{"id":"https://openalex.org/W4412887864","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1013","title":"Adaptive Detoxification: Safeguarding General Capabilities of LLMs through Toxicity-Aware Knowledge Editing","display_name":"Adaptive Detoxification: Safeguarding General Capabilities of LLMs through Toxicity-Aware Knowledge Editing","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412887864","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1013"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.1013","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1013","pdf_url":"https://aclanthology.org/2025.findings-acl.1013.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.1013.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101425156","display_name":"Yifan Lu","orcid":"https://orcid.org/0000-0003-3787-0440"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yifan Lu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101795243","display_name":"Jing Li","orcid":"https://orcid.org/0000-0001-5934-4359"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jing Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Yigeng Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yigeng Zhou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5015434235","display_name":"Yihui Zhang","orcid":"https://orcid.org/0000-0003-3587-9214"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yihui Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101936537","display_name":"Wenya Wang","orcid":"https://orcid.org/0000-0003-3902-4088"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wenya Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101701960","display_name":"Xiucheng Li","orcid":"https://orcid.org/0009-0006-4145-6698"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiucheng Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5004953265","display_name":"Meishan Zhang","orcid":"https://orcid.org/0000-0001-6335-1340"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Meishan Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048398271","display_name":"Fangming Liu","orcid":"https://orcid.org/0000-0002-8570-1345"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fangming Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100614702","display_name":"Jun Yu","orcid":"https://orcid.org/0000-0001-5008-2153"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jun Yu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100402892","display_name":"Min Zhang","orcid":"https://orcid.org/0000-0002-2225-8024"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Min Zhang","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":"19744","last_page":"19758"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T14351","display_name":"Statistical and Computational Modeling","score":0.8119999766349792,"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/T14351","display_name":"Statistical and Computational Modeling","score":0.8119999766349792,"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/T10215","display_name":"Semantic Web and Ontologies","score":0.6998000144958496,"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/safeguarding","display_name":"Safeguarding","score":0.8498725295066833},{"id":"https://openalex.org/keywords/detoxification","display_name":"Detoxification (alternative medicine)","score":0.8198729157447815},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5559270977973938},{"id":"https://openalex.org/keywords/toxicity","display_name":"Toxicity","score":0.44659629464149475},{"id":"https://openalex.org/keywords/computer-security","display_name":"Computer security","score":0.35532432794570923},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.33585458993911743},{"id":"https://openalex.org/keywords/chemistry","display_name":"Chemistry","score":0.12098133563995361},{"id":"https://openalex.org/keywords/medicine","display_name":"Medicine","score":0.10928770899772644}],"concepts":[{"id":"https://openalex.org/C2776743756","wikidata":"https://www.wikidata.org/wiki/Q5097921","display_name":"Safeguarding","level":2,"score":0.8498725295066833},{"id":"https://openalex.org/C2780497538","wikidata":"https://www.wikidata.org/wiki/Q1192006","display_name":"Detoxification (alternative medicine)","level":3,"score":0.8198729157447815},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5559270977973938},{"id":"https://openalex.org/C29730261","wikidata":"https://www.wikidata.org/wiki/Q274160","display_name":"Toxicity","level":2,"score":0.44659629464149475},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.35532432794570923},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.33585458993911743},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.12098133563995361},{"id":"https://openalex.org/C71924100","wikidata":"https://www.wikidata.org/wiki/Q11190","display_name":"Medicine","level":0,"score":0.10928770899772644},{"id":"https://openalex.org/C142724271","wikidata":"https://www.wikidata.org/wiki/Q7208","display_name":"Pathology","level":1,"score":0.0},{"id":"https://openalex.org/C204787440","wikidata":"https://www.wikidata.org/wiki/Q188504","display_name":"Alternative medicine","level":2,"score":0.0},{"id":"https://openalex.org/C178790620","wikidata":"https://www.wikidata.org/wiki/Q11351","display_name":"Organic chemistry","level":1,"score":0.0},{"id":"https://openalex.org/C159110408","wikidata":"https://www.wikidata.org/wiki/Q121176","display_name":"Nursing","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.1013","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1013","pdf_url":"https://aclanthology.org/2025.findings-acl.1013.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.1013","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1013","pdf_url":"https://aclanthology.org/2025.findings-acl.1013.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":[{"id":"https://metadata.un.org/sdg/2","score":0.4099999964237213,"display_name":"Zero hunger"}],"awards":[{"id":"https://openalex.org/G6570941292","display_name":null,"funder_award_id":"62125201","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412887864.pdf","grobid_xml":"https://content.openalex.org/works/W4412887864.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W4387497383","https://openalex.org/W3183948672","https://openalex.org/W3173606202","https://openalex.org/W3110381201","https://openalex.org/W2948807893","https://openalex.org/W2935909890","https://openalex.org/W2778153218","https://openalex.org/W2758277628","https://openalex.org/W1531601525"],"abstract_inverted_index":{"Large":[0],"language":[1,6],"models":[2,49],"(LLMs)":[3],"exhibit":[4],"impressive":[5],"capabilities":[7,130],"but":[8],"remain":[9],"vulnerable":[10],"to":[11,82],"malicious":[12],"prompts":[13],"and":[14,127],"jailbreaking":[15],"attacks.Existing":[16],"knowledge":[17,63],"editing":[18,64],"methods":[19,43,122],"for":[20],"LLM":[21],"detoxification":[22,125],"face":[23],"two":[24],"major":[25],"challenges.First,":[26],"they":[27],"often":[28],"rely":[29],"on":[30,112],"entityspecific":[31],"localization,":[32],"making":[33],"them":[34],"ineffective":[35],"against":[36],"adversarial":[37],"inputs":[38],"without":[39],"explicit":[40],"entities.Second,":[41],"these":[42],"suffer":[44],"from":[45],"overediting,":[46],"where":[47],"detoxified":[48],"reject":[50],"legitimate":[51],"queries,":[52],"compromising":[53],"overall":[54],"performance.In":[55],"this":[56],"paper,":[57],"we":[58,100],"propose":[59],"TOXEDIT,":[60],"a":[61],"toxicity-aware":[62],"approach":[65],"that":[66,116],"dynamically":[67],"detects":[68],"toxic":[69],"activation":[70],"patterns":[71],"during":[72],"forward":[73],"propagation.It":[74],"then":[75],"routes":[76],"computations":[77],"through":[78],"adaptive":[79],"inter-layer":[80],"pathways":[81],"mitigate":[83],"toxicity":[84,89],"effectively.This":[85],"design":[86],"ensures":[87],"precise":[88],"mitigation":[90],"while":[91],"preserving":[92],"LLMs'":[93],"general":[94,129],"capabilities.To":[95],"more":[96],"accurately":[97],"assess":[98],"over-editing,":[99],"also":[101],"enhance":[102],"the":[103],"SafeEdit":[104],"benchmark":[105],"by":[106],"incorporating":[107],"instruction-following":[108],"evaluation":[109],"tasks.Experimental":[110],"results":[111],"multiple":[113],"LLMs":[114],"demonstrate":[115],"our":[117],"TOXEDIT":[118],"outperforms":[119],"previous":[120],"state-of-the-art":[121],"in":[123],"both":[124],"performance":[126],"safeguarding":[128],"of":[131],"LLMs.":[132]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}