{"id":"https://openalex.org/W7166115268","doi":"https://doi.org/10.48550/arxiv.2606.26836","title":"The Capability Frontier: Benchmarks Miss 82% of Model Performance","display_name":"The Capability Frontier: Benchmarks Miss 82% of Model Performance","publication_year":2026,"publication_date":"2026-06-25","ids":{"openalex":"https://openalex.org/W7166115268","doi":"https://doi.org/10.48550/arxiv.2606.26836"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.26836","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.26836","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.2606.26836","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5025616612","display_name":"Bradley Fowler","orcid":"https://orcid.org/0000-0002-8447-7898"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fowler, Bradley","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139401059","display_name":"Ryan Smith","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Smith, Ryan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139460084","display_name":"Daniel Thi Graviet","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Graviet, Daniel Thi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139422598","display_name":"William Myers","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Myers, William","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075146281","display_name":"Joshua Greaves","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Greaves, Joshua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119862711","display_name":"Narmeen Oozeer","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Oozeer, Narmeen Fatimah","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108557771","display_name":"A Alejandro Montes Garcia","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Garc\u00eda, Ant\u00eda","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139456146","display_name":"Philip Quirke","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Quirke, Philip","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065974640","display_name":"Amirali Abdullah","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Abdullah, Amirali","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139439929","display_name":"Fazl Barez","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Barez, Fazl","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5094196399","display_name":"Shriyash Upadhyay","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Upadhyay, Shriyash Kaustubh","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/T13702","display_name":"Machine Learning in Healthcare","score":0.4041000008583069,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.4041000008583069,"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/T11710","display_name":"Biomedical Text Mining and Ontologies","score":0.10750000178813934,"subfield":{"id":"https://openalex.org/subfields/1312","display_name":"Molecular Biology"},"field":{"id":"https://openalex.org/fields/13","display_name":"Biochemistry, Genetics and Molecular Biology"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.07930000126361847,"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/oracle","display_name":"Oracle","score":0.5983999967575073},{"id":"https://openalex.org/keywords/probabilistic-logic","display_name":"Probabilistic logic","score":0.526199996471405},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.4765999913215637},{"id":"https://openalex.org/keywords/pareto-principle","display_name":"Pareto principle","score":0.4397999942302704},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.4065999984741211},{"id":"https://openalex.org/keywords/entropy","display_name":"Entropy (arrow of time)","score":0.4011000096797943},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.3873000144958496},{"id":"https://openalex.org/keywords/model-selection","display_name":"Model selection","score":0.32580000162124634}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7134000062942505},{"id":"https://openalex.org/C55166926","wikidata":"https://www.wikidata.org/wiki/Q2892946","display_name":"Oracle","level":2,"score":0.5983999967575073},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.526199996471405},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.4765999913215637},{"id":"https://openalex.org/C137635306","wikidata":"https://www.wikidata.org/wiki/Q182667","display_name":"Pareto principle","level":2,"score":0.4397999942302704},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.4065999984741211},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.4011000096797943},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.3873000144958496},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.34459999203681946},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.32580000162124634},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2953000068664551},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.29339998960494995},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.29269999265670776},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.2863999903202057},{"id":"https://openalex.org/C51485801","wikidata":"https://www.wikidata.org/wiki/Q16966861","display_name":"Efficient frontier","level":3,"score":0.27459999918937683},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.26919999718666077},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.26570001244544983},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.26019999384880066},{"id":"https://openalex.org/C115051666","wikidata":"https://www.wikidata.org/wiki/Q6522493","display_name":"Ranging","level":2,"score":0.25940001010894775},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2590999901294708},{"id":"https://openalex.org/C58489278","wikidata":"https://www.wikidata.org/wiki/Q1172284","display_name":"Data set","level":2,"score":0.25540000200271606}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.26836","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.26836","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.2606.26836","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.26836","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Existing":[0],"benchmarks":[1,113],"typically":[2],"report":[3],"accuracy":[4,157],"for":[5,89,137,148,206],"a":[6,10,38,57,61,141,180],"single":[7,11,149,193],"model":[8],"on":[9],"run.":[12],"This":[13],"systematically":[14],"understates":[15],"real-world":[16],"LLM":[17,199],"capabilities,":[18],"particularly":[19],"under":[20,75],"heterogeneous":[21],"data":[22],"distributions:":[23],"(i)":[24],"different":[25,28],"models":[26,64,79],"get":[27],"questions":[29],"correct":[30],"according":[31],"to":[32,131,172],"their":[33],"specializations,":[34],"and":[35,45,80,97,121,190,208],"(ii)":[36],"given":[37],"budget,":[39],"multiple":[40],"generations":[41,81],"can":[42],"be":[43],"sampled":[44],"selectively":[46],"retained.":[47],"To":[48],"quantify":[49],"this":[50],"gap,":[51],"we":[52,167],"introduce":[53],"the":[54,67,184,191],"Capability":[55,125],"Frontier:":[56],"Pareto":[58],"frontier":[59],"over":[60,102],"set":[62],"of":[63],"that":[65,174],"characterizes":[66],"best":[68,192],"achievable":[69],"performance":[70,127,185],"at":[71,128,159],"each":[72,132],"cost":[73,130,161],"level":[74],"optimal":[76],"selection":[77],"across":[78,109],"(i.e.,":[82],"via":[83],"an":[84,152],"oracle).":[85],"Our":[86,195],"construction":[87],"corrects":[88],"two":[90],"opposing":[91],"biases:":[92],"underestimation":[93],"from":[94,99],"single-model":[95,138],"evaluation":[96,139,207],"overestimation":[98],"taking":[100],"maxima":[101],"noisy":[103],"samples.":[104],"We":[105],"study":[106],"21":[107],"LLMs":[108],"16":[110],"widely":[111],"used":[112],"spanning":[114],"coding,":[115],"reasoning,":[116],"medicine,":[117],"factuality,":[118],"instruction":[119],"following,":[120],"agentic":[122],"tasks,":[123],"comparing":[124],"Frontier":[126],"matched":[129,158],"benchmark's":[133],"top-performing":[134],"model.":[135,194],"Correcting":[136],"yields":[140,151],"54%":[142],"error":[143],"rate":[144],"reduction;":[145],"additionally":[146],"correcting":[147],"runs":[150],"82%":[153],"improvement,":[154],"with":[155,204],"SOTA":[156],"85%":[160],"reduction.":[162],"Complementing":[163],"these":[164],"empirical":[165],"results,":[166],"use":[168],"controlled":[169],"probabilistic":[170],"simulations":[171],"show":[173],"higher":[175],"query":[176],"topic":[177],"entropy":[178],"produces":[179],"near-monotonic":[181],"increase":[182],"in":[183,210],"gap":[186],"between":[187],"oracle":[188],"routing":[189],"findings":[196],"suggest":[197],"collective":[198],"capabilities":[200],"are":[201],"substantially":[202],"underestimated,":[203],"implications":[205],"deployment":[209],"data-heterogeneous,":[211],"multi-domain":[212],"settings.":[213]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-06-27T00:00:00"}