{"id":"https://openalex.org/W4412164748","doi":"https://doi.org/10.1109/access.2025.3587387","title":"AsymGroup: Asymmetric Grouping and Communication Optimization for 2D Tensor Parallelism in LLM Inference","display_name":"AsymGroup: Asymmetric Grouping and Communication Optimization for 2D Tensor Parallelism in LLM Inference","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412164748","doi":"https://doi.org/10.1109/access.2025.3587387"},"language":"en","primary_location":{"id":"doi:10.1109/access.2025.3587387","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2025.3587387","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1109/access.2025.3587387","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5053477068","display_name":"K. Kim","orcid":"https://orcid.org/0000-0002-9115-7868"},"institutions":[{"id":"https://openalex.org/I193775966","display_name":"Yonsei University","ror":"https://ror.org/01wjejq96","country_code":"KR","type":"education","lineage":["https://openalex.org/I193775966"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Ki Tae Kim","raw_affiliation_strings":["Department of Electrical and Electronic Engineering, Yonsei University, Seoul, South Korea","Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea"],"raw_orcid":"https://orcid.org/0000-0002-9115-7868","affiliations":[{"raw_affiliation_string":"Department of Electrical and Electronic Engineering, Yonsei University, Seoul, South Korea","institution_ids":["https://openalex.org/I193775966"]},{"raw_affiliation_string":"Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea","institution_ids":["https://openalex.org/I193775966"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Seok-Ju Im","orcid":"https://orcid.org/0009-0003-4160-0758"},"institutions":[{"id":"https://openalex.org/I193775966","display_name":"Yonsei University","ror":"https://ror.org/01wjejq96","country_code":"KR","type":"education","lineage":["https://openalex.org/I193775966"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Seok-Ju Im","raw_affiliation_strings":["Department of Electrical and Electronic Engineering, Yonsei University, Seoul, South Korea"],"raw_orcid":"https://orcid.org/0009-0003-4160-0758","affiliations":[{"raw_affiliation_string":"Department of Electrical and Electronic Engineering, Yonsei University, Seoul, South Korea","institution_ids":["https://openalex.org/I193775966"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5084016800","display_name":"Eui-Young Chung","orcid":"https://orcid.org/0000-0003-2013-8763"},"institutions":[{"id":"https://openalex.org/I193775966","display_name":"Yonsei University","ror":"https://ror.org/01wjejq96","country_code":"KR","type":"education","lineage":["https://openalex.org/I193775966"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Eui-Young Chung","raw_affiliation_strings":["Department of Electrical and Electronic Engineering, Yonsei University, Seoul, South Korea","Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea"],"raw_orcid":"https://orcid.org/0000-0003-2013-8763","affiliations":[{"raw_affiliation_string":"Department of Electrical and Electronic Engineering, Yonsei University, Seoul, South Korea","institution_ids":["https://openalex.org/I193775966"]},{"raw_affiliation_string":"Department of Electrical and Electronic Engineering, Yonsei University, Seoul, Korea","institution_ids":["https://openalex.org/I193775966"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":{"value":1850,"currency":"USD","value_usd":1850},"apc_paid":{"value":1850,"currency":"USD","value_usd":1850},"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.1383433,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"13","issue":null,"first_page":"120591","last_page":"120602"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12303","display_name":"Tensor decomposition and applications","score":0.987500011920929,"subfield":{"id":"https://openalex.org/subfields/2605","display_name":"Computational Mathematics"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T12303","display_name":"Tensor decomposition and applications","score":0.987500011920929,"subfield":{"id":"https://openalex.org/subfields/2605","display_name":"Computational Mathematics"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9754999876022339,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T13650","display_name":"Computational Physics and Python Applications","score":0.9675999879837036,"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":0.7683881521224976},{"id":"https://openalex.org/keywords/parallelism","display_name":"Parallelism (grammar)","score":0.7675871253013611},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6978608965873718},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.5636327266693115},{"id":"https://openalex.org/keywords/tensor","display_name":"Tensor (intrinsic definition)","score":0.5005786418914795},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.37400397658348083},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.1780901849269867},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.16428562998771667}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7683881521224976},{"id":"https://openalex.org/C2781172179","wikidata":"https://www.wikidata.org/wiki/Q853109","display_name":"Parallelism (grammar)","level":2,"score":0.7675871253013611},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6978608965873718},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5636327266693115},{"id":"https://openalex.org/C155281189","wikidata":"https://www.wikidata.org/wiki/Q3518150","display_name":"Tensor (intrinsic definition)","level":2,"score":0.5005786418914795},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.37400397658348083},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.1780901849269867},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.16428562998771667},{"id":"https://openalex.org/C202444582","wikidata":"https://www.wikidata.org/wiki/Q837863","display_name":"Pure mathematics","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/access.2025.3587387","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2025.3587387","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},{"id":"pmh:oai:doaj.org/article:05c423aa336b4d708784c84b7b1a869a","is_oa":true,"landing_page_url":"https://doaj.org/article/05c423aa336b4d708784c84b7b1a869a","pdf_url":null,"source":{"id":"https://openalex.org/S4306401280","display_name":"DOAJ (DOAJ: Directory of Open Access Journals)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-sa","license_id":"https://openalex.org/licenses/cc-by-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"IEEE Access, Vol 13, Pp 120591-120602 (2025)","raw_type":"article"}],"best_oa_location":{"id":"doi:10.1109/access.2025.3587387","is_oa":true,"landing_page_url":"https://doi.org/10.1109/access.2025.3587387","pdf_url":null,"source":{"id":"https://openalex.org/S2485537415","display_name":"IEEE Access","issn_l":"2169-3536","issn":["2169-3536"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Access","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":30,"referenced_works":["https://openalex.org/W1558115001","https://openalex.org/W2147235956","https://openalex.org/W2170796499","https://openalex.org/W2973727699","https://openalex.org/W3039165326","https://openalex.org/W3072623287","https://openalex.org/W3092260380","https://openalex.org/W3097072756","https://openalex.org/W3129488589","https://openalex.org/W3159351344","https://openalex.org/W3206606249","https://openalex.org/W3209444800","https://openalex.org/W4221167110","https://openalex.org/W4288093768","https://openalex.org/W4322718191","https://openalex.org/W4384705453","https://openalex.org/W4394946008","https://openalex.org/W4395117348","https://openalex.org/W4396712969","https://openalex.org/W4404955085","https://openalex.org/W6767997687","https://openalex.org/W6769475105","https://openalex.org/W6790665546","https://openalex.org/W6799372109","https://openalex.org/W6802357070","https://openalex.org/W6810296985","https://openalex.org/W6839137647","https://openalex.org/W6846232141","https://openalex.org/W6850625674","https://openalex.org/W6856795522"],"related_works":["https://openalex.org/W2045183646","https://openalex.org/W2162409446","https://openalex.org/W2109463584","https://openalex.org/W2504075107","https://openalex.org/W2354058185","https://openalex.org/W2055243143","https://openalex.org/W3014500535","https://openalex.org/W2319812372","https://openalex.org/W1595672120","https://openalex.org/W4230999561"],"abstract_inverted_index":{"Recent":[0],"advances":[1],"in":[2,35,42,59],"large":[3],"language":[4,22],"models":[5],"(LLMs),":[6],"such":[7],"as":[8],"GPT":[9],"and":[10,45,91,99,120,143],"LLaMA,":[11],"have":[12],"demonstrated":[13],"remarkable":[14],"capabilities":[15],"across":[16,48],"a":[17,111,121],"wide":[18],"array":[19],"of":[20,85],"natural":[21],"processing":[23],"tasks.":[24],"Despite":[25],"these":[26],"successes,":[27],"efficient":[28,128],"inference":[29,137],"at":[30],"scale":[31],"remains":[32],"challenging,":[33],"particularly":[34],"heterogeneous":[36],"computing":[37],"environments":[38],"characterized":[39],"by":[40,139,146],"variations":[41],"GPU":[43],"counts,":[44],"computational":[46,98],"capacities":[47],"nodes.":[49],"Conventional":[50],"tensor":[51,78],"parallelism":[52,79],"approaches":[53],"typically":[54],"assume":[55],"homogeneous":[56],"hardware,":[57],"resulting":[58],"significant":[60],"performance":[61,158],"degradation":[62],"under":[63,159],"asymmetric":[64,76],"conditions.":[65,161],"To":[66],"address":[67],"this":[68],"challenge,":[69],"we":[70,109],"propose":[71],"<italic":[72],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[73],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">AsymGroup</i>,":[74],"an":[75],"2D":[77],"framework":[80],"that":[81,134],"dynamically":[82],"constructs":[83],"groups":[84],"uneven":[86],"sizes":[87],"based":[88],"on":[89],"node-":[90],"GPU-level":[92],"metrics.":[93],"AsymGroup":[94,135],"proportionally":[95],"allocates":[96],"both":[97],"communication":[100,113,144],"workloads":[101],"according":[102],"to":[103,116,125,141,148,151],"individual":[104],"device":[105],"capabilities.":[106],"In":[107],"addition,":[108],"introduce":[110],"formal":[112],"cost":[114],"model":[115],"accurately":[117],"quantify":[118],"bottlenecks":[119],"group":[122,129],"optimization":[123],"algorithm":[124],"systematically":[126],"determine":[127],"configurations.":[130],"Experimental":[131],"evaluations":[132],"demonstrate":[133],"reduces":[136],"latency":[138],"up":[140,147],"26.3%":[142],"overhead":[145],"33.4%":[149],"compared":[150],"state-of-the-art":[152],"frameworks,":[153],"while":[154],"also":[155],"maintaining":[156],"competitive":[157],"symmetric":[160]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}