{"id":"https://openalex.org/W4412888737","doi":"https://doi.org/10.18653/v1/2025.findings-acl.255","title":"Unlearning Backdoor Attacks for LLMs with Weak-to-Strong Knowledge Distillation","display_name":"Unlearning Backdoor Attacks for LLMs with Weak-to-Strong Knowledge Distillation","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888737","doi":"https://doi.org/10.18653/v1/2025.findings-acl.255"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.255","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.255","pdf_url":"https://aclanthology.org/2025.findings-acl.255.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: ACL 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.255.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5034822677","display_name":"Shuai Zhao","orcid":"https://orcid.org/0000-0001-5174-5182"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shuai Zhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011376608","display_name":"Xiaobao Wu","orcid":"https://orcid.org/0000-0003-0076-3924"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaobao Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5025645183","display_name":"Cong-Duy Nguyen","orcid":"https://orcid.org/0000-0002-0931-460X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cong-Duy T Nguyen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119181472","display_name":"Yanhao Jia","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yanhao Jia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024179471","display_name":"Meihuizi Jia","orcid":"https://orcid.org/0009-0003-5624-9980"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Meihuizi Jia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5116038830","display_name":"Feng Yichao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Feng Yichao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5050386762","display_name":"Anh Tuan Luu","orcid":"https://orcid.org/0000-0002-1927-9895"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Anh Tuan Luu","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":3.346,"has_fulltext":true,"cited_by_count":2,"citation_normalized_percentile":{"value":0.92309762,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":95,"max":96},"biblio":{"volume":null,"issue":null,"first_page":"4937","last_page":"4952"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.98089998960495,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.98089998960495,"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/T10237","display_name":"Cryptography and Data Security","score":0.9272000193595886,"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/backdoor","display_name":"Backdoor","score":0.9725158214569092},{"id":"https://openalex.org/keywords/distillation","display_name":"Distillation","score":0.5106959939002991},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.46429452300071716},{"id":"https://openalex.org/keywords/computer-security","display_name":"Computer security","score":0.4595346450805664},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.35975944995880127},{"id":"https://openalex.org/keywords/chemistry","display_name":"Chemistry","score":0.1722402274608612}],"concepts":[{"id":"https://openalex.org/C2781045450","wikidata":"https://www.wikidata.org/wiki/Q254569","display_name":"Backdoor","level":2,"score":0.9725158214569092},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.5106959939002991},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.46429452300071716},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.4595346450805664},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.35975944995880127},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.1722402274608612},{"id":"https://openalex.org/C178790620","wikidata":"https://www.wikidata.org/wiki/Q11351","display_name":"Organic chemistry","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.255","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.255","pdf_url":"https://aclanthology.org/2025.findings-acl.255.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: ACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.255","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.255","pdf_url":"https://aclanthology.org/2025.findings-acl.255.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: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888737.pdf","grobid_xml":"https://content.openalex.org/works/W4412888737.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W4320031223","https://openalex.org/W4200629851","https://openalex.org/W4281902577","https://openalex.org/W4309417370","https://openalex.org/W4292107232","https://openalex.org/W3009072493","https://openalex.org/W4401407399"],"abstract_inverted_index":{"Parameter-efficient":[0],"fine-tuning":[1,74],"(PEFT)":[2],"can":[3],"bridge":[4],"the":[5,31,78,86,93,102,106,115,118,138],"gap":[6],"between":[7],"large":[8,126],"language":[9,70,127],"models":[10,128],"(LLMs)":[11],"and":[12,129],"downstream":[13],"tasks.However,":[14],"PEFT":[15],"has":[16,101],"been":[17],"proven":[18],"vulnerable":[19],"to":[20,33,52,75,104,110],"malicious":[21],"attacks.Research":[22],"indicates":[23],"that":[24,99],"poisoned":[25,88],"LLMs,":[26],"even":[27],"after":[28],"PEFT,":[29],"retain":[30],"capability":[32],"activate":[34],"internalized":[35],"backdoors":[36],"when":[37],"input":[38],"samples":[39],"contain":[40],"predefined":[41],"triggers.In":[42],"this":[43,82],"paper,":[44],"we":[45,65],"introduce":[46],"a":[47,68],"novel":[48],"weak-to-strong":[49],"unlearning":[50,92],"algorithm":[51],"defend":[53],"against":[54,145],"backdoor":[55,112,132,146],"attacks":[56,147],"based":[57],"on":[58,123],"feature":[59],"alignment":[60],"knowledge":[61],"distillation,":[62],"named":[63],"W2SDefense.Specifically,":[64],"first":[66],"train":[67],"small-scale":[69],"model":[71,84,90,150],"through":[72],"full-parameter":[73],"serve":[76],"as":[77],"clean":[79],"teacher":[80,83],"model.Then,":[81],"guides":[85],"large-scale":[87],"student":[89,107],"in":[91,143],"backdoor,":[94],"leveraging":[95],"PEFT.Theoretical":[96],"analysis":[97],"suggests":[98],"W2SDefense":[100,142],"potential":[103],"enhance":[105],"model's":[108],"ability":[109],"unlearn":[111],"features,":[113],"preventing":[114],"activation":[116],"of":[117,141],"backdoor.We":[119],"conduct":[120],"comprehensive":[121],"experiments":[122],"three":[124],"state-ofthe-art":[125],"several":[130],"different":[131],"attack":[133],"algorithms.Our":[134],"empirical":[135],"results":[136],"demonstrate":[137],"outstanding":[139],"performance":[140,151],"defending":[144],"without":[148],"compromising":[149],"1":[152],".":[153]},"counts_by_year":[{"year":2025,"cited_by_count":2}],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}