{"id":"https://openalex.org/W4404386011","doi":"https://doi.org/10.1145/3698038.3698521","title":"MoEsaic: Shared Mixture of Experts","display_name":"MoEsaic: Shared Mixture of Experts","publication_year":2024,"publication_date":"2024-11-14","ids":{"openalex":"https://openalex.org/W4404386011","doi":"https://doi.org/10.1145/3698038.3698521"},"language":"en","primary_location":{"id":"doi:10.1145/3698038.3698521","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3698038.3698521","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM Symposium on Cloud Computing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5035364139","display_name":"U. D. Deshpande","orcid":"https://orcid.org/0000-0001-9506-0003"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Umesh Deshpande","raw_affiliation_strings":["IBM Research, USA"],"affiliations":[{"raw_affiliation_string":"IBM Research, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076308023","display_name":"Travis Janssen","orcid":"https://orcid.org/0009-0007-3033-3964"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Travis Janssen","raw_affiliation_strings":["IBM Research, USA"],"affiliations":[{"raw_affiliation_string":"IBM Research, USA","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5072108555","display_name":"Mudhakar Srivatsa","orcid":"https://orcid.org/0000-0002-5874-3750"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mudhakar Srivatsa","raw_affiliation_strings":["IBM Research, USA"],"affiliations":[{"raw_affiliation_string":"IBM Research, USA","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5099587073","display_name":"Swaminathan Sundararaman","orcid":"https://orcid.org/0000-0003-4468-3061"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Swaminathan Sundararaman","raw_affiliation_strings":["IBM Research, USA"],"affiliations":[{"raw_affiliation_string":"IBM Research, USA","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5035364139"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.19419137,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"434","last_page":"442"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10064","display_name":"Complex Network Analysis Techniques","score":0.9891999959945679,"subfield":{"id":"https://openalex.org/subfields/3109","display_name":"Statistical and Nonlinear Physics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10064","display_name":"Complex Network Analysis Techniques","score":0.9891999959945679,"subfield":{"id":"https://openalex.org/subfields/3109","display_name":"Statistical and Nonlinear Physics"},"field":{"id":"https://openalex.org/fields/31","display_name":"Physics and Astronomy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10824","display_name":"Image Retrieval and Classification Techniques","score":0.9876000285148621,"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"}},{"id":"https://openalex.org/T10799","display_name":"Data Visualization and Analytics","score":0.9846000075340271,"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/computer-science","display_name":"Computer science","score":0.5881110429763794}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5881110429763794}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3698038.3698521","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3698038.3698521","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the ACM Symposium on Cloud Computing","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Climate action","score":0.6000000238418579,"id":"https://metadata.un.org/sdg/13"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":9,"referenced_works":["https://openalex.org/W2809290718","https://openalex.org/W3175781807","https://openalex.org/W4385567093","https://openalex.org/W4386076670","https://openalex.org/W4388514131","https://openalex.org/W4390833687","https://openalex.org/W4390874259","https://openalex.org/W4392679398","https://openalex.org/W4404782031"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Mixture":[0,104],"of":[1,6,19,54,105],"Expert":[2],"(MoE)":[3],"models":[4],"consist":[5],"several":[7],"experts,":[8],"each":[9,78],"specializing":[10],"in":[11,60,93,107],"a":[12,17],"specific":[13],"task.":[14],"During":[15],"inference,":[16],"subset":[18],"the":[20,29,88,127,131,141],"experts":[21,55,73,81,115,133],"is":[22],"invoked":[23,76],"based":[24],"on":[25,167],"their":[26,37,121],"relevance":[27],"to":[28,45,86,135,152],"request.":[30],"MoE's":[31],"modular":[32],"architecture":[33],"lets":[34],"users":[35],"compose":[36],"model":[38,57,67,117,162],"from":[39],"popular":[40],"off-the-shelf":[41],"experts.":[42,51],"This":[43],"leads":[44],"multiple":[46],"MoE":[47,155],"deployments":[48],"with":[49,164],"identical":[50,114,132],"The":[52],"duplication":[53],"across":[56,116],"instances":[58,163],"results":[59],"excessive":[61],"GPU":[62,95],"memory":[63,122],"consumption":[64],"and":[65,112],"increased":[66],"serving":[68],"cost.":[69],"Moreover,":[70,124],"since":[71],"all":[72],"are":[74],"not":[75],"for":[77,147],"request,":[79],"individual":[80],"rarely":[82],"receive":[83],"enough":[84],"requests":[85,128],"exploit":[87],"GPUs'":[89],"computational":[90],"capabilities,":[91],"resulting":[92],"low":[94],"utilization.":[96],"To":[97],"address":[98],"these":[99],"problems,":[100],"we":[101],"propose":[102],"Shared":[103],"Experts":[106],"MoEsaic.":[108],"MoEsaic":[109,157],"automatically":[110],"identifies":[111],"deduplicates":[113],"instances,":[118,156],"thus":[119],"reducing":[120],"footprint.":[123],"it":[125],"batches":[126],"directed":[129],"toward":[130],"belonging":[134],"different":[136],"clients,":[137],"which":[138],"also":[139],"improves":[140],"processing":[142],"efficiency.":[143],"We":[144],"show":[145],"that":[146],"Mixtral-8x7B":[148],"model,":[149],"when":[150],"compared":[151],"deploying":[153],"dedicated":[154],"can":[158],"serve":[159],"7X":[160],"more":[161],"little":[165],"impact":[166],"inference":[168],"performance.":[169]},"counts_by_year":[],"updated_date":"2025-12-27T23:08:20.325037","created_date":"2025-10-10T00:00:00"}