{"id":"https://openalex.org/W7162509926","doi":"https://doi.org/10.48550/arxiv.2605.26761","title":"Once-For-All: A Train-Once and Select-Anytime Framework for Multimodal Instruction Tuning","display_name":"Once-For-All: A Train-Once and Select-Anytime Framework for Multimodal Instruction Tuning","publication_year":2026,"publication_date":"2026-05-26","ids":{"openalex":"https://openalex.org/W7162509926","doi":"https://doi.org/10.48550/arxiv.2605.26761"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.26761","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.26761","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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.2605.26761","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5036049280","display_name":"Mi Dong","orcid":"https://orcid.org/0000-0001-5373-5631"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong, Mingkang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136208979","display_name":"Hongyi Cai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cai, Hongyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137092280","display_name":"Xiwen Lei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lei, Xiwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137165542","display_name":"Jie Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Jie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137190025","display_name":"Tao Zhang","orcid":"https://orcid.org/0009-0001-4253-5578"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Tao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5005649659","display_name":"Muxin Pu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pu, Muxin","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":0.9528999924659729,"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":0.9528999924659729,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.00800000037997961,"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":0.005799999926239252,"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/selection","display_name":"Selection (genetic algorithm)","score":0.6723999977111816},{"id":"https://openalex.org/keywords/train","display_name":"Train","score":0.5814999938011169},{"id":"https://openalex.org/keywords/decoupling","display_name":"Decoupling (probability)","score":0.5062000155448914},{"id":"https://openalex.org/keywords/model-selection","display_name":"Model selection","score":0.4740999937057495},{"id":"https://openalex.org/keywords/de-facto","display_name":"De facto","score":0.39890000224113464},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.38190001249313354},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.3732999861240387}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8208000063896179},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.6723999977111816},{"id":"https://openalex.org/C190839683","wikidata":"https://www.wikidata.org/wiki/Q2448197","display_name":"Train","level":2,"score":0.5814999938011169},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5153999924659729},{"id":"https://openalex.org/C205606062","wikidata":"https://www.wikidata.org/wiki/Q5249645","display_name":"Decoupling (probability)","level":2,"score":0.5062000155448914},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.4740999937057495},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4564000070095062},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.40790000557899475},{"id":"https://openalex.org/C2992317946","wikidata":"https://www.wikidata.org/wiki/Q712144","display_name":"De facto","level":2,"score":0.39890000224113464},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.38190001249313354},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3732999861240387},{"id":"https://openalex.org/C2781235140","wikidata":"https://www.wikidata.org/wiki/Q275131","display_name":"Scratch","level":2,"score":0.3626999855041504},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.29820001125335693},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2849999964237213},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.28029999136924744},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.26429998874664307},{"id":"https://openalex.org/C132964779","wikidata":"https://www.wikidata.org/wiki/Q2110223","display_name":"Raw data","level":2,"score":0.2619999945163727},{"id":"https://openalex.org/C164866538","wikidata":"https://www.wikidata.org/wiki/Q367351","display_name":"Cluster (spacecraft)","level":2,"score":0.25040000677108765}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.26761","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.26761","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":"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":"doi:10.48550/arxiv.2605.26761","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.26761","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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":{"Multimodal":[0],"instruction":[1,15,237],"tuning":[2],"is":[3,114,137],"the":[4,40,47,96,120,125,153,160,174,177,188,217],"de":[5],"facto":[6],"recipe":[7],"for":[8,24,104,234],"adapting":[9],"vision":[10],"language":[11],"models":[12],"(VLMs),":[13],"yet":[14],"data":[16,21,64,167,182],"are":[17,117],"highly":[18],"redundant,":[19],"making":[20],"selection":[22,30,65,215],"critical":[23],"training":[25,183],"efficiency.":[26],"Existing":[27],"methods":[28],"derive":[29],"signals":[31],"from":[32,52,95,216],"a":[33,63,69,88,101,106,224],"specific":[34],"model":[35,42,80,133,212],"or":[36,43,79],"dataset,":[37],"so":[38],"whenever":[39],"target":[41,218],"candidate":[44],"pool":[45],"changes,":[46],"criteria":[48],"must":[49],"be":[50],"recomputed":[51],"scratch":[53],"at":[54,205],"substantial":[55],"cost.":[56],"To":[57],"address":[58],"this,":[59],"we":[60],"propose":[61],"OFA,":[62],"framework":[66],"that":[67,187,223],"trains":[68,100],"reusable":[70,232],"selector":[71,103,113,127,136,179,197,227],"once":[72,139],"and":[73,99,132,142,208,231],"applies":[74],"it":[75],"to":[76,145,152,192],"any":[77,150],"dataset":[78],"without":[81,149,210],"recomputation.":[82],"OFA":[83,162],"clusters":[84],"multimodal":[85,236],"instructions":[86],"in":[87],"frozen":[89,126],"CLIP":[90],"space,":[91],"derives":[92],"pseudo":[93],"labels":[94],"cluster":[97],"structure,":[98],"lightweight":[102],"only":[105,157],"few":[107],"epochs;":[108],"samples":[109],"on":[110,140,173],"which":[111],"this":[112],"least":[115],"confident":[116],"selected":[118,201],"as":[119],"most":[121],"informative.":[122],"Once":[123],"trained,":[124],"transfers":[128],"directly":[129],"across":[130,169],"datasets":[131,193],"scales.":[134],"The":[135,199],"trained":[138],"LLaVA-665K":[141,146],"applied":[143],"both":[144,206],"itself":[147],"and,":[148],"retraining,":[151],"unseen":[154],"Vision-Flan-186K.":[155],"Selecting":[156],"15%":[158],"of":[159,165],"data,":[161],"achieves":[163],"98.3%":[164],"full":[166,181],"performance":[168],"10":[170],"downstream":[171],"benchmarks;":[172],"smaller":[175],"Vision-Flan-186K,":[176],"transferred":[178],"surpasses":[180],"by":[184],"10.6%,":[185],"confirming":[186],"learned":[189],"signal":[190],"generalizes":[191],"never":[194],"seen":[195],"during":[196],"training.":[198],"same":[200],"subsets":[202],"benefit":[203],"VLMs":[204],"Qwen2.5-VL-3B":[207],"LLaVA-v1.5-7B":[209],"per":[211],"recomputation,":[213],"decoupling":[214],"model.":[219],"These":[220],"results":[221],"demonstrate":[222],"single,":[225],"transferable":[226],"provides":[228],"an":[229],"effective":[230],"solution":[233],"efficient":[235],"tuning.":[238]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-28T00:00:00"}