{"id":"https://openalex.org/W7162456501","doi":"https://doi.org/10.48550/arxiv.2605.25451","title":"BigMac: Breaking the Pareto Frontier of Compute and Memory in Multimodal LLM Training","display_name":"BigMac: Breaking the Pareto Frontier of Compute and Memory in Multimodal LLM Training","publication_year":2026,"publication_date":"2026-05-25","ids":{"openalex":"https://openalex.org/W7162456501","doi":"https://doi.org/10.48550/arxiv.2605.25451"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.25451","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.25451","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,"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.2605.25451","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137066763","display_name":"Zili Zhang","orcid":"https://orcid.org/0000-0001-9836-2138"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zili","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137056353","display_name":"Chengxu Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Chengxu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137059199","display_name":"Shenglong Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Shenglong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137073924","display_name":"Chenyu Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Chenyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137034108","display_name":"Yufan Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yufan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137001043","display_name":"Tuo Dai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dai, Tuo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101416241","display_name":"Zhouyang Li","orcid":"https://orcid.org/0009-0006-8570-9589"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Zhouyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137009548","display_name":"Yuhong Ge","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ge, Yuhong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136999421","display_name":"Chao Jin","orcid":"https://orcid.org/0000-0001-6590-333X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jin, Chao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137045952","display_name":"Xin Jin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jin, Xin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137029521","display_name":"Yuliang Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Yuliang","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/T10181","display_name":"Natural Language Processing Techniques","score":0.31450000405311584,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.31450000405311584,"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/T10028","display_name":"Topic Modeling","score":0.28929999470710754,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.09200000017881393,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.6207000017166138},{"id":"https://openalex.org/keywords/pareto-principle","display_name":"Pareto principle","score":0.578000009059906},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.566100001335144},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.5424000024795532},{"id":"https://openalex.org/keywords/generator","display_name":"Generator (circuit theory)","score":0.4724000096321106},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.4596000015735626},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.4262000024318695},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.37439998984336853}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.741100013256073},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.6207000017166138},{"id":"https://openalex.org/C137635306","wikidata":"https://www.wikidata.org/wiki/Q182667","display_name":"Pareto principle","level":2,"score":0.578000009059906},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.566100001335144},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5424000024795532},{"id":"https://openalex.org/C2780992000","wikidata":"https://www.wikidata.org/wiki/Q17016113","display_name":"Generator (circuit theory)","level":3,"score":0.4724000096321106},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.4596000015735626},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.4262000024318695},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.37439998984336853},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3720000088214874},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.36980000138282776},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.36730000376701355},{"id":"https://openalex.org/C68781425","wikidata":"https://www.wikidata.org/wiki/Q2052203","display_name":"Multi-objective optimization","level":2,"score":0.3328000009059906},{"id":"https://openalex.org/C176649486","wikidata":"https://www.wikidata.org/wiki/Q2308807","display_name":"Memory management","level":3,"score":0.3050999939441681},{"id":"https://openalex.org/C2781215313","wikidata":"https://www.wikidata.org/wiki/Q3493345","display_name":"SPARK (programming language)","level":2,"score":0.29820001125335693},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.2799000144004822},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.27559998631477356},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.26989999413490295},{"id":"https://openalex.org/C2986314615","wikidata":"https://www.wikidata.org/wiki/Q36829","display_name":"Pareto optimal","level":3,"score":0.26930001378059387},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.2599000036716461},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.2590999901294708},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.2556999921798706},{"id":"https://openalex.org/C51485801","wikidata":"https://www.wikidata.org/wiki/Q16966861","display_name":"Efficient frontier","level":3,"score":0.250900000333786}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.25451","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.25451","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,"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.2605.25451","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.25451","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,"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":{"Training":[0],"multimodal":[1,53],"large":[2],"language":[3],"models":[4],"(MLLMs)":[5],"is":[6,60],"challenged":[7],"by":[8,27],"both":[9,141],"model":[10],"and":[11,33,66,92,134,143,154],"data":[12],"heterogeneity.":[13],"Existing":[14],"systems":[15,169],"redesign":[16],"the":[17,40,43,64,70,85,90,98,103,107,112,117,128],"training":[18,50,155,165],"pipeline":[19,51,78],"to":[20,61,94],"address":[21],"these":[22],"challenges,":[23],"but":[24],"remain":[25],"bound":[26],"a":[28,48,75,124,163],"Pareto":[29,129],"frontier":[30,130],"between":[31,131],"compute":[32],"memory":[34,87,100,135,144,173],"efficiency,":[35],"improving":[36],"one":[37],"only":[38],"at":[39],"expense":[41],"of":[42,58,89,102,140],"other.":[44],"We":[45,148],"present":[46],"BigMac,":[47],"new":[49],"for":[52],"LLMs.":[54],"The":[55],"core":[56],"idea":[57],"BigMac":[59,83,126,150,161],"elegantly":[62],"nest":[63],"encoder":[65,91],"generator":[67,93],"computation":[68,142],"into":[69],"original":[71],"LLM":[72,104],"pipeline,":[73],"forming":[74],"dependency-safe":[76],"nested":[77],"structure.":[79],"With":[80],"this":[81],"design,":[82],"reduces":[84],"activation":[86,99],"complexity":[88,101],"O(1)":[95],"while":[96,170],"keeping":[97],"unchanged.":[105],"At":[106],"same":[108,113],"time,":[109],"it":[110],"achieves":[111,162],"computational":[114,132],"efficiency":[115,133],"as":[116,175],"idealized":[118],"setting":[119],"with":[120],"unlimited":[121],"memory.":[122],"As":[123],"result,":[125],"breaks":[127],"usage,":[136],"enabling":[137],"simultaneous":[138],"optimization":[139],"in":[145],"MLLM":[146],"training.":[147],"evaluate":[149],"on":[151],"multiple":[152],"MLLMs":[153],"workloads.":[156],"Experimental":[157],"results":[158],"show":[159],"that":[160],"1.08$\\times$-1.9$\\times$":[164],"speedup":[166],"over":[167],"baseline":[168],"maintaining":[171],"stable":[172],"usage":[174],"batch":[176],"size":[177],"increases.":[178]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-27T00:00:00"}