{"id":"https://openalex.org/W7154726984","doi":"https://doi.org/10.48550/arxiv.2604.14198","title":"MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining","display_name":"MixAtlas: Uncertainty-aware Data Mixture Optimization for Multimodal LLM Midtraining","publication_year":2026,"publication_date":"2026-04-03","ids":{"openalex":"https://openalex.org/W7154726984","doi":"https://doi.org/10.48550/arxiv.2604.14198"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.14198","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.14198","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"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.2604.14198","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133878446","display_name":"Bingbing Wen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wen, Bingbing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024976239","display_name":"Sirajul Salekin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Salekin, Sirajul","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133828989","display_name":"Feiyang Kang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kang, Feiyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133881911","display_name":"Bill Howe","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Howe, Bill","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133899215","display_name":"Lucy Lu Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Lucy Lu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5010136172","display_name":"Javier R. Movellan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Movellan, Javier","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5128771818","display_name":"Manjot Bilkhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bilkhu, Manjot","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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9225000143051147,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9225000143051147,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T10028","display_name":"Topic Modeling","score":0.030899999663233757,"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":0.01510000042617321,"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/task","display_name":"Task (project management)","score":0.6155999898910522},{"id":"https://openalex.org/keywords/proxy","display_name":"Proxy (statistics)","score":0.506600022315979},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.4830000102519989},{"id":"https://openalex.org/keywords/transfer-of-learning","display_name":"Transfer of learning","score":0.4156999886035919},{"id":"https://openalex.org/keywords/mixture-model","display_name":"Mixture model","score":0.3928000032901764},{"id":"https://openalex.org/keywords/sample","display_name":"Sample (material)","score":0.3865000009536743},{"id":"https://openalex.org/keywords/domain","display_name":"Domain (mathematical analysis)","score":0.35499998927116394},{"id":"https://openalex.org/keywords/sample-complexity","display_name":"Sample complexity","score":0.35030001401901245}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7440000176429749},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6155999898910522},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5267000198364258},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5083000063896179},{"id":"https://openalex.org/C2780148112","wikidata":"https://www.wikidata.org/wiki/Q1432581","display_name":"Proxy (statistics)","level":2,"score":0.506600022315979},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.4830000102519989},{"id":"https://openalex.org/C150899416","wikidata":"https://www.wikidata.org/wiki/Q1820378","display_name":"Transfer of learning","level":2,"score":0.4156999886035919},{"id":"https://openalex.org/C61224824","wikidata":"https://www.wikidata.org/wiki/Q2260434","display_name":"Mixture model","level":2,"score":0.3928000032901764},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.3865000009536743},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3797000050544739},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.35499998927116394},{"id":"https://openalex.org/C2778445095","wikidata":"https://www.wikidata.org/wiki/Q18354077","display_name":"Sample complexity","level":2,"score":0.35030001401901245},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.3479999899864197},{"id":"https://openalex.org/C2776145971","wikidata":"https://www.wikidata.org/wiki/Q30673951","display_name":"Labeled data","level":2,"score":0.3377000093460083},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.32190001010894775},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3199000060558319},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.31949999928474426},{"id":"https://openalex.org/C3020493868","wikidata":"https://www.wikidata.org/wiki/Q55631277","display_name":"Real world data","level":2,"score":0.3095000088214874},{"id":"https://openalex.org/C133462117","wikidata":"https://www.wikidata.org/wiki/Q4929239","display_name":"Data collection","level":2,"score":0.29190000891685486},{"id":"https://openalex.org/C21080849","wikidata":"https://www.wikidata.org/wiki/Q13611879","display_name":"Data point","level":2,"score":0.28299999237060547},{"id":"https://openalex.org/C35639132","wikidata":"https://www.wikidata.org/wiki/Q7452468","display_name":"Sequence labeling","level":3,"score":0.27410000562667847},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.2556999921798706},{"id":"https://openalex.org/C129848803","wikidata":"https://www.wikidata.org/wiki/Q2564360","display_name":"Sample size determination","level":2,"score":0.25119999051094055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.14198","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.14198","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.14198","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.14198","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":null,"license_id":null,"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":{"Domain":[0],"reweighting":[1],"can":[2,45],"improve":[3,132],"sample":[4],"efficiency":[5],"and":[6,49,71,82,94,125],"downstream":[7],"generalization,":[8],"but":[9,111],"data-mixture":[10],"optimization":[11],"for":[12],"multimodal":[13,19,126],"midtraining":[14],"remains":[15],"largely":[16],"unexplored.":[17],"Current":[18],"training":[20,57,150,167],"recipes":[21,43],"tune":[22],"mixtures":[23,131],"along":[24,59],"a":[25,37,91],"single":[26],"dimension,":[27],"typically":[28],"data":[29,42],"format":[30],"or":[31],"task":[32,72],"type.":[33],"We":[34,115],"introduce":[35],"MixAtlas,":[36],"method":[38],"that":[39,44],"produces":[40],"benchmark-targeted":[41],"be":[46],"inspected,":[47],"adapted,":[48],"transferred":[50],"to":[51,154,165],"new":[52],"corpora.":[53],"MixAtlas":[54,97],"decomposes":[55],"the":[56,99,104,138],"corpus":[58],"two":[60],"axes:":[61],"image":[62],"concepts":[63],"(10":[64],"visual-domain":[65],"clusters":[66],"discovered":[67,160],"via":[68],"CLIP":[69],"embeddings)":[70],"supervision":[73],"(5":[74],"objective":[75],"types":[76],"including":[77],"captioning,":[78],"OCR,":[79],"grounding,":[80],"detection,":[81],"VQA).":[83],"Using":[84],"small":[85],"proxy":[86,106],"models":[87],"(Qwen2-0.5B)":[88],"paired":[89],"with":[90,103],"Gaussian-process":[92],"surrogate":[93],"GP-UCB":[95],"acquisition,":[96],"searches":[98],"resulting":[100],"mixture":[101],"space":[102],"same":[105],"budget":[107],"as":[108],"regression-based":[109],"baselines":[110],"finds":[112],"better-performing":[113],"mixtures.":[114],"evaluate":[116],"on":[117,141,161],"10":[118],"benchmarks":[119],"spanning":[120],"visual":[121],"understanding,":[122],"document":[123],"reasoning,":[124],"reasoning.":[127],"On":[128],"Qwen2-7B,":[129],"optimized":[130],"average":[133],"performance":[134],"by":[135],"8.5%-17.6%":[136],"over":[137],"strongest":[139],"baseline;":[140],"Qwen2.5-7B,":[142],"gains":[143],"are":[144],"1.0%-3.3%.":[145],"Both":[146],"settings":[147],"reach":[148],"baseline-equivalent":[149],"loss":[151],"in":[152],"up":[153],"2":[155],"times":[156],"fewer":[157],"steps.":[158],"Recipes":[159],"0.5B":[162],"proxies":[163],"transfer":[164],"7B-scale":[166],"across":[168],"Qwen":[169],"model":[170],"families.":[171]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-18T00:00:00"}