{"id":"https://openalex.org/W7166803187","doi":"https://doi.org/10.18653/v1/2026.findings-acl.174","title":"ModularMoE: Fast LLM Customization with Parameter-Sharing Mixture-of-Experts for Low-Resource Settings","display_name":"ModularMoE: Fast LLM Customization with Parameter-Sharing Mixture-of-Experts for Low-Resource Settings","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166803187","doi":"https://doi.org/10.18653/v1/2026.findings-acl.174"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.174","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.174","pdf_url":"https://aclanthology.org/2026.findings-acl.174.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.174.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139838396","display_name":"Jiaxing Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiaxing Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139741526","display_name":"Qi Qi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qi Qi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139744993","display_name":"Haifeng Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haifeng Sun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139815250","display_name":"Dunjun Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dunjun Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076969590","display_name":"Zirui Zhuang","orcid":"https://orcid.org/0000-0003-3345-1732"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zirui Zhuang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113949081","display_name":"Bo He","orcid":"https://orcid.org/0000-0001-9092-9152"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bo He","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139769408","display_name":"Yang Xiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiang Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139724350","display_name":"Cong Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cong Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5055685073","display_name":"Jianxin Liao","orcid":"https://orcid.org/0000-0003-1486-0573"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jianxin Liao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139794894","display_name":"Jingyu Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jingyu Wang","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.79269071,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"3562","last_page":"3575"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12535","display_name":"Machine Learning and Data Classification","score":0.13619999587535858,"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":0.13619999587535858,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.09040000289678574,"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/T10260","display_name":"Software Engineering Research","score":0.08739999681711197,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/personalization","display_name":"Personalization","score":0.38589999079704285},{"id":"https://openalex.org/keywords/automation","display_name":"Automation","score":0.28870001435279846},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.27149999141693115},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.26829999685287476},{"id":"https://openalex.org/keywords/work","display_name":"Work (physics)","score":0.2635999917984009}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.559499979019165},{"id":"https://openalex.org/C183003079","wikidata":"https://www.wikidata.org/wiki/Q1000371","display_name":"Personalization","level":2,"score":0.38589999079704285},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.3066999912261963},{"id":"https://openalex.org/C115901376","wikidata":"https://www.wikidata.org/wiki/Q184199","display_name":"Automation","level":2,"score":0.28870001435279846},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.27149999141693115},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.26829999685287476},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.2635999917984009},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.2597000002861023},{"id":"https://openalex.org/C2780598303","wikidata":"https://www.wikidata.org/wiki/Q65921492","display_name":"Flexibility (engineering)","level":2,"score":0.25870001316070557},{"id":"https://openalex.org/C147494362","wikidata":"https://www.wikidata.org/wiki/Q2078905","display_name":"Troubleshooting","level":2,"score":0.2583000063896179}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.174","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.174","pdf_url":"https://aclanthology.org/2026.findings-acl.174.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.174","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.174","pdf_url":"https://aclanthology.org/2026.findings-acl.174.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/8","score":0.46988147497177124,"display_name":"Decent work and economic growth"}],"awards":[{"id":"https://openalex.org/G2150238061","display_name":null,"funder_award_id":"2025YZ005","funder_id":"https://openalex.org/F4320321470","funder_display_name":"Beijing University of Posts and Telecommunications"},{"id":"https://openalex.org/G3473387614","display_name":null,"funder_award_id":"62101064","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3591376991","display_name":null,"funder_award_id":"62171057","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3800536456","display_name":null,"funder_award_id":"U23B2001","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3829840669","display_name":null,"funder_award_id":"MCM20200202","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3885225667","display_name":null,"funder_award_id":"JYB2025XDXM107","funder_id":"https://openalex.org/F4320321106","funder_display_name":"Ministry of Education of the People's Republic of China"},{"id":"https://openalex.org/G4196386516","display_name":null,"funder_award_id":"MCM20180101","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5482006598","display_name":null,"funder_award_id":"62401080","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G7921905313","display_name":null,"funder_award_id":"62406039","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G8129143660","display_name":null,"funder_award_id":"62321001","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G817674981","display_name":null,"funder_award_id":"62471055","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G8615917232","display_name":null,"funder_award_id":"62201072","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320321106","display_name":"Ministry of Education of the People's Republic of China","ror":"https://ror.org/01mv9t934"},{"id":"https://openalex.org/F4320321470","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59"},{"id":"https://openalex.org/F4320323970","display_name":"Ministry of Industry and Information Technology of the People's Republic of China","ror":"https://ror.org/0385nmy68"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166803187.pdf","grobid_xml":"https://content.openalex.org/works/W7166803187.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"massive":[1],"size":[2],"of":[3,119,156],"Large":[4],"Language":[5],"Models":[6],"(LLMs)":[7],"imposes":[8],"substantial":[9],"computational":[10,51],"and":[11,25,127],"storage":[12],"burdens,":[13],"particularly":[14],"on":[15,76],"devices":[16],"with":[17],"limited":[18],"hardware":[19],"resources.Compared":[20],"to":[21,55,68,158,163],"foundation":[22],"models,":[23],"smaller":[24],"more":[26,31],"specialized":[27],"models":[28,94],"are":[29],"often":[30],"suitable":[32],"for":[33,95],"practical":[34],"deployment.Existing":[35],"customization":[36],"approaches,":[37],"such":[38,120],"as":[39,116],"the":[40,98,105,147],"conventional":[41],"\"prune-then-finetune\"":[42],"paradigm":[43],"or":[44,53],"task-agnostic":[45],"deployment":[46],"strategies,":[47],"either":[48],"incur":[49],"excessive":[50],"costs":[52],"lead":[54],"suboptimal":[56],"task":[57],"performance.The":[58],"recently":[59],"popular":[60],"Mixture-of-Experts":[61],"(MoE)":[62],"architecture":[63],"exhibits":[64],"a":[65,73,84,117],"strong":[66],"ability":[67],"mitigate":[69],"inter-task":[70],"interference,":[71],"offering":[72],"new":[74],"perspective":[75],"model":[77],"deployment.In":[78],"this":[79],"paper,":[80],"we":[81,103],"introduce":[82],"ModularMoE,":[83],"training":[85],"framework":[86],"that":[87,140],"converts":[88],"pre-trained":[89],"LLMs":[90],"into":[91,108],"parametersharing":[92],"MoE":[93],"lightweight":[96],"deployment.Exploiting":[97],"emergent":[99],"modularity":[100],"within":[101,132],"LLMs,":[102],"split":[104],"feed-forward":[106],"layers":[107],"multiple":[109,136],"disjoint":[110],"modules.Each":[111],"expert":[112],"is":[113],"then":[114],"constructed":[115],"combination":[118],"modules,":[121],"enabling":[122],"knowledge":[123],"sharing":[124],"across":[125,135],"experts":[126],"thereby":[128],"improving":[129],"parameter":[130],"efficiency":[131],"MoEs.Extensive":[133],"experiments":[134],"downstream":[137],"tasks":[138],"demonstrate":[139],"ModularMoE":[141],"outperforms":[142],"other":[143],"state-ofthe-art":[144],"baselines":[145],"at":[146],"same":[148],"sparsity":[149],"level,":[150],"achieving":[151],"an":[152],"average":[153],"performance":[154],"improvement":[155],"4.10%":[157],"28.75%":[159],"while":[160],"delivering":[161],"up":[162],"2.71\u00d7":[164],"inference":[165],"speedup.":[166]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}