country_code":"US","type":"education","lineage":["https://openalex.org/I859038795"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Adam Nguyen","raw_affiliation_strings":["Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech"],"affiliations":[{"raw_affiliation_string":"Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech","institution_ids":["https://openalex.org/I4210137942","https://openalex.org/I859038795"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063903215","display_name":"Charles Fleming","orcid":null},"institutions":[{"id":"https://openalex.org/I4210137942","display_name":"D-Tech (United States)","ror":"https://ror.org/03tw6b878","country_code":"US","type":"company","lineage":["https://openalex.org/I4210137942"]},{"id":"https://openalex.org/I859038795","display_name":"Virginia Tech","ror":"https://ror.org/02smfhw86","country_code":"US","type":"education","lineage":["https://openalex.org/I859038795"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Charles Fleming","raw_affiliation_strings":["Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech"],"affiliations":[{"raw_affiliation_string":"Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech","institution_ids":["https://openalex.org/I4210137942","https://openalex.org/I859038795"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101751858","display_name":"Jin Ming","orcid":"https://orcid.org/0000-0001-8260-9669"},"institutions":[{"id":"https://openalex.org/I4210137942","display_name":"D-Tech (United States)","ror":"https://ror.org/03tw6b878","country_code":"US","type":"company","lineage":["https://openalex.org/I4210137942"]},{"id":"https://openalex.org/I859038795","display_name":"Virginia Tech","ror":"https://ror.org/02smfhw86","country_code":"US","type":"education","lineage":["https://openalex.org/I859038795"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ming Jin","raw_affiliation_strings":["Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech"],"affiliations":[{"raw_affiliation_string":"Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech","institution_ids":["https://openalex.org/I4210137942","https://openalex.org/I859038795"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5032275274","display_name":"Ruoxi Jia","orcid":"https://orcid.org/0000-0001-9662-9556"},"institutions":[{"id":"https://openalex.org/I4210137942","display_name":"D-Tech (United States)","ror":"https://ror.org/03tw6b878","country_code":"US","type":"company","lineage":["https://openalex.org/I4210137942"]},{"id":"https://openalex.org/I859038795","display_name":"Virginia Tech","ror":"https://ror.org/02smfhw86","country_code":"US","type":"education","lineage":["https://openalex.org/I859038795"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ruoxi Jia","raw_affiliation_strings":["Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech"],"affiliations":[{"raw_affiliation_string":"Virginia Tech Virginia Tech Cisco Research Virginia Tech Virginia Tech","institution_ids":["https://openalex.org/I4210137942","https://openalex.org/I859038795"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5120270399"],"corresponding_institution_ids":["https://openalex.org/I4210137942","https://openalex.org/I859038795"],"apc_list":null,"apc_paid":null,"fwci":2.9892,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.93806827,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":96,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"35304","last_page":"35325"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11937","display_name":"Research Data Management Practices","score":0.10540000349283218,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T11937","display_name":"Research Data Management Practices","score":0.10540000349283218,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T14330","display_name":"Library Science and Information Systems","score":0.041999999433755875,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T12478","display_name":"Wikis in Education and Collaboration","score":0.03720000013709068,"subfield":{"id":"https://openalex.org/subfields/3315","display_name":"Communication"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.670199990272522},{"id":"https://openalex.org/keywords/work","display_name":"Work (physics)","score":0.24400000274181366},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.2313999980688095},{"id":"https://openalex.org/keywords/government","display_name":"Government (linguistics)","score":0.22269999980926514}],"concepts":[{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.670199990272522},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.38769999146461487},{"id":"https://openalex.org/C39549134","wikidata":"https://www.wikidata.org/wiki/Q133080","display_name":"Public relations","level":1,"score":0.32350000739097595},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.3059000074863434},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.2939000129699707},{"id":"https://openalex.org/C95457728","wikidata":"https://www.wikidata.org/wiki/Q309","display_name":"History","level":0,"score":0.2791000008583069},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.24400000274181366},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2313999980688095},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.22939999401569366},{"id":"https://openalex.org/C2778137410","wikidata":"https://www.wikidata.org/wiki/Q2732820","display_name":"Government (linguistics)","level":2,"score":0.22269999980926514}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.1789","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1789","pdf_url":"https://aclanthology.org/2025.emnlp-main.1789.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.1789","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1789","pdf_url":"https://aclanthology.org/2025.emnlp-main.1789.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416035110.pdf","grobid_xml":"https://content.openalex.org/works/W4416035110.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"memorization":[1,36,145],"of":[2,35,60,138],"training":[3,132,140],"data":[4,16,49,133],"in":[5,71,127],"large":[6],"language":[7],"models":[8],"(LLMs)":[9],"poses":[10],"significant":[11],"privacy":[12],"and":[13,27,103,116,130,150],"copyright":[14],"concerns.Existing":[15],"extraction":[17],"methods,":[18],"particularly":[19],"heuristic-based":[20],"divergence":[21,64],"attacks,":[22],"often":[23],"exhibit":[24],"limited":[25,29],"success":[26],"offer":[28,151],"insight":[30,77],"into":[31],"the":[32,58,139],"fundamental":[33],"drivers":[34],"leakage.This":[37],"paper":[38],"introduces":[39],"Confusion-Inducing":[40],"Attacks":[41],"(CIA),":[42],"a":[43,68,152],"principled":[44],"framework":[45],"for":[46,156],"extracting":[47,128],"memorized":[48,61],"by":[50,67,78],"systematically":[51],"maximizing":[52],"model":[53],"uncertainty.We":[54],"empirically":[55],"demonstrate":[56,119],"that":[57,120],"emission":[59],"text":[62],"during":[63],"is":[65],"preceded":[66],"sustained":[69],"spike":[70],"token-level":[72],"prediction":[73],"entropy.CIA":[74],"leverages":[75],"this":[76,85],"optimizing":[79],"input":[80],"snippets":[81],"to":[82,98,110],"deliberately":[83],"induce":[84,104],"consecutive":[86],"highentropy":[87],"state.For":[88],"aligned":[89,117],"LLMs,":[90],"we":[91],"further":[92],"propose":[93],"mismatched":[94],"Supervised":[95],"Fine-tuning":[96],"(SFT)":[97],"simultaneously":[99],"weaken":[100],"their":[101],"alignment":[102],"targeted":[105],"confusion,":[106],"thereby":[107],"increasing":[108],"susceptibility":[109],"our":[111,121],"attacks.Experiments":[112],"on":[113],"various":[114,148],"unaligned":[115],"LLMs":[118,149],"proposed":[122],"attacks":[123],"outperform":[124],"existing":[125],"baselines":[126],"verbatim":[129],"near-verbatim":[131],"without":[134],"requiring":[135],"prior":[136],"knowledge":[137],"data.Our":[141],"findings":[142],"highlight":[143],"persistent":[144],"risks":[146],"across":[147],"more":[153],"systematic":[154],"method":[155],"assessing":[157],"these":[158],"vulnerabilities.":[159]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-04-17T18:11:37.981687","created_date":"2025-11-08T00:00:00"}