{"id":"https://openalex.org/W4392490241","doi":"https://doi.org/10.48550/arxiv.2403.00952","title":"MediSwift: Efficient Sparse Pre-trained Biomedical Language Models","display_name":"MediSwift: Efficient Sparse Pre-trained Biomedical Language Models","publication_year":2024,"publication_date":"2024-03-01","ids":{"openalex":"https://openalex.org/W4392490241","doi":"https://doi.org/10.48550/arxiv.2403.00952"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2403.00952","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.00952","pdf_url":"https://arxiv.org/pdf/2403.00952","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":"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/2403.00952","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5022019488","display_name":"Vithursan Thangarasa","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Thangarasa, Vithursan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5063557363","display_name":"Mahmoud Salem","orcid":"https://orcid.org/0000-0002-9787-0525"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Salem, Mahmoud","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5025685148","display_name":"Shreyas Saxena","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saxena, Shreyas","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102722010","display_name":"K.M.K.H. Leong","orcid":"https://orcid.org/0000-0002-5968-4342"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leong, Kevin","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089825123","display_name":"Joel Hestness","orcid":"https://orcid.org/0000-0001-6920-0906"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hestness, Joel","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5001174692","display_name":"Sean Lie","orcid":"https://orcid.org/0000-0001-6074-4640"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lie, Sean","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5022019488"],"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":0.9505000114440918,"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":0.9505000114440918,"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/T11710","display_name":"Biomedical Text Mining and Ontologies","score":0.9424999952316284,"subfield":{"id":"https://openalex.org/subfields/1312","display_name":"Molecular Biology"},"field":{"id":"https://openalex.org/fields/13","display_name":"Biochemistry, Genetics and Molecular Biology"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5972725749015808},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.43518948554992676},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.43294936418533325}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5972725749015808},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.43518948554992676},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43294936418533325}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2403.00952","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.00952","pdf_url":"https://arxiv.org/pdf/2403.00952","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},{"id":"doi:10.48550/arxiv.2403.00952","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2403.00952","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":"article"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2403.00952","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2403.00952","pdf_url":"https://arxiv.org/pdf/2403.00952","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":"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/W4392490241.pdf","grobid_xml":"https://content.openalex.org/works/W4392490241.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W2358668433","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W2382290278","https://openalex.org/W2478288626","https://openalex.org/W4391913857","https://openalex.org/W2350741829","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,28,134,177],"(LLMs)":[3],"are":[4],"typically":[5],"trained":[6],"on":[7,68,98,142,150],"general":[8],"source":[9],"data":[10],"for":[11,172],"various":[12],"domains,":[13],"but":[14],"a":[15,59,86],"recent":[16],"surge":[17],"in":[18,29,89,178],"domain-specific":[19,30,35,69],"LLMs":[20,50,137],"has":[21],"shown":[22],"their":[23],"potential":[24],"to":[25,41,76,107,139],"outperform":[26,135],"general-purpose":[27],"tasks":[31,151],"(e.g.,":[32],"biomedicine).":[33],"Although":[34],"pre-training":[36,67,82,95],"enhances":[37],"efficiency":[38,120],"and":[39,129,165],"leads":[40],"smaller":[42],"models,":[43],"the":[44,81,99,109,119,122],"computational":[45],"costs":[46],"of":[47,61,121],"training":[48,90],"these":[49],"remain":[51],"high,":[52],"posing":[53],"budgeting":[54],"challenges.":[55],"We":[56],"introduce":[57],"MediSwift,":[58],"suite":[60],"biomedical":[62,70,143],"LMs":[63],"that":[64,158],"leverage":[65],"sparse":[66,94,159],"text":[71],"data.":[72],"By":[73],"inducing":[74],"up":[75,138],"75%":[77],"weight":[78,114],"sparsity":[79],"during":[80],"phase,":[83],"MediSwift":[84,123,133],"achieves":[85],"2-2.5x":[87],"reduction":[88],"FLOPs.":[91],"Notably,":[92],"all":[93],"was":[96],"performed":[97],"Cerebras":[100],"CS-2":[101],"system,":[102],"which":[103],"is":[104],"specifically":[105],"designed":[106],"realize":[108],"acceleration":[110],"benefits":[111],"from":[112],"unstructured":[113],"sparsity,":[115],"thereby":[116],"significantly":[117],"enhancing":[118],"models.":[124],"Through":[125],"subsequent":[126],"dense":[127,163],"fine-tuning":[128,164],"strategic":[130],"soft":[131,166],"prompting,":[132,167],"existing":[136],"7B":[140],"parameters":[141],"tasks,":[144],"setting":[145],"new":[146],"benchmarks":[147],"w.r.t":[148],"efficiency-accuracy":[149],"such":[152],"as":[153],"PubMedQA.":[154],"Our":[155],"results":[156],"show":[157],"pre-training,":[160],"along":[161],"with":[162],"offers":[168],"an":[169],"effective":[170],"method":[171],"creating":[173],"high-performing,":[174],"computationally":[175],"efficient":[176],"specialized":[179],"domains.":[180]},"counts_by_year":[],"updated_date":"2026-03-10T16:38:18.471706","created_date":"2025-10-10T00:00:00"}