{"id":"https://openalex.org/W7204265183","doi":"https://doi.org/10.17605/osf.io/vnxcy","title":"The Audit Regime Gap: Regime-Conditional Bias Evaluation in Large Language Models","display_name":"The Audit Regime Gap: Regime-Conditional Bias Evaluation in Large Language Models","publication_year":2026,"publication_date":"2026-08-26","ids":{"openalex":"https://openalex.org/W7204265183","doi":"https://doi.org/10.17605/osf.io/vnxcy"},"language":"en","primary_location":{"id":"pmh:doi:10.17605/osf.io/vnxcy","doi":null,"is_oa":true,"landing_page_url":"https://osf.io/vnxcy","pdf_url":null,"source":null,"license":"cc-by-nd","license_id":"https://openalex.org/licenses/cc-by-nd","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"StudyRegistration"},"volume":null,"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://osf.io/vnxcy","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124910313","display_name":"Noor Islam S. Mohammad","orcid":"https://orcid.org/0009-0007-8823-6681","observed_orcids":["https://orcid.org/0009-0007-8823-6681"]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Noor Islam S. Mohammad","raw_affiliation_strings":[],"raw_orcid":"https://orcid.org/0009-0007-8823-6681","affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5124910313"],"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":0.9735999703407288,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.9735999703407288,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.004699999932199717,"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":0.0032999999821186066,"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/islamophobia","display_name":"Islamophobia","score":0.7910000085830688},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.8309999704360962},{"id":"https://openalex.org/keywords/bias-measurement","display_name":"bias measurement","score":0.46000000834465027},{"id":"https://openalex.org/keywords/chain-of-thought-reasoning","display_name":"chain-of-thought reasoning","score":0.8489999771118164},{"id":"https://openalex.org/keywords/bias-amplification","display_name":"bias amplification","score":0.6890000104904175},{"id":"https://openalex.org/keywords/bias-suppression","display_name":"bias suppression","score":0.6439999938011169},{"id":"https://openalex.org/keywords/gpt-4o","display_name":"GPT-4o","score":0.7369999885559082},{"id":"https://openalex.org/keywords/mistral","display_name":"Mistral","score":0.6389999985694885}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.646399974822998},{"id":"https://openalex.org/C199521495","wikidata":"https://www.wikidata.org/wiki/Q181487","display_name":"Audit","level":2,"score":0.5554999709129333},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.49810001254081726},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.47920000553131104},{"id":"https://openalex.org/C50817715","wikidata":"https://www.wikidata.org/wiki/Q79895177","display_name":"Regret","level":2,"score":0.47530001401901245},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.42809998989105225},{"id":"https://openalex.org/C2777382242","wikidata":"https://www.wikidata.org/wiki/Q6017816","display_name":"Index (typography)","level":2,"score":0.41370001435279846},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.3682999908924103},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.36419999599456787},{"id":"https://openalex.org/C18619997","wikidata":"https://www.wikidata.org/wiki/Q47627","display_name":"Experimental economics","level":2,"score":0.3330000042915344},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.32829999923706055},{"id":"https://openalex.org/C40423286","wikidata":"https://www.wikidata.org/wiki/Q284172","display_name":"Selection bias","level":2,"score":0.32409998774528503},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.320499986410141},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.3140000104904175},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.31369999051094055},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.30970001220703125},{"id":"https://openalex.org/C100660578","wikidata":"https://www.wikidata.org/wiki/Q18733","display_name":"Recall","level":2,"score":0.29490000009536743},{"id":"https://openalex.org/C2779458634","wikidata":"https://www.wikidata.org/wiki/Q24963715","display_name":"Debiasing","level":2,"score":0.2922999858856201},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.289900004863739},{"id":"https://openalex.org/C77805123","wikidata":"https://www.wikidata.org/wiki/Q161272","display_name":"Social psychology","level":1,"score":0.28940001130104065},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.2888999879360199},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.28859999775886536},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.28130000829696655},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.2784000039100647},{"id":"https://openalex.org/C166052673","wikidata":"https://www.wikidata.org/wiki/Q83021","display_name":"Empirical evidence","level":2,"score":0.26989999413490295},{"id":"https://openalex.org/C174106493","wikidata":"https://www.wikidata.org/wiki/Q1057880","display_name":"External validity","level":2,"score":0.2635999917984009},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26350000500679016},{"id":"https://openalex.org/C134121241","wikidata":"https://www.wikidata.org/wiki/Q899301","display_name":"Yield (engineering)","level":2,"score":0.262800008058548},{"id":"https://openalex.org/C2777267654","wikidata":"https://www.wikidata.org/wiki/Q3519023","display_name":"Test (biology)","level":2,"score":0.2621000111103058},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.2603999972343445},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.25870001316070557},{"id":"https://openalex.org/C87007009","wikidata":"https://www.wikidata.org/wiki/Q210832","display_name":"Statistical hypothesis testing","level":2,"score":0.2513999938964844}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.17605/osf.io/vnxcy","doi":null,"is_oa":true,"landing_page_url":"https://osf.io/vnxcy","pdf_url":null,"source":null,"license":"cc-by-nd","license_id":"https://openalex.org/licenses/cc-by-nd","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"StudyRegistration"},{"id":"doi:10.17605/osf.io/vnxcy","doi":"https://doi.org/10.17605/osf.io/vnxcy","is_oa":true,"landing_page_url":"https://doi.org/10.17605/osf.io/vnxcy","pdf_url":null,"source":{"id":"https://openalex.org/S7407050956","display_name":"Open Science Framework","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I2799848540","host_organization_name":"Center for Open Science","host_organization_lineage":["https://openalex.org/I2799848540"],"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":"StudyRegistration"}],"best_oa_location":{"id":"pmh:doi:10.17605/osf.io/vnxcy","doi":null,"is_oa":true,"landing_page_url":"https://osf.io/vnxcy","pdf_url":null,"source":null,"license":"cc-by-nd","license_id":"https://openalex.org/licenses/cc-by-nd","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"StudyRegistration"},"sustainable_development_goals":[{"score":0.4399999976158142,"display_name":"Peace, justice, and strong institutions","id":"https://metadata.un.org/sdg/16"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7164912273","https://openalex.org/W7138943196","https://openalex.org/W3184144760","https://openalex.org/W7166878127","https://openalex.org/W7204997586","https://openalex.org/W7154985695","https://openalex.org/W7166803902","https://openalex.org/W7212033163","https://openalex.org/W4386185160","https://openalex.org/W4403706476"],"abstract_inverted_index":{"This":[0],"pre-registration":[1],"documents":[2],"the":[3,24,52,57,63,143,166,187,212,224,267,275],"experimental":[4],"protocol":[5],"for":[6,140],"\"The":[7],"Audit":[8],"Regime":[9],"Gap,\"":[10],"a":[11,38,41,74,117,123,263,272],"study":[12],"investigating":[13],"whether":[14],"anti-Muslim":[15],"bias":[16,36,257],"measurement":[17],"in":[18,72,81],"large":[19],"language":[20],"models":[21,155,202],"depends":[22],"on":[23],"inference":[25,46,93],"regime":[26],"used":[27],"(completion,":[28],"reasoning,":[29,103],"or":[30],"agentic":[31,105,249,279],"decision-making).":[32],"Purpose:":[33],"We":[34,68,238],"formalize":[35],"as":[37],"functional":[39],"of":[40,76,165,274],"model,":[42],"item":[43],"distribution,":[44],"and":[45,62,122,136,186,261,305],"regime,":[47],"deriving":[48],"three":[49],"estimable":[50],"quantities:":[51],"audit":[53,58,220,265],"validity":[54],"gap":[55],"(G),":[56],"transfer":[59,171],"coefficient":[60,320],"(\u03c4),":[61],"mitigation":[64],"misranking":[65],"index":[66,145],"(M).":[67],"instantiate":[69],"this":[70],"framework":[71],"MIRAGE,":[73],"benchmark":[75],"1,200":[77],"minimal-edit":[78],"matched-pair":[79],"prompts":[80],"five":[82],"identity":[83],"variants":[84],"(Muslim,":[85],"Christian,":[86],"Jewish,":[87],"Hindu,":[88],"secular),":[89],"evaluated":[90],"under":[91,253,278],"four":[92,109],"modes":[94],"(C\u2081":[95],"direct":[96],"completion,":[97],"C\u2082\u2090":[98],"elicited":[99],"chain-of-thought,":[100],"C\u2082_b":[101],"native":[102],"C\u2083":[104],"decision)":[106],"crossed":[107],"with":[108,116,181,203,211,223,296,303],"retrieval-recency":[110],"conditions":[111],"(empty,":[112],"neutral,":[113],"historical,":[114],"recent),":[115],"400-prompt":[118],"parallel":[119],"Arabic":[120],"subset":[121],"placebo-swap":[124],"null":[125],"calibration.":[126],"Primary":[127],"Hypotheses":[128],"(pre-registered):":[129],"H1":[130],"(Reasoning":[131],"amplification):":[132,198],"B_C\u2082\u2090":[133],">":[134,138,150,184],"B_C\u2081":[135,139],"B_C\u2082_b":[137,152],"all":[141],"models;":[142],"amplification":[144,213],"exceeds":[146,311],"1.":[147],"H2":[148],"(Native":[149],"elicited):":[151],"\u2265":[153],"B_C\u2082\u2090:":[154],"trained":[156],"to":[157,172],"reason":[158],"natively":[159],"produce":[160],"longer":[161],"intermediate":[162],"output":[163],"upstream":[164],"output-stage":[167],"suppressor.":[168],"H3":[169],"(Weak":[170],"decisions):":[173],"\u03c4_C\u2081\u2192C\u2083":[174],"is":[175,191],"substantially":[176],"below":[177],"\u03c4_C\u2081\u2192C\u2082\u2090.":[178],"H4":[179],"(Misranking":[180],"consequence):":[182],"M_C\u2081\u2192C\u2083":[183],"0":[185],"induced":[188],"selection":[189],"regret":[190],"strictly":[192],"positive.":[193],"H5":[194],"(Suppression":[195],"depth":[196],"predicts":[197],"In":[199],"reasoning-capable":[200],"open":[201],"probe":[204],"access,":[205],"later":[206],"suppression":[207],"onset":[208],"correlates":[209],"positively":[210],"index.":[214],"H6":[215],"(Recall-bound":[216],"agreement):":[217],"Empirical":[218],"top-q":[219],"recall":[221],"agrees":[222],"theoretical":[225],"bound":[226],"within":[227,316,328],"\u00b15":[228],"pp":[229],"at":[230],"q":[231],"\u2208":[232],"{0.05,":[233],"0.10,":[234],"0.20}.":[235],"Expected":[236],"Outcomes:":[237],"predict":[239],"that":[240,248,256,262],"reasoning":[241],"amplifies":[242],"rather":[243],"than":[244],"suppresses":[245],"Muslim\u2013violence":[246],"association,":[247],"decision":[250],"asymmetry":[251],"persists":[252],"prompt-based":[254],"defences,":[255],"tracks":[258],"retrieval":[259],"recency,":[260],"completion-only":[264],"inspecting":[266],"worst":[268,276],"decile":[269,277],"recovers":[270],"only":[271],"fraction":[273],"decision.":[280],"Models":[281],"Evaluated:":[282],"GPT-4o,":[283],"Claude":[284],"Opus":[285],"4,":[286],"Mistral":[287],"Large,":[288],"Llama-3.3-70B-Instruct,":[289],"Qwen2.5-72B-Instruct,":[290],"DeepSeek-R1.":[291],"Statistical":[292],"Protocol:":[293],"Paired":[294],"bootstrap":[295,309],"B=10,000":[297],"prompt-level":[298],"resamples":[299],"stratified":[300],"by":[301],"template,":[302],"bias-corrected":[304],"accelerated":[306],"adjustment":[307],"where":[308],"skewness":[310],"0.5.":[312],"Holm-Bonferroni":[313],"correction":[314],"applied":[315],"pre-declared":[317],"families.":[318],"Time-coupling":[319],"tested":[321],"via":[322],"permutation":[323],"test":[324],"shuffling":[325],"pool":[326],"labels":[327],"prompt.":[329]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-27T00:00:00"}