{"id":"https://openalex.org/W4412888412","doi":"https://doi.org/10.18653/v1/2025.findings-acl.535","title":"From Evasion to Concealment: Stealthy Knowledge Unlearning for LLMs","display_name":"From Evasion to Concealment: Stealthy Knowledge Unlearning for LLMs","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888412","doi":"https://doi.org/10.18653/v1/2025.findings-acl.535"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.535","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.535","pdf_url":"https://aclanthology.org/2025.findings-acl.535.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.535.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102499785","display_name":"Tianle Gu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tianle Gu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101733674","display_name":"Kexin Huang","orcid":"https://orcid.org/0009-0001-5697-6710"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kexin Huang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Ruilin Luo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ruilin Luo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111230251","display_name":"Yuanqi Yao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuanqi Yao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101568161","display_name":"Xiuying Chen","orcid":"https://orcid.org/0000-0001-7598-0363"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiuying Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5020953714","display_name":"Yujiu Yang","orcid":"https://orcid.org/0000-0002-6427-1024"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yujiu Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100525584","display_name":"Yan Teng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yan Teng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100613148","display_name":"Yingchun Wang","orcid":"https://orcid.org/0000-0003-4968-8398"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yingchun Wang","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":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"10261","last_page":"10279"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T14431","display_name":"Diverse Research and Applications","score":0.3865000009536743,"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/T14431","display_name":"Diverse Research and Applications","score":0.3865000009536743,"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/T11903","display_name":"Private Equity and Venture Capital","score":0.3741999864578247,"subfield":{"id":"https://openalex.org/subfields/1402","display_name":"Accounting"},"field":{"id":"https://openalex.org/fields/14","display_name":"Business, Management and Accounting"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/evasion","display_name":"Evasion (ethics)","score":0.687996506690979},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5534241795539856},{"id":"https://openalex.org/keywords/computer-security","display_name":"Computer security","score":0.3413350582122803},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.33148378133773804},{"id":"https://openalex.org/keywords/biology","display_name":"Biology","score":0.09148207306861877},{"id":"https://openalex.org/keywords/immunology","display_name":"Immunology","score":0.06426408886909485}],"concepts":[{"id":"https://openalex.org/C2781251061","wikidata":"https://www.wikidata.org/wiki/Q5416089","display_name":"Evasion (ethics)","level":3,"score":0.687996506690979},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5534241795539856},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.3413350582122803},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.33148378133773804},{"id":"https://openalex.org/C86803240","wikidata":"https://www.wikidata.org/wiki/Q420","display_name":"Biology","level":0,"score":0.09148207306861877},{"id":"https://openalex.org/C203014093","wikidata":"https://www.wikidata.org/wiki/Q101929","display_name":"Immunology","level":1,"score":0.06426408886909485},{"id":"https://openalex.org/C8891405","wikidata":"https://www.wikidata.org/wiki/Q1059","display_name":"Immune system","level":2,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.535","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.535","pdf_url":"https://aclanthology.org/2025.findings-acl.535.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.535","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.535","pdf_url":"https://aclanthology.org/2025.findings-acl.535.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/W4412888412.pdf","grobid_xml":"https://content.openalex.org/works/W4412888412.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/W2373230814","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890"],"abstract_inverted_index":{"LLM":[0,108],"Unlearning":[1,161],"plays":[2],"a":[3,75,117],"crucial":[4],"role":[5],"in":[6,170],"removing":[7],"sensitive":[8,147],"information":[9,41,148],"from":[10],"language":[11],"models":[12,58,87],"to":[13,49,65,125,190],"mitigate":[14],"potential":[15],"misuse.However,":[16],"previous":[17],"approaches":[18],"often":[19],"treat":[20],"nonsensical":[21],"responses":[22],"or":[23,59,88],"template-based":[24],"refusals":[25],"(e.g.,":[26],"Sorry,":[27],"I":[28],"cannot":[29],"answer.)as":[30],"the":[31,37,44,66,83,122,137,141,154,158,194],"unlearning":[32,67,79],"target,":[33],"which":[34,62],"can":[35,206],"give":[36],"impression":[38],"of":[39,98,131,140],"deliberate":[40],"suppression,":[42],"making":[43],"process":[45],"even":[46],"more":[47],"vulnerable":[48],"attacks":[50],"and":[51,77,109,165,183,186,200,205],"jailbreaks.Moreover,":[52],"most":[53],"methods":[54],"rely":[55],"on":[56,157],"auxiliary":[57,86],"retaining":[60,89],"datasets,":[61],"adds":[63],"complexity":[64],"process.To":[68],"address":[69],"these":[70],"challenges,":[71],"we":[72,114],"propose":[73],"MEOW,":[74],"streamlined":[76],"stealthy":[78],"method":[80],"that":[81,120,146],"eliminates":[82],"need":[84],"for":[85],"data":[90],"while":[91,173],"avoiding":[92],"leakage":[93],"through":[94],"its":[95],"innovative":[96],"use":[97,130],"inverted":[99,101,132],"facts.These":[100],"facts":[102,133],"are":[103],"generated":[104],"by":[105],"an":[106],"offline":[107],"serve":[110],"as":[111],"fine-tuning":[112,128],"labels.Meanwhile,":[113],"introduce":[115],"MEMO,":[116],"novel":[118],"metric":[119],"measures":[121],"model's":[123],"memorization,":[124],"select":[126],"optimal":[127],"targets.The":[129],"not":[134],"only":[135],"maintains":[136,178],"covert":[138],"nature":[139],"model":[142,175],"but":[143],"also":[144,177],"ensures":[145],"is":[149],"effectively":[150],"forgotten":[151],"without":[152],"revealing":[153],"target":[155],"data.Evaluated":[156],"ToFU":[159],"Knowledge":[160],"dataset":[162],"using":[163],"Llama2-7B-Chat":[164],"Phi-1.5,":[166],"MEOW":[167],"outperforms":[168],"baselines":[169],"forgetting":[171],"quality":[172],"preserving":[174],"utility.MEOW":[176],"strong":[179],"performance":[180],"across":[181],"NLU":[182],"NLG":[184],"tasks":[185],"demonstrates":[187],"superior":[188],"resilience":[189],"attacks,":[191],"validated":[192],"via":[193],"Min-K%":[195],"membership":[196],"inference":[197],"method.All":[198],"codes":[199],"datasets":[201],"have":[202],"been":[203],"released":[204],"be":[207],"accessed":[208],"at":[209],"MEOW-repo.":[210]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}