{"id":"https://openalex.org/W4412945081","doi":"https://doi.org/10.18653/v1/2025.acl-long.862","title":"Understanding Large Language Model Vulnerabilities to Social Bias Attacks","display_name":"Understanding Large Language Model Vulnerabilities to Social Bias Attacks","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945081","doi":"https://doi.org/10.18653/v1/2025.acl-long.862"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.862","is_oa":false,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.862","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5081885159","display_name":"Jiaxu Zhao","orcid":"https://orcid.org/0009-0003-7731-0453"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiaxu Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100400497","display_name":"Meng Fang","orcid":"https://orcid.org/0000-0003-0793-9187"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Meng Fang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078520300","display_name":"Fanghua Ye","orcid":"https://orcid.org/0000-0002-1063-6534"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fanghua Ye","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5055309934","display_name":"Ke Xu","orcid":"https://orcid.org/0000-0002-6472-7052"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ke Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056626064","display_name":"Qin Zhang","orcid":"https://orcid.org/0000-0001-5010-9715"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qin Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Joey Tianyi Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Joey Tianyi Zhou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5022601535","display_name":"Mykola Pechenizkiy","orcid":"https://orcid.org/0000-0003-4955-0743"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mykola Pechenizkiy","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"17620","last_page":"17636"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.817300021648407,"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":0.817300021648407,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.7991999983787537,"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":0.7231317162513733},{"id":"https://openalex.org/keywords/computer-security","display_name":"Computer security","score":0.5715804696083069},{"id":"https://openalex.org/keywords/internet-privacy","display_name":"Internet privacy","score":0.3521832823753357}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7231317162513733},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.5715804696083069},{"id":"https://openalex.org/C108827166","wikidata":"https://www.wikidata.org/wiki/Q175975","display_name":"Internet privacy","level":1,"score":0.3521832823753357}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.862","is_oa":false,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.862","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},{"id":"pmh:oai:pure.tue.nl:publications/405cbfd7-f1b1-4def-8c8e-02c019f8b1ff","is_oa":false,"landing_page_url":"https://research.tue.nl/en/publications/405cbfd7-f1b1-4def-8c8e-02c019f8b1ff","pdf_url":null,"source":{"id":"https://openalex.org/S4406922641","display_name":"TU/e Research Portal","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Zhao, J, Fang, M, Ye, F, Xu, K, Zhang, Q, Zhou, J T & Pechenizkiy, M 2025, Understanding Large Language Model Vulnerabilities to Social Bias Attacks. in Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics : Volume 1: Long Papers. Association for Computational Linguistics (ACL), pp. 17620-17636, 63rd Annual Meeting of the Association for Computational Linguistics, ACL 2025, Vienna, Austria, 27/07/25. https://doi.org/10.18653/v1/2025.acl-long.862","raw_type":"info:eu-repo/semantics/conferenceObject"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.7400000095367432,"id":"https://metadata.un.org/sdg/10","display_name":"Reduced inequalities"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4],"become":[5],"foundational":[6],"in":[7,31],"human-computer":[8],"interaction,":[9],"demonstrating":[10],"remarkable":[11],"linguistic":[12],"capabilities":[13],"across":[14,70,112],"various":[15,47],"tasks.":[16],"However,":[17],"there":[18],"is":[19],"a":[20],"growing":[21],"concern":[22],"about":[23],"their":[24,32],"potential":[25,153],"to":[26,46,86,91,130,136,164],"perpetuate":[27],"social":[28,48,171],"biases":[29],"present":[30],"training":[33],"data.":[34],"In":[35],"this":[36],"paper,":[37],"we":[38],"comprehensively":[39],"investigate":[40],"the":[41,137,147,166],"vulnerabilities":[42],"of":[43,110,139,149,168],"contemporary":[44],"LLMs":[45,169],"bias":[49,75,88,113,131,154,172],"attacks,":[50,106],"including":[51],"prefix":[52],"injection,":[53],"refusal":[54],"suppression,":[55],"and":[56,68,73,104,122,142,151,162],"learned":[57],"attack":[58],"prompts.":[59],"We":[60,96,156],"evaluate":[61],"popular":[62],"models":[63,81,121,125],"such":[64,101],"as":[65,102],"LLaMA-2,":[66],"GPT-3.5,":[67],"GPT-4":[69],"gender,":[71],"racial,":[72],"religious":[74,94],"types.":[76,114],"Our":[77],"findings":[78],"reveal":[79],"that":[80,119],"are":[82],"generally":[83],"more":[84,140],"susceptible":[85],"gender":[87],"attacks":[89],"compared":[90],"racial":[92],"or":[93],"biases.":[95],"also":[97],"explore":[98],"novel":[99],"aspects":[100],"cross-bias":[103],"multiple-bias":[105],"finding":[107],"varying":[108],"degrees":[109],"transferability":[111],"Additionally,":[115],"our":[116],"results":[117],"show":[118],"larger":[120],"pretrained":[123],"base":[124],"often":[126],"exhibit":[127],"higher":[128],"susceptibility":[129],"attacks.":[132,173],"These":[133],"insights":[134],"contribute":[135],"development":[138],"inclusive":[141],"ethically":[143],"responsible":[144],"LLMs,":[145],"emphasizing":[146],"importance":[148],"understanding":[150],"mitigating":[152],"vulnerabilities.":[155],"offer":[157],"recommendations":[158],"for":[159],"model":[160],"developers":[161],"users":[163],"enhance":[165],"robustness":[167],"against":[170]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}