{"id":"https://openalex.org/W7128693191","doi":"https://doi.org/10.48448/t9aq-j125","title":"Silenced Biases: The Dark Side LLMs Learned to Refuse","display_name":"Silenced Biases: The Dark Side LLMs Learned to Refuse","publication_year":2026,"publication_date":"2026-01-07","ids":{"openalex":"https://openalex.org/W7128693191","doi":"https://doi.org/10.48448/t9aq-j125"},"language":"en","primary_location":{"id":"doi:10.48448/t9aq-j125","doi":"https://doi.org/10.48448/t9aq-j125","is_oa":true,"landing_page_url":"https://doi.org/10.48448/t9aq-j125","pdf_url":null,"source":{"id":"https://openalex.org/S7407053148","display_name":"Underline Science Inc.","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Audiovisual"},"volume":null,"type":"other","is_traditional":false,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48448/t9aq-j125","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5125741655","display_name":"Association for Artificial Intelligence 2026","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Association for Artificial Intelligence 2026","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5116293067","display_name":"Rom Himelstein","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Himelstein, Rom","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124977942","display_name":"Amit LeVi","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I174306211","display_name":"Technion \u2013 Israel Institute of Technology","ror":"https://ror.org/03qryx823","country_code":"IL","type":"education","lineage":["https://openalex.org/I174306211"]}],"countries":["IL"],"is_corresponding":false,"raw_author_name":"LeVi, Amit","raw_affiliation_strings":["Technion"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technion","institution_ids":["https://openalex.org/I174306211"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125754277","display_name":"Avi Mendelson","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mendelson, Avi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066182660","display_name":"Yaniv Nemcovsky","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nemcovsky, Yaniv","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5026215174","display_name":"Brit Youngmann","orcid":"https://orcid.org/0000-0002-0031-5550","observed_orcids":["https://orcid.org/0000-0002-0031-5550"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Youngmann, Brit","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.9940999746322632,"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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.9940999746322632,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.0010000000474974513,"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/T10028","display_name":"Topic Modeling","score":0.0007999999797903001,"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/large-language-models","display_name":"large language models","score":0.8029999732971191},{"id":"https://openalex.org/keywords/fairness-evaluation","display_name":"fairness evaluation","score":0.8180000185966492},{"id":"https://openalex.org/keywords/safety-alignment","display_name":"safety alignment","score":0.7419999837875366},{"id":"https://openalex.org/keywords/activation-steering","display_name":"activation steering","score":0.777999997138977},{"id":"https://openalex.org/keywords/demographic-bias","display_name":"demographic bias","score":0.414000004529953},{"id":"https://openalex.org/keywords/question-answering","display_name":"question answering","score":0.3370000123977661}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6165000200271606},{"id":"https://openalex.org/C2777402240","wikidata":"https://www.wikidata.org/wiki/Q6783436","display_name":"Masking (illustration)","level":2,"score":0.5701000094413757},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5507000088691711},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.5217000246047974},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5202000141143799},{"id":"https://openalex.org/C99394141","wikidata":"https://www.wikidata.org/wiki/Q1085203","display_name":"Great Rift","level":2,"score":0.4821000099182129},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.41609999537467957},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.39820000529289246},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.37610000371932983},{"id":"https://openalex.org/C77805123","wikidata":"https://www.wikidata.org/wiki/Q161272","display_name":"Social psychology","level":1,"score":0.374099999666214},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.3666999936103821},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.3610999882221222},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.3402999937534332},{"id":"https://openalex.org/C2780598303","wikidata":"https://www.wikidata.org/wiki/Q65921492","display_name":"Flexibility (engineering)","level":2,"score":0.326200008392334},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.32170000672340393},{"id":"https://openalex.org/C2781118332","wikidata":"https://www.wikidata.org/wiki/Q430460","display_name":"Capability approach","level":2,"score":0.31299999356269836},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.2671999931335449},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.26579999923706055},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.26159998774528503},{"id":"https://openalex.org/C108827166","wikidata":"https://www.wikidata.org/wiki/Q175975","display_name":"Internet privacy","level":1,"score":0.2567000091075897},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.25619998574256897},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.25200000405311584},{"id":"https://openalex.org/C195094911","wikidata":"https://www.wikidata.org/wiki/Q14167904","display_name":"Process management","level":1,"score":0.25049999356269836}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48448/t9aq-j125","doi":"https://doi.org/10.48448/t9aq-j125","is_oa":true,"landing_page_url":"https://doi.org/10.48448/t9aq-j125","pdf_url":null,"source":{"id":"https://openalex.org/S7407053148","display_name":"Underline Science Inc.","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Audiovisual"}],"best_oa_location":{"id":"doi:10.48448/t9aq-j125","doi":"https://doi.org/10.48448/t9aq-j125","is_oa":true,"landing_page_url":"https://doi.org/10.48448/t9aq-j125","pdf_url":null,"source":{"id":"https://openalex.org/S7407053148","display_name":"Underline Science Inc.","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Audiovisual"},"sustainable_development_goals":[{"display_name":"Reduced inequalities","score":0.7578999996185303,"id":"https://metadata.un.org/sdg/10"},{"display_name":"Peace, justice, and strong institutions","score":0.4399999976158142,"id":"https://metadata.un.org/sdg/16"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7128732841","https://openalex.org/W7152934385","https://openalex.org/W4411113090","https://openalex.org/W7168998171","https://openalex.org/W4403577306","https://openalex.org/W4416034919","https://openalex.org/W4400720077","https://openalex.org/W4414855969","https://openalex.org/W4415273677","https://openalex.org/W7171809955"],"abstract_inverted_index":{"Safety-aligned":[0],"large":[1],"language":[2],"models":[3,29,168],"(LLMs)":[4],"are":[5,80,89],"becoming":[6],"increasingly":[7],"widespread,":[8],"especially":[9],"in":[10],"sensitive":[11],"applications":[12],"where":[13,185],"fairness":[14,27,60,158,199],"is":[15,30],"essential":[16],"and":[17,34,88,114,154,169,196],"biased":[18],"outputs":[19],"can":[20],"cause":[21],"significant":[22],"harm.":[23],"However,":[24],"evaluating":[25],"the":[26,54,74,117,125,163,172],"of":[28,67,76,166,175],"a":[31,64,157],"complex":[32],"challenge,":[33],"approaches":[35,95],"that":[36,96,161],"do":[37],"so":[38],"typically":[39],"utilize":[40],"standard":[41],"question-answer":[42],"(QA)":[43],"styled":[44],"schemes.":[45],"Such":[46],"methods":[47],"often":[48,101],"overlook":[49],"deeper":[50],"issues":[51],"by":[52,92,136],"interpreting":[53],"model's":[55],"refusal":[56],"responses":[57,195],"as":[58],"positive":[59],"measurements,":[61],"which":[62,79,110,130],"creates":[63],"false":[65],"sense":[66],"fairness.":[68],"In":[69],"this":[70],"work,":[71],"we":[72],"introduce":[73],"concept":[75],"silenced":[77],"biases,":[78],"unfair":[81],"preferences":[82],"encoded":[83],"within":[84],"models'":[85,193],"latent":[86],"space":[87],"effectively":[90],"concealed":[91],"safety-alignment.":[93],"Previous":[94],"considered":[97],"similar":[98],"indirect":[99],"biases":[100,135],"relied":[102],"on":[103],"prompt":[104],"manipulation":[105],"or":[106],"handcrafted":[107],"implicit":[108],"queries,":[109],"present":[111],"limited":[112],"scalability":[113],"risk":[115],"contaminating":[116],"evaluation":[118,159],"process":[119],"with":[120],"additional":[121],"biases.":[122],"We":[123,178],"propose":[124],"Silenced":[126],"Bias":[127],"Benchmark":[128],"(SBB),":[129],"aims":[131],"to":[132,140,150],"uncover":[133],"these":[134],"employing":[137],"activation":[138],"steering":[139],"reduce":[141],"model":[142],"refusals":[143],"during":[144],"QA.":[145],"SBB":[146],"supports":[147],"easy":[148],"expansion":[149],"new":[151],"demographic":[152],"groups":[153],"subjects,":[155],"presenting":[156],"framework":[160],"encourages":[162],"future":[164],"development":[165],"fair":[167],"tools":[170],"beyond":[171],"masking":[173],"effects":[174],"alignment":[176],"training.":[177],"demonstrate":[179],"our":[180,186],"approach":[181],"over":[182],"multiple":[183],"LLMs,":[184],"findings":[187],"expose":[188],"an":[189],"alarming":[190],"distinction":[191],"between":[192],"direct":[194],"their":[197],"underlying":[198],"issues.":[200]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-02-13T00:00:00"}