{"id":"https://openalex.org/W4416645798","doi":"https://doi.org/10.48550/arxiv.2506.17989","title":"Data Curation Matters: Model Collapse and Spurious Shift Performance Prediction from Training on Uncurated Text Embeddings","display_name":"Data Curation Matters: Model Collapse and Spurious Shift Performance Prediction from Training on Uncurated Text Embeddings","publication_year":2025,"publication_date":"2025-06-22","ids":{"openalex":"https://openalex.org/W4416645798","doi":"https://doi.org/10.48550/arxiv.2506.17989"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2506.17989","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2506.17989","pdf_url":"https://arxiv.org/pdf/2506.17989","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2506.17989","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5093988909","display_name":"Lucas Mattioli","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I3018083178","display_name":"Institut de Recherche Technologique SystemX","ror":"https://ror.org/03crmsn52","country_code":"FR","type":"facility","lineage":["https://openalex.org/I3018083178"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Mattioli, Lucas","raw_affiliation_strings":["IRT SystemX (Institut de recherche technologique SystemX - 8 Avenue de la Vauve, 91120 Palaiseau - France)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IRT SystemX (Institut de recherche technologique SystemX - 8 Avenue de la Vauve, 91120 Palaiseau - France)","institution_ids":["https://openalex.org/I3018083178"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120518178","display_name":"Youness Ait-Hadichou","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I3018083178","display_name":"Institut de Recherche Technologique SystemX","ror":"https://ror.org/03crmsn52","country_code":"FR","type":"facility","lineage":["https://openalex.org/I3018083178"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Hadichou, Youness Ait","raw_affiliation_strings":["IRT SystemX (Institut de recherche technologique SystemX - 8 Avenue de la Vauve, 91120 Palaiseau - France)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IRT SystemX (Institut de recherche technologique SystemX - 8 Avenue de la Vauve, 91120 Palaiseau - France)","institution_ids":["https://openalex.org/I3018083178"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029900412","display_name":"Sabrina Chaouche","orcid":"https://orcid.org/0009-0005-2638-4182","observed_orcids":["https://orcid.org/0009-0005-2638-4182"]},"institutions":[{"id":"https://openalex.org/I3018083178","display_name":"Institut de Recherche Technologique SystemX","ror":"https://ror.org/03crmsn52","country_code":"FR","type":"facility","lineage":["https://openalex.org/I3018083178"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Chaouche, Sabrina","raw_affiliation_strings":["IRT SystemX (Institut de recherche technologique SystemX - 8 Avenue de la Vauve, 91120 Palaiseau - France)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IRT SystemX (Institut de recherche technologique SystemX - 8 Avenue de la Vauve, 91120 Palaiseau - France)","institution_ids":["https://openalex.org/I3018083178"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101788847","display_name":"Martin Gonzalez","orcid":"https://orcid.org/0000-0002-1176-2022","observed_orcids":["https://orcid.org/0000-0002-1176-2022"]},"institutions":[{"id":"https://openalex.org/I3018083178","display_name":"Institut de Recherche Technologique SystemX","ror":"https://ror.org/03crmsn52","country_code":"FR","type":"facility","lineage":["https://openalex.org/I3018083178"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Gonzalez, Martin","raw_affiliation_strings":["IRT SystemX (Institut de recherche technologique SystemX - 8 Avenue de la Vauve, 91120 Palaiseau - France)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IRT SystemX (Institut de recherche technologique SystemX - 8 Avenue de la Vauve, 91120 Palaiseau - France)","institution_ids":["https://openalex.org/I3018083178"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I3018083178"],"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":0.5752000212669373,"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":0.5752000212669373,"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":0.1776999980211258,"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/T11273","display_name":"Advanced Graph Neural Networks","score":0.030700000002980232,"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/model-collapse","display_name":"model collapse","score":0.9279999732971191},{"id":"https://openalex.org/keywords/text-embeddings","display_name":"text embeddings","score":0.8299999833106995},{"id":"https://openalex.org/keywords/data-curation","display_name":"data curation","score":0.9549999833106995},{"id":"https://openalex.org/keywords/tabular-data","display_name":"tabular data","score":0.9160000085830688},{"id":"https://openalex.org/keywords/embedding-quality","display_name":"embedding quality","score":0.625},{"id":"https://openalex.org/keywords/out-of-distribution-generalization","display_name":"out-of-distribution generalization","score":0.8650000095367432}],"concepts":[{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.9391999840736389},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6789000034332275},{"id":"https://openalex.org/C132964779","wikidata":"https://www.wikidata.org/wiki/Q2110223","display_name":"Raw data","level":2,"score":0.6416000127792358},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.5659000277519226},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.5446000099182129},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5200999975204468},{"id":"https://openalex.org/C91632574","wikidata":"https://www.wikidata.org/wiki/Q15088675","display_name":"Data curation","level":2,"score":0.5026000142097473},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.47279998660087585},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.47209998965263367},{"id":"https://openalex.org/C24756922","wikidata":"https://www.wikidata.org/wiki/Q1757694","display_name":"Data quality","level":3,"score":0.460099995136261},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.4584999978542328},{"id":"https://openalex.org/C58489278","wikidata":"https://www.wikidata.org/wiki/Q1172284","display_name":"Data set","level":2,"score":0.4480000138282776},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43860000371932983},{"id":"https://openalex.org/C2780148112","wikidata":"https://www.wikidata.org/wiki/Q1432581","display_name":"Proxy (statistics)","level":2,"score":0.4088999927043915},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.3765000104904175},{"id":"https://openalex.org/C48677424","wikidata":"https://www.wikidata.org/wiki/Q6888088","display_name":"Mode (computer interface)","level":2,"score":0.35280001163482666},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.3483000099658966},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.34380000829696655},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.32910001277923584},{"id":"https://openalex.org/C145741570","wikidata":"https://www.wikidata.org/wiki/Q7251534","display_name":"Proteogenomics","level":5,"score":0.2994000017642975},{"id":"https://openalex.org/C2776321320","wikidata":"https://www.wikidata.org/wiki/Q857525","display_name":"Annotation","level":2,"score":0.28600001335144043},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.26510000228881836},{"id":"https://openalex.org/C2776904630","wikidata":"https://www.wikidata.org/wiki/Q356336","display_name":"Adept","level":3,"score":0.2614000141620636},{"id":"https://openalex.org/C189474733","wikidata":"https://www.wikidata.org/wiki/Q917912","display_name":"Model building","level":2,"score":0.2524999976158142},{"id":"https://openalex.org/C134121241","wikidata":"https://www.wikidata.org/wiki/Q899301","display_name":"Yield (engineering)","level":2,"score":0.2522999942302704}],"mesh":[],"locations_count":3,"locations":[{"id":"pmh:oai:arXiv.org:2506.17989","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2506.17989","pdf_url":"https://arxiv.org/pdf/2506.17989","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":"pmh:oai:HAL:hal-05124551v1","doi":null,"is_oa":true,"landing_page_url":"https://hal.science/hal-05124551","pdf_url":"https://hal.science/hal-05124551/document","source":{"id":"https://openalex.org/S4306402512","display_name":"HAL (Le Centre pour la Communication Scientifique Directe)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I1294671590","host_organization_name":"Centre National de la Recherche Scientifique","host_organization_lineage":["https://openalex.org/I1294671590"],"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":"Workshop \"DataWorld: Unifying Data Curation Frameworks Across Domains\" at International Conference on Machine Learning (ICML), Jul 2025, Vancouver (Canada), Canada","raw_type":"info:eu-repo/semantics/conferenceObject"},{"id":"doi:10.48550/arxiv.2506.17989","doi":"https://doi.org/10.48550/arxiv.2506.17989","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2506.17989","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":"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:2506.17989","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2506.17989","pdf_url":"https://arxiv.org/pdf/2506.17989","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":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Training":[0],"models":[1,35],"on":[2,41,80],"uncurated":[3],"Text":[4],"Embeddings":[5],"(TEs)":[6],"derived":[7],"from":[8],"raw":[9,43],"tabular":[10,44],"data":[11,45,87],"can":[12,122],"lead":[13],"to":[14,26],"a":[15,27,55,65,77,84,100],"severe":[16],"failure":[17,57],"mode":[18,58],"known":[19],"as":[20,83,99],"model":[21,74,120],"collapse,":[22,75],"where":[23],"predictions":[24],"converge":[25],"single":[28],"class":[29],"regardless":[30],"of":[31,67,73,119,141],"input.":[32],"By":[33],"comparing":[34],"trained":[36],"with":[37],"identical":[38],"hyper-parameter":[39],"configurations":[40],"both":[42],"and":[46,104,126,139],"their":[47,106],"TE-derived":[48],"counterparts,":[49],"we":[50,114],"find":[51],"that":[52,69,92,105,116],"collapse":[53,121],"is":[54],"consistent":[56],"in":[59,145],"the":[60,71,117,133],"latter":[61],"setting.":[62],"We":[63],"introduce":[64],"set":[66],"metrics":[68],"capture":[70],"extent":[72],"offering":[76],"new":[78],"perspective":[79],"TE":[81,93],"quality":[82,107],"proxy":[85],"for":[86,135],"curation.":[88],"Our":[89],"results":[90],"reveal":[91],"alone":[94],"does":[95],"not":[96],"effectively":[97],"function":[98],"curation":[101,138],"layer":[102],"-":[103],"significantly":[108],"influences":[109],"downstream":[110],"learning.":[111],"More":[112],"insidiously,":[113],"observe":[115],"presence":[118],"yield":[123],"artificially":[124],"inflated":[125],"spurious":[127],"Accuracy-on-the-Line":[128],"correlation.":[129],"These":[130],"findings":[131],"highlight":[132],"need":[134],"more":[136],"nuanced":[137],"evaluation":[140],"embedding-based":[142],"representations,":[143],"particularly":[144],"out-of-distribution":[146],"settings.":[147]},"counts_by_year":[],"updated_date":"2026-09-21T07:43:02.447757","created_date":"2025-10-10T00:00:00"}