{"id":"https://openalex.org/W4403575587","doi":"https://doi.org/10.48550/arxiv.2410.10181","title":"Scalable Multi-Domain Adaptation of Language Models using Modular Experts","display_name":"Scalable Multi-Domain Adaptation of Language Models using Modular Experts","publication_year":2024,"publication_date":"2024-10-14","ids":{"openalex":"https://openalex.org/W4403575587","doi":"https://doi.org/10.48550/arxiv.2410.10181"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2410.10181","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.10181","pdf_url":"https://arxiv.org/pdf/2410.10181","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2410.10181","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5070342607","display_name":"Peter Schafhalter","orcid":"https://orcid.org/0009-0007-9865-0456","observed_orcids":["https://orcid.org/0009-0007-9865-0456"]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Schafhalter, Peter","raw_affiliation_strings":["UC Berkeley"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UC Berkeley","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Liao, Shun","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Liao, Shun","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5081567725","display_name":"Yanqi Zhou","orcid":"https://orcid.org/0000-0003-2051-7616","observed_orcids":["https://orcid.org/0000-0003-2051-7616","https://orcid.org/0009-0003-9304-6090"]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhou, Yanqi","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045708570","display_name":"Chih\u2010Kuan Yeh","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yeh, Chih-Kuan","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043377709","display_name":"Arun Kandoor","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Kandoor, Arun","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5017565803","display_name":"James P. Laudon","orcid":"https://orcid.org/0000-0003-1061-1780","observed_orcids":["https://orcid.org/0000-0003-1061-1780"]},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Laudon, James","raw_affiliation_strings":["Google"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google","institution_ids":["https://openalex.org/I1291425158"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9970999956130981,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9970999956130981,"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":0.0027000000700354576,"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":9.999999747378752e-05,"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/multi-domain-adaptation","display_name":"multi-domain adaptation","score":0.9340000152587891},{"id":"https://openalex.org/keywords/pre-trained-language-models","display_name":"pre-trained language models","score":0.9169999957084656},{"id":"https://openalex.org/keywords/mixture-of-experts","display_name":"mixture of experts","score":0.8650000095367432},{"id":"https://openalex.org/keywords/domain-specific-fine-tuning","display_name":"domain-specific fine-tuning","score":0.7590000033378601},{"id":"https://openalex.org/keywords/knowledge-retention","display_name":"knowledge retention","score":0.824999988079071},{"id":"https://openalex.org/keywords/edge-devices","display_name":"edge devices","score":0.5910000205039978},{"id":"https://openalex.org/keywords/distributed-training","display_name":"distributed training","score":0.7070000171661377},{"id":"https://openalex.org/keywords/sharding","display_name":"sharding","score":0.5590000152587891},{"id":"https://openalex.org/keywords/low-rank-adaptation","display_name":"low-rank adaptation","score":0.6159999966621399}],"concepts":[{"id":"https://openalex.org/C101468663","wikidata":"https://www.wikidata.org/wiki/Q1620158","display_name":"Modular design","level":2,"score":0.8326762914657593},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7451510429382324},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.7435926198959351},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.7301801443099976},{"id":"https://openalex.org/C2776434776","wikidata":"https://www.wikidata.org/wiki/Q19246213","display_name":"Domain adaptation","level":3,"score":0.6959372758865356},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.6025938391685486},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.35181763768196106},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.33461666107177734},{"id":"https://openalex.org/C77088390","wikidata":"https://www.wikidata.org/wiki/Q8513","display_name":"Database","level":1,"score":0.11768916249275208},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.06531742215156555},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.052641600370407104},{"id":"https://openalex.org/C169760540","wikidata":"https://www.wikidata.org/wiki/Q207011","display_name":"Neuroscience","level":1,"score":0.0},{"id":"https://openalex.org/C16910744","wikidata":"https://www.wikidata.org/wiki/Q7705759","display_name":"Test data","level":2,"score":0.0},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2410.10181","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.10181","pdf_url":"https://arxiv.org/pdf/2410.10181","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2410.10181","doi":"https://doi.org/10.48550/arxiv.2410.10181","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2410.10181","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,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2410.10181","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.10181","pdf_url":"https://arxiv.org/pdf/2410.10181","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4403575587.pdf","grobid_xml":"https://content.openalex.org/works/W4403575587.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7155086021","https://openalex.org/W7216476598","https://openalex.org/W4410356707","https://openalex.org/W7168235484","https://openalex.org/W4385569751","https://openalex.org/W7147548055","https://openalex.org/W4287896269","https://openalex.org/W4416036110","https://openalex.org/W4416766941","https://openalex.org/W4313182274"],"abstract_inverted_index":{"Domain-specific":[0],"adaptation":[1,89],"is":[2,58],"critical":[3],"to":[4,33,86,119,142],"maximizing":[5],"the":[6],"performance":[7],"of":[8,38,95],"pre-trained":[9],"language":[10],"models":[11],"(PLMs)":[12],"on":[13],"one":[14],"or":[15],"multiple":[16],"targeted":[17],"tasks,":[18],"especially":[19],"under":[20],"resource-constrained":[21],"use":[22],"cases,":[23],"such":[24],"as":[25],"edge":[26],"devices.":[27],"However,":[28],"existing":[29],"methods":[30],"often":[31],"struggle":[32],"balance":[34],"domain-specific":[35],"performance,":[36],"retention":[37,127],"general":[39,65],"knowledge,":[40],"and":[41,45,76,106,136],"efficiency":[42],"for":[43],"training":[44,82,104,138,147],"inference.":[46],"To":[47],"address":[48],"these":[49],"challenges,":[50],"we":[51],"propose":[52],"Modular":[53],"Domain":[54],"Experts":[55],"(MoDE).":[56],"MoDE":[57,92,114],"a":[59,64,80],"mixture-of-experts":[60],"architecture":[61,131],"that":[62,113],"augments":[63],"PLMs":[66],"with":[67,102],"modular,":[68],"domain-specialized":[69],"experts.":[70],"These":[71],"experts":[72],"are":[73],"trained":[74],"independently":[75],"composed":[77],"together":[78],"via":[79],"lightweight":[81],"process.":[83],"In":[84],"contrast":[85],"standard":[87],"low-rank":[88],"methods,":[90],"each":[91],"expert":[93],"consists":[94],"several":[96],"transformer":[97],"layers":[98],"which":[99],"scale":[100],"better":[101,126],"more":[103],"examples":[105],"larger":[107],"parameter":[108,121],"counts.":[109],"Our":[110],"evaluation":[111],"demonstrates":[112],"achieves":[115],"comparable":[116],"target":[117],"performances":[118],"full":[120],"fine-tuning":[122],"while":[123],"achieving":[124],"1.65%":[125],"performance.":[128],"Moreover,":[129],"MoDE's":[130],"enables":[132],"flexible":[133],"sharding":[134],"configurations":[135],"improves":[137],"speeds":[139],"by":[140],"up":[141],"38%":[143],"over":[144],"state-of-the-art":[145],"distributed":[146],"configurations.":[148]},"counts_by_year":[],"updated_date":"2026-10-01T08:54:57.152042","created_date":"2024-10-20T00:00:00"}