{"id":"https://openalex.org/W4403488063","doi":"https://doi.org/10.3233/faia240662","title":"WPN: An Unlearning Method Based on N-pair Contrastive Learning in Language Models","display_name":"WPN: An Unlearning Method Based on N-pair Contrastive Learning in Language Models","publication_year":2024,"publication_date":"2024-10-16","ids":{"openalex":"https://openalex.org/W4403488063","doi":"https://doi.org/10.3233/faia240662"},"language":"en","primary_location":{"id":"doi:10.3233/faia240662","is_oa":true,"landing_page_url":"https://doi.org/10.3233/faia240662","pdf_url":null,"source":{"id":"https://openalex.org/S4210201731","display_name":"Frontiers in artificial intelligence and applications","issn_l":"0922-6389","issn":["0922-6389","1879-8314"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"journal"},"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Frontiers in Artificial Intelligence and Applications","raw_type":"book-chapter"},"type":"book-chapter","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.3233/faia240662","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5030454718","display_name":"Guitao Chen","orcid":null},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guitao Chen","raw_affiliation_strings":["Beijing University of Posts and Telecommunications"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100754860","display_name":"Yunsheng Wang","orcid":"https://orcid.org/0000-0002-2552-8253"},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yunshen Wang","raw_affiliation_strings":["Beijing University of Posts and Telecommunications"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001965522","display_name":"Hongye Sun","orcid":null},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hongye Sun","raw_affiliation_strings":["Beijing University of Posts and Telecommunications"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5102992722","display_name":"Guang Chen","orcid":"https://orcid.org/0000-0003-1487-8044"},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guang Chen","raw_affiliation_strings":["Beijing University of Posts and Telecommunications"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications","institution_ids":["https://openalex.org/I139759216"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I139759216"],"apc_list":null,"apc_paid":null,"fwci":0.6858,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.7321165,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":91,"max":95},"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.8773999810218811,"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":0.8773999810218811,"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":0.8704000115394592,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.8338000178337097,"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/computer-science","display_name":"Computer science","score":0.6004540324211121},{"id":"https://openalex.org/keywords/contrastive-analysis","display_name":"Contrastive analysis","score":0.5277605056762695},{"id":"https://openalex.org/keywords/linguistics","display_name":"Linguistics","score":0.4771106541156769},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.39970341324806213},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.3981906771659851},{"id":"https://openalex.org/keywords/psychology","display_name":"Psychology","score":0.39689403772354126},{"id":"https://openalex.org/keywords/philosophy","display_name":"Philosophy","score":0.12800195813179016}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6004540324211121},{"id":"https://openalex.org/C2777629044","wikidata":"https://www.wikidata.org/wiki/Q614959","display_name":"Contrastive analysis","level":2,"score":0.5277605056762695},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.4771106541156769},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.39970341324806213},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3981906771659851},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.39689403772354126},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.12800195813179016}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.3233/faia240662","is_oa":true,"landing_page_url":"https://doi.org/10.3233/faia240662","pdf_url":null,"source":{"id":"https://openalex.org/S4210201731","display_name":"Frontiers in artificial intelligence and applications","issn_l":"0922-6389","issn":["0922-6389","1879-8314"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"journal"},"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Frontiers in Artificial Intelligence and Applications","raw_type":"book-chapter"}],"best_oa_location":{"id":"doi:10.3233/faia240662","is_oa":true,"landing_page_url":"https://doi.org/10.3233/faia240662","pdf_url":null,"source":{"id":"https://openalex.org/S4210201731","display_name":"Frontiers in artificial intelligence and applications","issn_l":"0922-6389","issn":["0922-6389","1879-8314"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"journal"},"license":"cc-by-nc","license_id":"https://openalex.org/licenses/cc-by-nc","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Frontiers in Artificial Intelligence and Applications","raw_type":"book-chapter"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2945865340","https://openalex.org/W4294495400","https://openalex.org/W4367176324","https://openalex.org/W2972340244","https://openalex.org/W3204019825","https://openalex.org/W1834428557","https://openalex.org/W3200487689","https://openalex.org/W2394697227","https://openalex.org/W3192129151","https://openalex.org/W2922545124"],"abstract_inverted_index":{"Generative":[0],"language":[1],"models":[2],"(LMs)":[3],"offer":[4],"numerous":[5],"advantages":[6],"but":[7],"may":[8],"produce":[9],"inappropriate":[10],"or":[11],"harmful":[12,17,99,135,172],"outputs":[13,100],"due":[14],"to":[15,34,41,89,117,143,165,169],"the":[16,56,91,110,132,171],"knowledge":[18,23],"acquired":[19],"during":[20],"pre-training.":[21],"This":[22],"often":[24],"manifests":[25],"as":[26,30,180],"undesirable":[27],"correspondences,":[28],"such":[29],"\u201charmful":[31,35,114,118],"prompts\u201d":[32],"leading":[33],"outputs,\u201d":[36],"which":[37,75],"our":[38],"research":[39],"aims":[40],"mitigate":[42],"through":[43],"unlearning":[44,47],"techniques.However,":[45],"existing":[46],"methods":[48],"based":[49],"on":[50,122,149,158,182],"gradient":[51],"ascent":[52],"can":[53],"significantly":[54],"impair":[55],"performance":[57,148],"of":[58,94,134,141,176],"LMs.":[59],"To":[60],"address":[61],"this":[62],"issue,":[63],"we":[64,161],"propose":[65],"a":[66,138],"novel":[67],"approach":[68],"called":[69],"Weighted":[70],"Positional":[71],"N-pair":[72],"(WPN)":[73],"Learning,":[74],"leverages":[76],"position-weighted":[77],"mean":[78],"pooling":[79],"within":[80],"an":[81],"n-pair":[82],"contrastive":[83],"learning":[84],"framework.":[85],"WPN":[86,129],"is":[87],"designed":[88],"modify":[90],"output":[92],"distribution":[93],"LMs":[95,126],"by":[96],"eliminating":[97],"specific":[98],"(e.g.,":[101],"replacing":[102],"toxic":[103],"responses":[104],"with":[105],"neutral":[106],"ones),":[107],"thereby":[108],"transforming":[109],"model\u2019s":[111],"behavior":[112],"from":[113],"prompt-harmful":[115],"output\u201d":[116],"prompt-harmless":[119],"response\u201d.":[120],"Experiments":[121],"OPT":[123],"and":[124,178,186],"GPT-NEO":[125],"show":[127],"that":[128],"effectively":[130],"reduces":[131],"proportion":[133],"responses,":[136],"achieving":[137],"harmless":[139],"rate":[140],"up":[142],"95.8%":[144],"while":[145],"maintaining":[146],"stable":[147],"nine":[150],"common":[151],"benchmarks":[152],"(with":[153],"less":[154],"than":[155],"2%":[156],"degradation":[157],"average).":[159],"Moreover,":[160],"provide":[162],"empirical":[163],"evidence":[164],"demonstrate":[166],"WPN\u2019s":[167],"ability":[168],"weaken":[170],"correspondences":[173],"in":[174],"terms":[175],"generalizability":[177],"robustness,":[179],"evaluated":[181],"out-of-distribution":[183],"test":[184],"sets":[185],"under":[187],"adversarial":[188],"attacks.":[189]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-05-21T06:26:12.895304","created_date":"2025-10-10T00:00:00"}