{"id":"https://openalex.org/W7155171866","doi":"https://doi.org/10.48550/arxiv.2604.18775","title":"An Empirical Study of Multi-Generation Sampling for Jailbreak Detection in Large Language Models","display_name":"An Empirical Study of Multi-Generation Sampling for Jailbreak Detection in Large Language Models","publication_year":2026,"publication_date":"2026-04-20","ids":{"openalex":"https://openalex.org/W7155171866","doi":"https://doi.org/10.48550/arxiv.2604.18775"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.18775","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.18775","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.18775","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5134268282","display_name":"Hanrui Luo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Luo, Hanrui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134252814","display_name":"Shreyank N Gowda","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gowda, Shreyank N","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":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/T10028","display_name":"Topic Modeling","score":0.12060000002384186,"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/T10028","display_name":"Topic Modeling","score":0.12060000002384186,"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/T13910","display_name":"Computational and Text Analysis Methods","score":0.08420000225305557,"subfield":{"id":"https://openalex.org/subfields/3300","display_name":"General Social Sciences"},"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/T12380","display_name":"Authorship Attribution and Profiling","score":0.06419999897480011,"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/detector","display_name":"Detector","score":0.6262999773025513},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.6259999871253967},{"id":"https://openalex.org/keywords/generator","display_name":"Generator (circuit theory)","score":0.6080999970436096},{"id":"https://openalex.org/keywords/sample","display_name":"Sample (material)","score":0.4742000102996826},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.46639999747276306},{"id":"https://openalex.org/keywords/empirical-research","display_name":"Empirical research","score":0.44449999928474426},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.382999986410141}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6978999972343445},{"id":"https://openalex.org/C94915269","wikidata":"https://www.wikidata.org/wiki/Q1834857","display_name":"Detector","level":2,"score":0.6262999773025513},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.6259999871253967},{"id":"https://openalex.org/C2780992000","wikidata":"https://www.wikidata.org/wiki/Q17016113","display_name":"Generator (circuit theory)","level":3,"score":0.6080999970436096},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49160000681877136},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.4742000102996826},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.46639999747276306},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.44449999928474426},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.39579999446868896},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.382999986410141},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3580000102519989},{"id":"https://openalex.org/C129848803","wikidata":"https://www.wikidata.org/wiki/Q2564360","display_name":"Sample size determination","level":2,"score":0.3560999929904938},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.35499998927116394},{"id":"https://openalex.org/C96608239","wikidata":"https://www.wikidata.org/wiki/Q1199823","display_name":"Statistical power","level":2,"score":0.3440000116825104},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3264000117778778},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C133199616","wikidata":"https://www.wikidata.org/wiki/Q25386885","display_name":"Empirical modelling","level":2,"score":0.2824999988079071},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.2791999876499176},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.27219998836517334},{"id":"https://openalex.org/C2776175482","wikidata":"https://www.wikidata.org/wiki/Q1195816","display_name":"Transfer (computing)","level":2,"score":0.271699994802475},{"id":"https://openalex.org/C134121241","wikidata":"https://www.wikidata.org/wiki/Q899301","display_name":"Yield (engineering)","level":2,"score":0.2517000138759613}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.18775","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.18775","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.18775","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.18775","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","score":0.7591114640235901,"id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Detecting":[0],"jailbreak":[1,30,74,173],"behaviour":[2],"in":[3,175],"large":[4,176],"language":[5,177],"models":[6,13,43],"remains":[7],"challenging,":[8],"particularly":[9],"when":[10,92],"strongly":[11],"aligned":[12],"produce":[14],"harmful":[15,85,149],"outputs":[16],"only":[17],"rarely.":[18],"In":[19],"this":[20],"work,":[21],"we":[22],"present":[23],"an":[24],"empirical":[25],"study":[26],"of":[27,80,139],"output":[28,70],"based":[29,59],"detection":[31,113,174],"under":[32],"realistic":[33],"conditions":[34],"using":[35],"the":[36,78],"JailbreakBench":[37],"Behaviors":[38],"dataset":[39],"and":[40,55,142,164,171],"multiple":[41],"generator":[42,109],"with":[44,119],"varying":[45],"alignment":[46],"strengths.":[47],"We":[48],"evaluate":[49],"both":[50],"a":[51,56,95,137,161],"lexical":[52,134],"TF-IDF":[53],"detector":[54,60],"generation":[57,97],"inconsistency":[58],"across":[61,117],"different":[62],"sampling":[63,103],"budgets.":[64],"Our":[65],"results":[66],"show":[67],"that":[68,112,133,155],"single":[69,96],"evaluation":[71],"systematically":[72],"underestimates":[73],"vulnerability,":[75],"as":[76],"increasing":[77],"number":[79],"sampled":[81],"generations":[82],"reveals":[83,132],"additional":[84],"behaviour.":[86,150],"The":[87],"most":[88],"significant":[89],"improvements":[90],"occur":[91],"moving":[93],"from":[94],"to":[98],"moderate":[99,156],"sampling,":[100],"while":[101],"larger":[102],"budgets":[104],"yield":[105],"diminishing":[106],"returns.":[107],"Cross":[108],"experiments":[110],"demonstrate":[111],"signals":[114,141],"partially":[115],"generalise":[116],"models,":[118],"stronger":[120],"transfer":[121],"observed":[122],"within":[123],"related":[124],"model":[125,169],"families.":[126],"A":[127],"category":[128],"level":[129],"analysis":[130],"further":[131],"detectors":[135],"capture":[136],"mixture":[138],"behavioural":[140],"topic":[143],"specific":[144],"cues,":[145],"rather":[146],"than":[147],"purely":[148],"Overall,":[151],"our":[152],"findings":[153],"suggest":[154],"multi":[157],"sample":[158],"auditing":[159],"provides":[160],"more":[162],"reliable":[163],"practical":[165],"approach":[166],"for":[167],"estimating":[168],"vulnerability":[170],"improving":[172],"models.":[178],"Code":[179],"will":[180],"be":[181],"released.":[182]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-23T00:00:00"}