{"id":"https://openalex.org/W7162121979","doi":"https://doi.org/10.48550/arxiv.2605.21543","title":"Provable Joint Decontamination for Benchmarking Multiple Large Language Models","display_name":"Provable Joint Decontamination for Benchmarking Multiple Large Language Models","publication_year":2026,"publication_date":"2026-05-20","ids":{"openalex":"https://openalex.org/W7162121979","doi":"https://doi.org/10.48550/arxiv.2605.21543"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.21543","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21543","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"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.2605.21543","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5079652468","display_name":"Zhenlong Liu","orcid":"https://orcid.org/0000-0001-9149-7179"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Zhenlong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136733035","display_name":"Hao Zeng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Hao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136744036","display_name":"Hongxin Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Hongxin","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":0.22169999778270721,"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":0.22169999778270721,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.13539999723434448,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.11309999972581863,"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/benchmark","display_name":"Benchmark (surveying)","score":0.7835999727249146},{"id":"https://openalex.org/keywords/benchmarking","display_name":"Benchmarking","score":0.6549999713897705},{"id":"https://openalex.org/keywords/envelope","display_name":"Envelope (radar)","score":0.5454999804496765},{"id":"https://openalex.org/keywords/joint","display_name":"Joint (building)","score":0.5297999978065491},{"id":"https://openalex.org/keywords/conformal-map","display_name":"Conformal map","score":0.4602999985218048},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.4366999864578247},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.3666999936103821},{"id":"https://openalex.org/keywords/model-selection","display_name":"Model selection","score":0.36309999227523804}],"concepts":[{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.7835999727249146},{"id":"https://openalex.org/C86251818","wikidata":"https://www.wikidata.org/wiki/Q816754","display_name":"Benchmarking","level":2,"score":0.6549999713897705},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6204000115394592},{"id":"https://openalex.org/C65155139","wikidata":"https://www.wikidata.org/wiki/Q5380912","display_name":"Envelope (radar)","level":3,"score":0.5454999804496765},{"id":"https://openalex.org/C18555067","wikidata":"https://www.wikidata.org/wiki/Q8375051","display_name":"Joint (building)","level":2,"score":0.5297999978065491},{"id":"https://openalex.org/C98214594","wikidata":"https://www.wikidata.org/wiki/Q850275","display_name":"Conformal map","level":2,"score":0.4602999985218048},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.44589999318122864},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.4366999864578247},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.41929998993873596},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.40380001068115234},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.3666999936103821},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.36309999227523804},{"id":"https://openalex.org/C163258240","wikidata":"https://www.wikidata.org/wiki/Q25342","display_name":"Power (physics)","level":2,"score":0.34450000524520874},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2903999984264374},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2888999879360199},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2854999899864197},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.2791000008583069},{"id":"https://openalex.org/C96608239","wikidata":"https://www.wikidata.org/wiki/Q1199823","display_name":"Statistical power","level":2,"score":0.27810001373291016},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.2702000141143799},{"id":"https://openalex.org/C133462117","wikidata":"https://www.wikidata.org/wiki/Q4929239","display_name":"Data collection","level":2,"score":0.26649999618530273},{"id":"https://openalex.org/C27158222","wikidata":"https://www.wikidata.org/wiki/Q5532422","display_name":"Generalizability theory","level":2,"score":0.26409998536109924},{"id":"https://openalex.org/C67226441","wikidata":"https://www.wikidata.org/wiki/Q1665389","display_name":"Robust statistics","level":3,"score":0.2603999972343445},{"id":"https://openalex.org/C18653775","wikidata":"https://www.wikidata.org/wiki/Q1333358","display_name":"Joint probability distribution","level":2,"score":0.2522999942302704}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.21543","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21543","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.21543","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21543","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Benchmark":[0],"data":[1,18,53],"contamination":[2,115],"has":[3],"become":[4,33],"a":[5,48,51,70,98,109,136,148,164],"central":[6],"challenge":[7],"in":[8,15],"LLM":[9],"evaluation:":[10],"when":[11],"evaluation":[12],"examples":[13],"appear":[14],"the":[16,131,140,153,159,184,190],"training":[17],"of":[19,38,139],"one":[20],"or":[21],"more":[22],"audited":[23],"models,":[24],"reported":[25],"performance":[26],"can":[27,80],"be":[28],"inflated":[29],"and":[30,102,134,175],"cross-model":[31],"comparisons":[32],"unreliable.":[34],"A":[35],"broad":[36],"line":[37],"training-data":[39],"detection":[40],"work":[41],"designs":[42],"scores":[43],"to":[44,77,158],"quantify":[45],"how":[46],"strongly":[47],"model":[49,79],"memorizes":[50],"given":[52],"point,":[54],"but":[55],"these":[56],"score-based":[57],"methods":[58],"lack":[59],"theoretical":[60],"guarantees.":[61],"Recent":[62],"conformal":[63,110,126],"approaches":[64],"provide":[65],"provable":[66,167],"false-identification":[67],"control":[68,118],"for":[69],"single":[71],"model;":[72],"however,":[73],"applying":[74,152],"them":[75,129],"separately":[76],"each":[78],"produce":[81],"model-specific":[82],"benchmarks,":[83],"undermining":[84],"fair":[85],"comparison":[86],"across":[87,172],"models.":[88],"In":[89],"this":[90],"work,":[91],"we":[92,162],"formalize":[93],"multi-model":[94],"benchmark":[95,165],"decontamination":[96],"as":[97],"joint":[99],"selection":[100],"problem":[101],"propose":[103],"Joint":[104],"Envelope":[105],"Conformal":[106],"Selection":[107],"(JECS),":[108],"procedure":[111,157],"that":[112,178],"enables":[113],"global":[114],"rate":[116],"(GCR)":[117],"under":[119],"stated":[120],"assumptions.":[121],"Specifically,":[122],"JECS":[123,179],"computes":[124],"per-model":[125],"p-values,":[127],"aggregates":[128],"by":[130],"per-item":[132],"maximum,":[133],"reconstructs":[135],"conservative":[137],"envelope":[138],"max-p":[141,185],"null":[142],"distribution":[143],"from":[144],"right-tail":[145],"observations":[146],"above":[147],"data-driven":[149],"threshold.":[150],"By":[151],"adaptive":[154],"Benjamini-Hochberg":[155],"(BH)":[156],"envelope-rescaled":[160],"values,":[161],"select":[163],"with":[166],"GCR":[168,192],"control.":[169,193],"Extensive":[170],"experiments":[171],"various":[173],"models":[174],"benchmarks":[176],"demonstrate":[177],"achieves":[180],"higher":[181],"power":[182],"than":[183],"baseline":[186],"while":[187],"consistently":[188],"maintaining":[189],"target":[191]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-23T00:00:00"}