{"id":"https://openalex.org/W4416034843","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.435","title":"Distributional Surgery for Language Model Activations","display_name":"Distributional Surgery for Language Model Activations","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416034843","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.435"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.435","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.435","pdf_url":"https://aclanthology.org/2025.findings-emnlp.435.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: EMNLP 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-emnlp.435.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5103431981","display_name":"Bao D. Nguyen","orcid":null},"institutions":[{"id":"https://openalex.org/I1333535994","display_name":"University of North Carolina Health Care","ror":"https://ror.org/00qz24g20","country_code":"US","type":"healthcare","lineage":["https://openalex.org/I1333535994"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK","US"],"is_corresponding":false,"raw_author_name":"Bao Nguyen","raw_affiliation_strings":["UNC-Chapel Hill CUHK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UNC-Chapel Hill CUHK","institution_ids":["https://openalex.org/I1333535994","https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5091142923","display_name":"Binh P. Nguyen","orcid":"https://orcid.org/0000-0001-6203-6664"},"institutions":[{"id":"https://openalex.org/I1333535994","display_name":"University of North Carolina Health Care","ror":"https://ror.org/00qz24g20","country_code":"US","type":"healthcare","lineage":["https://openalex.org/I1333535994"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK","US"],"is_corresponding":false,"raw_author_name":"Binh Nguyen","raw_affiliation_strings":["UNC-Chapel Hill CUHK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UNC-Chapel Hill CUHK","institution_ids":["https://openalex.org/I1333535994","https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101928156","display_name":"Duy Lap Nguyen","orcid":"https://orcid.org/0009-0001-4239-9174"},"institutions":[{"id":"https://openalex.org/I1333535994","display_name":"University of North Carolina Health Care","ror":"https://ror.org/00qz24g20","country_code":"US","type":"healthcare","lineage":["https://openalex.org/I1333535994"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK","US"],"is_corresponding":false,"raw_author_name":"Duy Nguyen","raw_affiliation_strings":["UNC-Chapel Hill CUHK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UNC-Chapel Hill CUHK","institution_ids":["https://openalex.org/I1333535994","https://openalex.org/I889458895"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100625770","display_name":"Viet Anh Nguyen","orcid":"https://orcid.org/0000-0002-2515-9494"},"institutions":[{"id":"https://openalex.org/I1333535994","display_name":"University of North Carolina Health Care","ror":"https://ror.org/00qz24g20","country_code":"US","type":"healthcare","lineage":["https://openalex.org/I1333535994"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["HK","US"],"is_corresponding":false,"raw_author_name":"Viet Anh Nguyen","raw_affiliation_strings":["UNC-Chapel Hill CUHK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UNC-Chapel Hill CUHK","institution_ids":["https://openalex.org/I1333535994","https://openalex.org/I889458895"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.38263555,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"8192","last_page":"8212"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10730","display_name":"Language Development and Disorders","score":0.19499999284744263,"subfield":{"id":"https://openalex.org/subfields/3204","display_name":"Developmental and Educational Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T10730","display_name":"Language Development and Disorders","score":0.19499999284744263,"subfield":{"id":"https://openalex.org/subfields/3204","display_name":"Developmental and Educational Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10465","display_name":"Neurobiology of Language and Bilingualism","score":0.050200000405311584,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.04989999905228615,"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/language-model","display_name":"Language model","score":0.43380001187324524},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.2816999852657318},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.27709999680519104},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.23520000278949738}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.527999997138977},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.45649999380111694},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.43380001187324524},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4147999882698059},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.28780001401901245},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2816999852657318},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.27709999680519104},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.26570001244544983},{"id":"https://openalex.org/C71924100","wikidata":"https://www.wikidata.org/wiki/Q11190","display_name":"Medicine","level":0,"score":0.2614000141620636},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.23520000278949738}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.435","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.435","pdf_url":"https://aclanthology.org/2025.findings-emnlp.435.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: EMNLP 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.435","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.435","pdf_url":"https://aclanthology.org/2025.findings-emnlp.435.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: EMNLP 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/W4416034843.pdf","grobid_xml":"https://content.openalex.org/works/W4416034843.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Language":[0],"models,":[1],"while":[2,91],"capable":[3],"of":[4,44,58,96,116],"generating":[5],"remarkably":[6],"coherent":[7],"and":[8,32,104],"seemingly":[9],"accurate":[10],"text,":[11],"can":[12],"occasionally":[13],"produce":[14],"undesirable":[15,34,49,64,117],"content":[16,35,50],"including":[17],"harmful":[18],"or":[19],"toxic":[20],"outputs.In":[21],"this":[22],"paper,":[23],"we":[24,40,66],"present":[25],"a":[26,55],"new":[27],"two-stage":[28],"approach":[29],"to":[30,47,85],"detect":[31,48],"mitigate":[33],"generations":[36],"by":[37,53],"rectifying":[38],"activations.First,":[39],"train":[41],"an":[42],"ensemble":[43],"layerwise":[45,68],"classifiers":[46],"using":[51],"activations":[52],"minimizing":[54],"smooth":[56],"surrogate":[57],"the":[59,74,88,94,97,114],"risk-aware":[60],"score.Then,":[61],"for":[62],"detected":[63],"contents,":[65],"propose":[67],"distributional":[69],"steering":[70],"policies":[71,77],"that":[72,107],"transform":[73],"attention":[75,89],"heads.These":[76],"are":[78],"computed":[79],"through":[80],"principled":[81],"semidefinite":[82],"programming":[83],"aims":[84],"minimally":[86],"perturb":[87],"distribution":[90],"probabilistically":[92],"guaranteeing":[93],"effectiveness":[95],"editions.Empirical":[98],"evaluations":[99],"across":[100],"multiple":[101],"language":[102],"models":[103],"datasets":[105],"show":[106],"our":[108],"method":[109],"outperforms":[110],"baselines":[111],"in":[112],"reducing":[113],"generation":[115],"output.":[118]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-11-08T00:00:00"}