{"id":"https://openalex.org/W7224838683","doi":"https://doi.org/10.48550/arxiv.2610.37842","title":"Improving Data-Constrained Scaling with Hyperparameter Transfer and Ensembling","display_name":"Improving Data-Constrained Scaling with Hyperparameter Transfer and Ensembling","publication_year":2026,"publication_date":"2026-10-27","ids":{"openalex":"https://openalex.org/W7224838683","doi":"https://doi.org/10.48550/arxiv.2610.37842"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.37842","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.37842","pdf_url":"https://arxiv.org/pdf/2610.37842","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.37842","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5115771698","display_name":"Ethan Baron","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I57206974","display_name":"New York University","ror":"https://ror.org/0190ak572","country_code":"US","type":"education","lineage":["https://openalex.org/I57206974"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Baron, Ethan","raw_affiliation_strings":["New York University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"New York University","institution_ids":["https://openalex.org/I57206974"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041128557","display_name":"Shmuel Berman","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I20089843","display_name":"Princeton University","ror":"https://ror.org/00hx57361","country_code":"US","type":"education","lineage":["https://openalex.org/I20089843"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Berman, Shmuel","raw_affiliation_strings":["Princeton University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Princeton University","institution_ids":["https://openalex.org/I20089843"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109022197","display_name":"Anirudh Satheesh","orcid":"https://orcid.org/0009-0004-9480-3601","observed_orcids":["https://orcid.org/0009-0004-9480-3601"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Satheesh, Anirudh","raw_affiliation_strings":["Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Labs","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119759511","display_name":"Akshay Vegesna","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Vegesna, Akshay","raw_affiliation_strings":["Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Labs","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088513490","display_name":"Samip Dahal","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dahal, Samip","raw_affiliation_strings":["Labs"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Labs","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101694421","display_name":"Andrew Gordon Wilson","orcid":"https://orcid.org/0000-0002-2011-3315","observed_orcids":["https://orcid.org/0000-0002-2011-3315"]},"institutions":[{"id":"https://openalex.org/I57206974","display_name":"New York University","ror":"https://ror.org/0190ak572","country_code":"US","type":"education","lineage":["https://openalex.org/I57206974"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wilson, Andrew Gordon","raw_affiliation_strings":["New York University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"New York University","institution_ids":["https://openalex.org/I57206974"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"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/T10028","display_name":"Topic Modeling","score":0.7598000168800354,"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":0.7598000168800354,"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":0.12160000205039978,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.02759999968111515,"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/hyperparameter-transfer","display_name":"hyperparameter transfer","score":0.8930000066757202},{"id":"https://openalex.org/keywords/ensembling","display_name":"ensembling","score":0.800000011920929},{"id":"https://openalex.org/keywords/large-language-model-training","display_name":"large language model training","score":0.7699999809265137},{"id":"https://openalex.org/keywords/scaling-laws","display_name":"scaling laws","score":0.7739999890327454},{"id":"https://openalex.org/keywords/power-law","display_name":"power law","score":0.8519999980926514},{"id":"https://openalex.org/keywords/weight-decay","display_name":"weight decay","score":0.9269999861717224},{"id":"https://openalex.org/keywords/training-data-efficiency","display_name":"training data efficiency","score":0.5680000185966492},{"id":"https://openalex.org/keywords/perplexity-reduction","display_name":"perplexity reduction","score":0.593999981880188}],"concepts":[{"id":"https://openalex.org/C8642999","wikidata":"https://www.wikidata.org/wiki/Q4171168","display_name":"Hyperparameter","level":2,"score":0.892799973487854},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.6996999979019165},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6276000142097473},{"id":"https://openalex.org/C72169020","wikidata":"https://www.wikidata.org/wiki/Q194404","display_name":"Monotonic function","level":2,"score":0.5896000266075134},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.4957999885082245},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.47679999470710754},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.4625999927520752},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.46239998936653137},{"id":"https://openalex.org/C150899416","wikidata":"https://www.wikidata.org/wiki/Q1820378","display_name":"Transfer of learning","level":2,"score":0.45739999413490295},{"id":"https://openalex.org/C2776175482","wikidata":"https://www.wikidata.org/wiki/Q1195816","display_name":"Transfer (computing)","level":2,"score":0.40230000019073486},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3562000095844269},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.35010001063346863},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.336899995803833},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3100000023841858},{"id":"https://openalex.org/C163258240","wikidata":"https://www.wikidata.org/wiki/Q25342","display_name":"Power (physics)","level":2,"score":0.3012999892234802},{"id":"https://openalex.org/C10485038","wikidata":"https://www.wikidata.org/wiki/Q48996162","display_name":"Hyperparameter optimization","level":3,"score":0.2939000129699707},{"id":"https://openalex.org/C2776289891","wikidata":"https://www.wikidata.org/wiki/Q1931511","display_name":"Neglect","level":2,"score":0.2856000065803528},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2718999981880188}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.37842","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.37842","pdf_url":"https://arxiv.org/pdf/2610.37842","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2610.37842","doi":"https://doi.org/10.48550/arxiv.2610.37842","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.37842","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,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2610.37842","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.37842","pdf_url":"https://arxiv.org/pdf/2610.37842","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7224838683.pdf","grobid_xml":"https://content.openalex.org/works/W7224838683.grobid-xml"},"referenced_works_count":12,"referenced_works":["https://openalex.org/W3001279689","https://openalex.org/W3006861283","https://openalex.org/W4398795990","https://openalex.org/W4412945484","https://openalex.org/W4415798387","https://openalex.org/W4417143336","https://openalex.org/W7133225605","https://openalex.org/W7135075179","https://openalex.org/W7160889183","https://openalex.org/W7163360564","https://openalex.org/W7171663499","https://openalex.org/W7196935208"],"related_works":["https://openalex.org/W7133188041","https://openalex.org/W7160889183","https://openalex.org/W4417082077","https://openalex.org/W4396821442","https://openalex.org/W7163919700","https://openalex.org/W7147008202","https://openalex.org/W7171934181","https://openalex.org/W4417088129","https://openalex.org/W4406793945","https://openalex.org/W7163200956"],"abstract_inverted_index":{"As":[0,77],"growing":[1],"compute":[2,19,90,116,148,168],"budgets":[3],"outpace":[4],"the":[5,50],"growth":[6],"of":[7,62],"high-quality":[8],"training":[9],"data,":[10],"large":[11],"language":[12],"models":[13],"must":[14],"be":[15],"trained":[16],"with":[17,47,57,89,106,117,147],"both":[18,94],"and":[20,60,108,143,161],"data":[21,75],"constraints":[22],"in":[23,141],"mind.":[24],"Unfortunately,":[25],"existing":[26],"prescriptions":[27,82,156],"for":[28,72],"data-constrained":[29],"scaling":[30,73,109],"suffer":[31],"from":[32,123],"two":[33],"important":[34],"limitations.":[35],"First,":[36],"past":[37,65],"works":[38,66],"hold":[39],"key":[40],"hyperparameters":[41,54,105],"such":[42],"as":[43],"weight":[44],"decay":[45],"fixed":[46],"scale,":[48],"despite":[49],"fact":[51],"that":[52,87,100,111],"optimal":[53],"vary":[55],"wildly":[56],"model":[58],"size":[59],"number":[61],"epochs.":[63],"Second,":[64],"neglect":[67],"ensembling,":[68],"a":[69,78,138],"promising":[70],"axis":[71],"under":[74],"constraints.":[76],"surprising":[79],"consequence,":[80],"prior":[81,136],"can":[83],"lead":[84],"to":[85,103,114,135,158,165],"loss":[86,129],"increases":[88],"budget.":[91],"We":[92],"address":[93],"limitations":[95],"by":[96,130],"fitting":[97],"power":[98],"laws":[99,110],"prescribe":[101,112],"how":[102,113],"transfer":[104],"scale":[107],"allocate":[115],"ensembling.":[118,150],"Training":[119],"on":[120],"100M":[121],"tokens":[122],"FineWeb,":[124],"our":[125],"recipe":[126],"reduces":[127],"validation":[128],"over":[131],"0.15":[132],"nats":[133],"compared":[134],"recipes,":[137],"14%":[139],"decrease":[140],"perplexity,":[142],"achieves":[144],"monotonic":[145],"improvements":[146],"through":[149],"These":[151],"gains":[152],"are":[153],"achieved":[154],"using":[155],"fit":[157],"smaller":[159],"runs":[160],"validated":[162],"at":[163],"up":[164],"12$\\times$":[166],"larger":[167],"budgets.":[169]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-31T00:00:00"}