{"id":"https://openalex.org/W7152443959","doi":"https://doi.org/10.48550/arxiv.2604.06836","title":"STQuant: Spatio-Temporal Adaptive Framework for Optimizer Quantization in Large Multimodal Model Training","display_name":"STQuant: Spatio-Temporal Adaptive Framework for Optimizer Quantization in Large Multimodal Model Training","publication_year":2026,"publication_date":"2026-04-08","ids":{"openalex":"https://openalex.org/W7152443959","doi":"https://doi.org/10.48550/arxiv.2604.06836"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.06836","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06836","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.06836","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133242656","display_name":"Minglu Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Liu, Minglu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063322816","display_name":"Cunchen Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Cunchen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112656299","display_name":"Lida Xu","orcid":"https://orcid.org/0000-0003-2467-6120"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xu, Liangliang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102666462","display_name":"Fengming Tang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tang, Fengming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133298239","display_name":"Ruijia Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Ruijia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100610546","display_name":"Fu Yu","orcid":"https://orcid.org/0000-0002-1713-2160"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Fu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5133242656"],"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/T12535","display_name":"Machine Learning and Data Classification","score":0.2863999903202057,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.2863999903202057,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.13490000367164612,"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":0.13079999387264252,"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/quantization","display_name":"Quantization (signal processing)","score":0.7978000044822693},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.4878999888896942},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.4291999936103821},{"id":"https://openalex.org/keywords/memory-footprint","display_name":"Memory footprint","score":0.39070001244544983},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.3612000048160553},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.35429999232292175},{"id":"https://openalex.org/keywords/state","display_name":"State (computer science)","score":0.3517000079154968}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7978000044822693},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.722100019454956},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.4878999888896942},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.4291999936103821},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4219000041484833},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.4034000039100647},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.39070001244544983},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.3612000048160553},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3598000109195709},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.35429999232292175},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.3517000079154968},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.33500000834465027},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.328900009393692},{"id":"https://openalex.org/C151376022","wikidata":"https://www.wikidata.org/wiki/Q168698","display_name":"Exponential function","level":2,"score":0.321399986743927},{"id":"https://openalex.org/C75235859","wikidata":"https://www.wikidata.org/wiki/Q582659","display_name":"Exponential growth","level":2,"score":0.29010000824928284},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.28790000081062317},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.28130000829696655},{"id":"https://openalex.org/C163175372","wikidata":"https://www.wikidata.org/wiki/Q3339222","display_name":"Linear model","level":2,"score":0.25540000200271606},{"id":"https://openalex.org/C30390489","wikidata":"https://www.wikidata.org/wiki/Q4680748","display_name":"Adaptive memory","level":3,"score":0.25099998712539673}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.06836","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06836","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.06836","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.06836","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":"article"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","score":0.5098769068717957,"id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Quantization":[0],"is":[1,84],"an":[2,169],"effective":[3],"way":[4],"to":[5,152],"reduce":[6],"the":[7,23,57,132,147],"memory":[8,58,165],"cost":[9,149],"of":[10,60,172],"large-scale":[11],"model":[12,76],"training.":[13],"However,":[14],"most":[15,133],"existing":[16,180],"methods":[17],"adopt":[18],"fixed-precision":[19],"policies,":[20],"which":[21],"ignore":[22],"fact":[24],"that":[25,55,129,145,161],"optimizer-state":[26,164],"distributions":[27],"vary":[28],"significantly":[29],"across":[30,67],"layers":[31,107],"and":[32,71,95,106,158,189],"training":[33,53,72,83],"steps.":[34],"Such":[35],"uniform":[36],"designs":[37],"often":[38],"introduce":[39],"noticeable":[40],"accuracy":[41],"degradation.":[42],"To":[43],"move":[44],"beyond":[45],"fixed":[46],"quantization,":[47],"we":[48],"propose":[49],"STQuant,":[50],"a":[51,109,123,140],"distributed":[52],"framework":[54],"reduces":[56,146,163],"footprint":[59],"optimizer":[61,90],"states":[62,91,105],"via":[63],"dynamic":[64,80,141],"precision":[65,137],"allocation":[66],"layers,":[68],"state":[69],"variables,":[70],"steps,":[73],"while":[74],"maintaining":[75],"quality.":[77,100],"Naively":[78],"applying":[79],"quantization":[81,96],"during":[82],"challenging":[85],"for":[86,136],"two":[87,119],"reasons.":[88],"First,":[89],"are":[92],"numerically":[93],"sensitive,":[94],"noise":[97],"can":[98],"destabilize":[99],"Second,":[101],"jointly":[102],"considering":[103],"multiple":[104],"induces":[108],"large":[110],"combinatorial":[111],"search":[112,148],"space.":[113,193],"STQuant":[114,162,183],"addresses":[115],"these":[116],"challenges":[117],"with":[118,179],"key":[120],"techniques:":[121],"1)":[122],"provably":[124],"near-optimal":[125],"factor":[126],"selection":[127],"strategy":[128],"accurately":[130],"identifies":[131],"influential":[134],"factors":[135],"adaptation.":[138],"2)":[139],"transition":[142],"decision":[143],"algorithm":[144],"from":[150],"exponential":[151],"linear":[153],"complexity.":[154],"Experiments":[155],"on":[156],"GPT-2":[157],"ViT":[159],"show":[160],"by":[166],"84.4%,":[167],"achieving":[168],"average":[170],"bit-width":[171],"as":[173,175],"low":[174],"5.1":[176],"bits,":[177],"compared":[178],"solutions.":[181],"Moreover,":[182],"incurs":[184],"only":[185],"O(N/K)":[186],"computational":[187],"overhead":[188],"requires":[190],"O(1)":[191],"extra":[192]},"counts_by_year":[],"updated_date":"2026-05-05T08:41:31.759640","created_date":"2026-04-10T00:00:00"}