{"id":"https://openalex.org/W4414978685","doi":"https://doi.org/10.48550/arxiv.2510.05709","title":"Correcting Prompt Dependence in LLM Benchmarks: A Bayesian Hierarchical Model with Embedding-Space Clustering","display_name":"Correcting Prompt Dependence in LLM Benchmarks: A Bayesian Hierarchical Model with Embedding-Space Clustering","publication_year":2025,"publication_date":"2025-10-07","ids":{"openalex":"https://openalex.org/W4414978685","doi":"https://doi.org/10.48550/arxiv.2510.05709"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2510.05709","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2510.05709","pdf_url":"https://arxiv.org/pdf/2510.05709","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2510.05709","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5075072583","display_name":"Mary Llewellyn","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Llewellyn, Mary","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Thornton, Isobel","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Thornton, Isobel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Bishop, James","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bishop, James","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5032458943","display_name":"Annie Gray","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gray, Annie","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":true,"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/T13999","display_name":"Digital Rights Management and Security","score":0.9291999936103821,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T13999","display_name":"Digital Rights Management and Security","score":0.9291999936103821,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/transformer","display_name":"Transformer","score":0.5645999908447266},{"id":"https://openalex.org/keywords/heuristic","display_name":"Heuristic","score":0.5218999981880188},{"id":"https://openalex.org/keywords/bayesian-probability","display_name":"Bayesian probability","score":0.45170000195503235},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.4043000042438507},{"id":"https://openalex.org/keywords/bayesian-inference","display_name":"Bayesian inference","score":0.3813999891281128},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.26809999346733093}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6966999769210815},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5645999908447266},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.5218999981880188},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.498199999332428},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.45170000195503235},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.4043000042438507},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.382999986410141},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.3813999891281128},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.32499998807907104},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.29100000858306885},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.26809999346733093},{"id":"https://openalex.org/C12174686","wikidata":"https://www.wikidata.org/wiki/Q1058438","display_name":"Risk assessment","level":2,"score":0.26159998774528503},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.25429999828338623},{"id":"https://openalex.org/C111065885","wikidata":"https://www.wikidata.org/wiki/Q1189053","display_name":"Fuzz testing","level":3,"score":0.2533999979496002},{"id":"https://openalex.org/C2777267654","wikidata":"https://www.wikidata.org/wiki/Q3519023","display_name":"Test (biology)","level":2,"score":0.2515999972820282}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2510.05709","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2510.05709","pdf_url":"https://arxiv.org/pdf/2510.05709","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2510.05709","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.05709","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":"pmh:oai:arXiv.org:2510.05709","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2510.05709","pdf_url":"https://arxiv.org/pdf/2510.05709","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4414978685.pdf","grobid_xml":"https://content.openalex.org/works/W4414978685.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"LLM":[0],"benchmarking":[1],"metrics":[2,52],"often":[3],"misstate":[4],"performance":[5,51,79],"and":[6,32,88],"uncertainty":[7],"as":[8],"they":[9],"rely":[10],"on":[11],"two":[12],"assumptions":[13],"that":[14,48],"frequently":[15],"do":[16],"not":[17],"hold":[18],"in":[19,53,76],"practice:":[20],"(i)":[21],"a":[22,40],"sufficient":[23],"number":[24],"of":[25,72],"evaluations":[26],"are":[27,36],"available":[28],"for":[29,58],"classical":[30],"inference,":[31],"(ii)":[33],"test":[34],"prompts":[35],"independent.":[37],"We":[38,61],"propose":[39],"corrective":[41],"Bayesian":[42],"hierarchical":[43],"model":[44],"with":[45,81],"embedding-space":[46],"clustering":[47,73],"provides":[49],"robust":[50],"limited-data":[54],"settings":[55],"while":[56],"correcting":[57],"prompt":[59],"dependence.":[60],"apply":[62],"the":[63],"approach":[64],"to":[65,84,92],"adversarial":[66],"robustness":[67],"benchmarks,":[68],"showing":[69],"consistent":[70],"recovery":[71],"structure,":[74],"resulting":[75],"more":[77],"reliable":[78],"metrics,":[80],"4-73%":[82],"improvements":[83,91],"mean":[85],"absolute":[86],"errors":[87],"40-450":[89],"unit":[90],"expected":[93],"log":[94],"posterior":[95],"densities.":[96]},"counts_by_year":[],"updated_date":"2026-08-11T07:18:39.950985","created_date":"2025-10-10T00:00:00"}