{"id":"https://openalex.org/W7223114658","doi":"https://doi.org/10.48550/arxiv.2610.28036","title":"When Does Training on Synthetic Data Provably Work?","display_name":"When Does Training on Synthetic Data Provably Work?","publication_year":2026,"publication_date":"2026-10-20","ids":{"openalex":"https://openalex.org/W7223114658","doi":"https://doi.org/10.48550/arxiv.2610.28036"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.28036","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.28036","pdf_url":"https://arxiv.org/pdf/2610.28036","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/2610.28036","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5159167654","display_name":"Leroy Odunlami","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4577782","display_name":"Technische Universit\u00e4t Berlin","ror":"https://ror.org/03v4gjf40","country_code":"DE","type":"education","lineage":["https://openalex.org/I4577782"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Odunlami, Leroy","raw_affiliation_strings":["Institute of Mathematics , Technische Universit\u00e4t Berlin Berlin , Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Mathematics , Technische Universit\u00e4t Berlin Berlin , Germany","institution_ids":["https://openalex.org/I4577782"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017760654","display_name":"Tobias Riedlinger","orcid":"https://orcid.org/0000-0002-1953-8607","observed_orcids":["https://orcid.org/0000-0002-1953-8607"]},"institutions":[{"id":"https://openalex.org/I4577782","display_name":"Technische Universit\u00e4t Berlin","ror":"https://ror.org/03v4gjf40","country_code":"DE","type":"education","lineage":["https://openalex.org/I4577782"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Riedlinger, Tobias","raw_affiliation_strings":["Institute of Mathematics , Technische Universit\u00e4t Berlin Berlin , Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Mathematics , Technische Universit\u00e4t Berlin Berlin , Germany","institution_ids":["https://openalex.org/I4577782"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5029365974","display_name":"Hanno Gottschalk","orcid":"https://orcid.org/0000-0003-2167-2028","observed_orcids":["https://orcid.org/0000-0003-2167-2028"]},"institutions":[{"id":"https://openalex.org/I4577782","display_name":"Technische Universit\u00e4t Berlin","ror":"https://ror.org/03v4gjf40","country_code":"DE","type":"education","lineage":["https://openalex.org/I4577782"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Gottschalk, Hanno","raw_affiliation_strings":["Institute of Mathematics , Technische Universit\u00e4t Berlin Berlin , Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Mathematics , Technische Universit\u00e4t Berlin Berlin , Germany","institution_ids":["https://openalex.org/I4577782"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I4577782"],"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.6269999742507935,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.6269999742507935,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.11940000206232071,"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":0.10109999775886536,"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/synthetic-data","display_name":"synthetic data","score":0.875},{"id":"https://openalex.org/keywords/domain-generalization","display_name":"domain generalization","score":0.7250000238418579},{"id":"https://openalex.org/keywords/binary-classification","display_name":"binary classification","score":0.5149999856948853},{"id":"https://openalex.org/keywords/finite-sample-analysis","display_name":"finite-sample analysis","score":0.6330000162124634},{"id":"https://openalex.org/keywords/density-ratio","display_name":"density ratio","score":0.5879999995231628},{"id":"https://openalex.org/keywords/singular-distributions","display_name":"singular distributions","score":0.460999995470047},{"id":"https://openalex.org/keywords/wasserstein-distance","display_name":"Wasserstein distance","score":0.9520000219345093},{"id":"https://openalex.org/keywords/generative-models","display_name":"generative models","score":0.6119999885559082},{"id":"https://openalex.org/keywords/adversarial-image-generation","display_name":"adversarial image generation","score":0.45100000500679016},{"id":"https://openalex.org/keywords/model-validation","display_name":"model validation","score":0.4339999854564667}],"concepts":[{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.885200023651123},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5918999910354614},{"id":"https://openalex.org/C95623464","wikidata":"https://www.wikidata.org/wiki/Q1096149","display_name":"Classifier (UML)","level":2,"score":0.5878000259399414},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5188000202178955},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.454800009727478},{"id":"https://openalex.org/C43555835","wikidata":"https://www.wikidata.org/wiki/Q2300258","display_name":"Conditional probability distribution","level":2,"score":0.45339998602867126},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.45210000872612},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.4242999851703644},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.423799991607666},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.42329999804496765},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4180000126361847},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3874000012874603},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.36730000376701355},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3499999940395355},{"id":"https://openalex.org/C66905080","wikidata":"https://www.wikidata.org/wiki/Q17005494","display_name":"Binary classification","level":3,"score":0.3382999897003174},{"id":"https://openalex.org/C3020493868","wikidata":"https://www.wikidata.org/wiki/Q55631277","display_name":"Real world data","level":2,"score":0.33469998836517334},{"id":"https://openalex.org/C16910744","wikidata":"https://www.wikidata.org/wiki/Q7705759","display_name":"Test data","level":2,"score":0.31439998745918274},{"id":"https://openalex.org/C79772020","wikidata":"https://www.wikidata.org/wiki/Q5159264","display_name":"Conditional independence","level":2,"score":0.3116999864578247},{"id":"https://openalex.org/C21080849","wikidata":"https://www.wikidata.org/wiki/Q13611879","display_name":"Data point","level":2,"score":0.3012000024318695},{"id":"https://openalex.org/C44492722","wikidata":"https://www.wikidata.org/wiki/Q327069","display_name":"Conditional probability","level":2,"score":0.3005000054836273},{"id":"https://openalex.org/C197055811","wikidata":"https://www.wikidata.org/wiki/Q207522","display_name":"Probability density function","level":2,"score":0.296099990606308},{"id":"https://openalex.org/C2779190172","wikidata":"https://www.wikidata.org/wiki/Q4913888","display_name":"Binary data","level":3,"score":0.2897999882698059},{"id":"https://openalex.org/C119043178","wikidata":"https://www.wikidata.org/wiki/Q320723","display_name":"Covariate","level":2,"score":0.289000004529953},{"id":"https://openalex.org/C87007009","wikidata":"https://www.wikidata.org/wiki/Q210832","display_name":"Statistical hypothesis testing","level":2,"score":0.2800000011920929},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.26989999413490295},{"id":"https://openalex.org/C197096303","wikidata":"https://www.wikidata.org/wiki/Q869887","display_name":"Probability mass function","level":3,"score":0.26159998774528503},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.2572999894618988},{"id":"https://openalex.org/C186215838","wikidata":"https://www.wikidata.org/wiki/Q772232","display_name":"Conditional expectation","level":2,"score":0.25699999928474426},{"id":"https://openalex.org/C2781170535","wikidata":"https://www.wikidata.org/wiki/Q30587856","display_name":"Noisy data","level":2,"score":0.2549000084400177}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.28036","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.28036","pdf_url":"https://arxiv.org/pdf/2610.28036","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.2610.28036","doi":"https://doi.org/10.48550/arxiv.2610.28036","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.28036","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:2610.28036","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.28036","pdf_url":"https://arxiv.org/pdf/2610.28036","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":[{"id":"https://openalex.org/F4320323803","display_name":"Bundesministerium f\u00fcr Wirtschaft und Energie","ror":"https://ror.org/02vgg2808"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7223114658.pdf","grobid_xml":"https://content.openalex.org/works/W7223114658.grobid-xml"},"referenced_works_count":21,"referenced_works":["https://openalex.org/W1498436455","https://openalex.org/W1564947197","https://openalex.org/W1959608418","https://openalex.org/W2097117768","https://openalex.org/W2104094955","https://openalex.org/W2112796928","https://openalex.org/W2752782242","https://openalex.org/W2947121476","https://openalex.org/W2963163009","https://openalex.org/W2963446712","https://openalex.org/W2981985696","https://openalex.org/W3128484727","https://openalex.org/W3138516171","https://openalex.org/W3177349073","https://openalex.org/W4321760728","https://openalex.org/W4362496454","https://openalex.org/W4362641980","https://openalex.org/W4377111665","https://openalex.org/W4394861491","https://openalex.org/W4400949264","https://openalex.org/W4416383082"],"related_works":["https://openalex.org/W4226008296","https://openalex.org/W4416383082","https://openalex.org/W7134265814","https://openalex.org/W4387725201","https://openalex.org/W1992697579","https://openalex.org/W3201972112","https://openalex.org/W4327653014","https://openalex.org/W2735740596","https://openalex.org/W7171676053","https://openalex.org/W2947121476"],"abstract_inverted_index":{"High-quality":[0],"training":[1],"data":[2,26,106,224],"are":[3,120,139],"a":[4,20,30,34,60,98,143,151,163],"major":[5],"bottleneck":[6],"in":[7,123,145,167,175],"machine":[8],"learning,":[9],"motivating":[10],"increasing":[11],"use":[12],"of":[13,134,147,158],"artificially":[14],"generated":[15,183],"data.":[16,237],"We":[17],"study":[18,66],"when":[19,45,117],"binary":[21],"classifier":[22,99,196],"trained":[23,213],"on":[24,104,111,180,200,207,235],"synthetic":[25,57,81,90,105,160,202,218],"transfers":[27],"successfully":[28],"to":[29,55,186],"real":[31,47,75,114,152,187,209,220,236],"distribution,":[32],"under":[33,59],"shared":[35],"conditional":[36,136],"label":[37,137],"distribution.":[38],"First,":[39],"we":[40,65,141],"derive":[41],"finite-sample":[42],"excess-error":[43],"bounds":[44],"the":[46,56,67,74,80,84,112,118,132,135,148,156,159,168,201,208],"covariate":[48],"distribution":[49,58,76],"is":[50,87],"absolutely":[51],"continuous":[52],"with":[53],"respect":[54],"controlled":[61,176],"density":[62],"ratio.":[63],"Second,":[64],"partially":[68],"and":[69,131,179,219,232],"fully":[70,169],"singular":[71,170],"case,":[72],"where":[73,129],"has":[77],"mass":[78],"outside":[79,155],"one.":[82],"If":[83],"hypothesis":[85],"class":[86],"sufficiently":[88],"expressive,":[89],"risk":[91],"alone":[92],"cannot":[93],"uniformly":[94],"control":[95],"real-world":[96],"error:":[97],"can":[100],"be":[101,226],"nearly":[102],"optimal":[103],"while":[107,211],"approaching":[108],"worst-case":[109],"error":[110,199,206],"uncovered":[113],"region,":[115],"even":[116,166],"distributions":[119],"arbitrarily":[121],"close":[122],"Wasserstein":[124],"distance.":[125],"For":[126],"function":[127],"classes":[128],"hypotheses":[130],"log-odds":[133],"probability":[138,149],"Lipschitz,":[140],"recover":[142],"bound":[144],"terms":[146],"that":[150],"sample":[153,230],"lies":[154],"reach":[157],"data,":[161],"restoring":[162],"transfer":[164],"guarantee":[165],"case.":[171],"Experiments":[172],"confirm":[173],"this":[174],"low-dimensional":[177],"settings":[178],"32,200":[181],"images":[182,189],"as":[184],"counterparts":[185],"ImageNet":[188],"by":[190],"four":[191],"state-of-the-art":[192],"generators:":[193],"an":[194],"adversarial":[195],"reaches":[197],"0.5%":[198],"domain":[203],"but":[204],"99.25%":[205],"domain,":[210],"ordinarily":[212],"architectures":[214],"show":[215],"positively":[216],"correlated":[217],"test":[221],"errors.":[222],"Synthetic":[223],"should":[225],"budgeted":[227],"at":[228],"high":[229],"counts":[231],"models":[233],"validated":[234]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-22T00:00:00"}