{"id":"https://openalex.org/W4399795577","doi":"https://doi.org/10.48550/arxiv.2406.10260","title":"Flextron: Many-in-One Flexible Large Language Model","display_name":"Flextron: Many-in-One Flexible Large Language Model","publication_year":2024,"publication_date":"2024-06-11","ids":{"openalex":"https://openalex.org/W4399795577","doi":"https://doi.org/10.48550/arxiv.2406.10260"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2406.10260","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2406.10260","pdf_url":"https://arxiv.org/pdf/2406.10260","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2406.10260","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5111233264","display_name":"Ruisi Cai","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Cai, Ruisi","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5099241641","display_name":"Saurav Muralidharan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Muralidharan, Saurav","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030133314","display_name":"Greg Heinrich","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Heinrich, Greg","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108868459","display_name":"Hongxu Yin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yin, Hongxu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048522863","display_name":"Zhangyang Wang","orcid":"https://orcid.org/0000-0002-2050-5693"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zhangyang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056503617","display_name":"Jan Kautz","orcid":"https://orcid.org/0000-0002-8830-429X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kautz, Jan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5066945976","display_name":"Pavlo Molchanov","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Molchanov, Pavlo","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5111233264"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10028","display_name":"Topic Modeling","score":0.8877000212669373,"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/T10028","display_name":"Topic Modeling","score":0.8877000212669373,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.8327000141143799,"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/computer-science","display_name":"Computer science","score":0.523561418056488},{"id":"https://openalex.org/keywords/linguistics","display_name":"Linguistics","score":0.3515903949737549},{"id":"https://openalex.org/keywords/philosophy","display_name":"Philosophy","score":0.11303001642227173}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.523561418056488},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.3515903949737549},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.11303001642227173}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2406.10260","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2406.10260","pdf_url":"https://arxiv.org/pdf/2406.10260","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},{"id":"doi:10.48550/arxiv.2406.10260","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2406.10260","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2406.10260","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2406.10260","pdf_url":"https://arxiv.org/pdf/2406.10260","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4399795577.pdf","grobid_xml":"https://content.openalex.org/works/W4399795577.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052","https://openalex.org/W2382290278","https://openalex.org/W4395014643"],"abstract_inverted_index":{"Training":[0],"modern":[1],"LLMs":[2],"is":[3,24,70],"extremely":[4],"resource":[5],"intensive,":[6],"and":[7,18,35,59,73,84,92,113,118,127],"customizing":[8],"them":[9],"for":[10,81,96],"various":[11],"deployment":[12],"scenarios":[13],"characterized":[14],"by":[15],"limited":[16],"compute":[17],"memory":[19],"resources":[20],"through":[21,78],"repeated":[22],"training":[23,90],"impractical.":[25],"In":[26],"this":[27],"paper,":[28],"we":[29],"introduce":[30],"Flextron,":[31],"a":[32,48,88,104,134,140],"network":[33],"architecture":[34,46],"post-training":[36],"model":[37,42],"optimization":[38],"framework":[39],"supporting":[40],"flexible":[41],"deployment.":[43],"The":[44],"Flextron":[45,105,109],"utilizes":[47],"nested":[49],"elastic":[50,130],"structure":[51],"to":[52,55,145],"rapidly":[53],"adapt":[54],"specific":[56],"user-defined":[57],"latency":[58],"accuracy":[60],"targets":[61],"during":[62],"inference":[63],"with":[64,133],"no":[65],"additional":[66],"fine-tuning":[67],"required.":[68],"It":[69],"also":[71],"input-adaptive,":[72],"can":[74],"automatically":[75],"route":[76],"tokens":[77,143],"its":[79],"sub-networks":[80],"improved":[82],"performance":[83,121],"efficiency.":[85],"We":[86,107],"present":[87],"sample-efficient":[89],"method":[91],"associated":[93],"routing":[94],"algorithms":[95],"systematically":[97],"transforming":[98],"an":[99],"existing":[100],"trained":[101,125],"LLM":[102],"into":[103],"model.":[106],"evaluate":[108],"on":[110],"the":[111],"GPT-3":[112],"LLama-2":[114],"family":[115],"of":[116],"LLMs,":[117],"demonstrate":[119],"superior":[120],"over":[122],"multiple":[123],"end-to-end":[124],"variants":[126],"other":[128],"state-of-the-art":[129],"networks,":[131],"all":[132],"single":[135],"pretraining":[136],"run":[137],"that":[138],"consumes":[139],"mere":[141],"7.63%":[142],"compared":[144],"original":[146],"pretraining.":[147]},"counts_by_year":[],"updated_date":"2026-03-07T16:01:11.037858","created_date":"2025-10-10T00:00:00"}