{"id":"https://openalex.org/W7146975323","doi":"https://doi.org/10.48550/arxiv.2603.26786","title":"A Step Toward Federated Pretraining of Multimodal Large Language Models","display_name":"A Step Toward Federated Pretraining of Multimodal Large Language Models","publication_year":2026,"publication_date":"2026-03-25","ids":{"openalex":"https://openalex.org/W7146975323","doi":"https://doi.org/10.48550/arxiv.2603.26786"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.26786","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.26786","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.2603.26786","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5050573340","display_name":"Baochen Xiong","orcid":"https://orcid.org/0009-0005-9799-2898"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiong, Baochen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132562590","display_name":"Yifan Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Yifan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132564887","display_name":"Xiaoshan Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Xiaoshan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132606157","display_name":"Yaguang Song","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Song, Yaguang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132604924","display_name":"Yaowei Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yaowei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5132544145","display_name":"Changsheng Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Changsheng","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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.26179999113082886,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.26179999113082886,"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":0.15469999611377716,"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":0.08540000021457672,"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/construct","display_name":"Construct (python library)","score":0.5616999864578247},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.5188000202178955},{"id":"https://openalex.org/keywords/federated-learning","display_name":"Federated learning","score":0.43050000071525574},{"id":"https://openalex.org/keywords/basis","display_name":"Basis (linear algebra)","score":0.41609999537467957},{"id":"https://openalex.org/keywords/multimodality","display_name":"Multimodality","score":0.4077000021934509},{"id":"https://openalex.org/keywords/space","display_name":"Space (punctuation)","score":0.33570000529289246}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7724999785423279},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.5616999864578247},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.5188000202178955},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4672999978065491},{"id":"https://openalex.org/C2992525071","wikidata":"https://www.wikidata.org/wiki/Q50818671","display_name":"Federated learning","level":2,"score":0.43050000071525574},{"id":"https://openalex.org/C12426560","wikidata":"https://www.wikidata.org/wiki/Q189569","display_name":"Basis (linear algebra)","level":2,"score":0.41609999537467957},{"id":"https://openalex.org/C2780910867","wikidata":"https://www.wikidata.org/wiki/Q1952416","display_name":"Multimodality","level":2,"score":0.4077000021934509},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.3628999888896942},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3521000146865845},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.33570000529289246},{"id":"https://openalex.org/C135641252","wikidata":"https://www.wikidata.org/wiki/Q738567","display_name":"Multimodal interaction","level":2,"score":0.30489999055862427},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3009999990463257},{"id":"https://openalex.org/C2776321320","wikidata":"https://www.wikidata.org/wiki/Q857525","display_name":"Annotation","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.27880001068115234},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2728999853134155},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.26980000734329224},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.25099998712539673}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.26786","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.26786","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.2603.26786","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.26786","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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"rapid":[1],"evolution":[2],"of":[3,14,21],"Multimodal":[4],"Large":[5],"Language":[6],"Models":[7],"(MLLMs)":[8],"is":[9],"bottlenecked":[10],"by":[11],"the":[12,50,62,74,82,161],"saturation":[13],"high-quality":[15],"public":[16,184],"data,":[17],"while":[18,79,172],"vast":[19],"amounts":[20],"diverse":[22],"multimodal":[23],"data":[24],"remain":[25],"inaccessible":[26],"in":[27,90,96,104],"privacy-sensitive":[28],"silos.":[29],"Federated":[30,63],"Learning":[31],"(FL)":[32],"offers":[33],"a":[34,68,115,129,137],"promising":[35],"solution":[36],"to":[37,132,148,160],"unlock":[38],"these":[39,110],"distributed":[40],"resources,":[41],"but":[42],"existing":[43,194],"research":[44],"focuses":[45],"predominantly":[46],"on":[47,183],"fine-tuning,":[48],"leaving":[49],"foundational":[51],"pre-training":[52,70,180],"phase":[53],"largely":[54],"unexplored.":[55],"In":[56],"this":[57,91],"paper,":[58],"we":[59,112],"formally":[60],"introduce":[61],"MLLM":[64,120],"Alignment":[65],"(Fed-MA)":[66],"task,":[67],"lightweight":[69],"paradigm":[71],"that":[72,190],"freezes":[73],"vision":[75],"encoder":[76],"and":[77,100,141,186],"LLM":[78],"collaboratively":[80],"training":[81],"cross-modal":[83],"projector.":[84],"We":[85,176],"identify":[86],"two":[87],"critical":[88],"challenges":[89],"setting:":[92],"(i)":[93],"parameter":[94,150],"interference":[95],"aggregating":[97],"local":[98],"projectors;":[99],"(ii)":[101],"gradient":[102],"oscillations":[103],"one-pass":[105],"collaborative":[106],"SGD.":[107],"To":[108],"address":[109],"challenges,":[111],"propose":[113],"Fed-CMP,":[114],"pioneering":[116],"framework":[117],"for":[118],"federated":[119,179],"pre-training.":[121],"Fed-CMP":[122,153,191],"employs":[123],"Canonical":[124],"Reliability-Aware":[125],"Aggregation,":[126],"which":[127,157],"constructs":[128],"canonical":[130],"space":[131],"decompose":[133],"client":[134],"projectors":[135],"into":[136],"shared":[138,162],"alignment":[139,163],"basis":[140,164],"client-specific":[142],"coefficients,":[143],"then":[144],"performs":[145],"reliability-weighted":[146],"fusion":[147],"suppress":[149],"interference.":[151],"Furthermore,":[152],"introduces":[154],"Orthogonality-Preserved":[155],"Momentum,":[156],"applies":[158],"momentum":[159],"via":[165],"orthogonal":[166],"projection,":[167],"accumulating":[168],"historical":[169],"optimization":[170],"directions":[171],"preserving":[173],"geometric":[174],"structure.":[175],"construct":[177],"four":[178],"scenarios":[181],"based":[182],"datasets,":[185],"extensive":[187],"experiments":[188],"validate":[189],"significantly":[192],"outperforms":[193],"baselines.":[195]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-02T00:00:00"}