{"id":"https://openalex.org/W4380875572","doi":"https://doi.org/10.1145/3579371.3589040","title":"TaskFusion: An Efficient Transfer Learning Architecture with Dual Delta Sparsity for Multi-Task Natural Language Processing","display_name":"TaskFusion: An Efficient Transfer Learning Architecture with Dual Delta Sparsity for Multi-Task Natural Language Processing","publication_year":2023,"publication_date":"2023-06-16","ids":{"openalex":"https://openalex.org/W4380875572","doi":"https://doi.org/10.1145/3579371.3589040"},"language":"en","primary_location":{"id":"doi:10.1145/3579371.3589040","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3579371.3589040","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 50th Annual International Symposium on Computer Architecture","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/A5008585508","display_name":"Zichen Fan","orcid":"https://orcid.org/0000-0002-8181-2996"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Zichen Fan","raw_affiliation_strings":["Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA"],"raw_orcid":"https://orcid.org/0000-0002-8181-2996","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036388111","display_name":"Qirui Zhang","orcid":"https://orcid.org/0000-0001-8113-3558"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Qirui Zhang","raw_affiliation_strings":["Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA"],"raw_orcid":"https://orcid.org/0000-0001-8113-3558","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5069810675","display_name":"Pierre Abillama","orcid":"https://orcid.org/0000-0001-9523-6152"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Pierre Abillama","raw_affiliation_strings":["Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA"],"raw_orcid":"https://orcid.org/0000-0001-9523-6152","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5048967406","display_name":"Sara Shoouri","orcid":"https://orcid.org/0009-0004-7699-7852"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Sara Shoouri","raw_affiliation_strings":["Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA"],"raw_orcid":"https://orcid.org/0009-0004-7699-7852","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026925435","display_name":"C. S. George Lee","orcid":"https://orcid.org/0000-0002-5610-0671"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Changwoo Lee","raw_affiliation_strings":["Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA"],"raw_orcid":"https://orcid.org/0000-0002-5610-0671","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026311377","display_name":"David Blaauw","orcid":"https://orcid.org/0000-0001-6744-7075"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"David Blaauw","raw_affiliation_strings":["Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA"],"raw_orcid":"https://orcid.org/0000-0001-6744-7075","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014196508","display_name":"Hun-Seok Kim","orcid":"https://orcid.org/0000-0002-6658-5502"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hun-Seok Kim","raw_affiliation_strings":["Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA"],"raw_orcid":"https://orcid.org/0000-0002-6658-5502","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA","institution_ids":["https://openalex.org/I27837315"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5000767141","display_name":"Dennis Sylvester","orcid":"https://orcid.org/0000-0003-2598-0458"},"institutions":[{"id":"https://openalex.org/I27837315","display_name":"University of Michigan","ror":"https://ror.org/00jmfr291","country_code":"US","type":"education","lineage":["https://openalex.org/I27837315"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Dennis Sylvester","raw_affiliation_strings":["Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA"],"raw_orcid":"https://orcid.org/0000-0003-2598-0458","affiliations":[{"raw_affiliation_string":"Electrical and Computer Engineering, University of Michigan, Ann Arbor, Ann Arbor, MI, USA","institution_ids":["https://openalex.org/I27837315"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":8,"corresponding_author_ids":["https://openalex.org/A5008585508"],"corresponding_institution_ids":["https://openalex.org/I27837315"],"apc_list":null,"apc_paid":null,"fwci":2.2153,"has_fulltext":false,"cited_by_count":13,"citation_normalized_percentile":{"value":0.90064529,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":90,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"14"},"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":0.9997000098228455,"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":0.9997000098228455,"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":0.9987999796867371,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.9980999827384949,"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":0.8511717915534973},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.5808445811271667},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5444501638412476},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.5426218509674072},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.49751022458076477},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.4473242461681366},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.44721531867980957},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.43889087438583374},{"id":"https://openalex.org/keywords/transfer-of-learning","display_name":"Transfer of learning","score":0.42413920164108276},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.4109901189804077},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.3817247450351715},{"id":"https://openalex.org/keywords/computer-architecture","display_name":"Computer architecture","score":0.37204521894454956},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.3313825726509094},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.13920602202415466}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8511717915534973},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.5808445811271667},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5444501638412476},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.5426218509674072},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.49751022458076477},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.4473242461681366},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.44721531867980957},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43889087438583374},{"id":"https://openalex.org/C150899416","wikidata":"https://www.wikidata.org/wiki/Q1820378","display_name":"Transfer of learning","level":2,"score":0.42413920164108276},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.4109901189804077},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3817247450351715},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.37204521894454956},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3313825726509094},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.13920602202415466},{"id":"https://openalex.org/C187736073","wikidata":"https://www.wikidata.org/wiki/Q2920921","display_name":"Management","level":1,"score":0.0},{"id":"https://openalex.org/C21547014","wikidata":"https://www.wikidata.org/wiki/Q1423657","display_name":"Operations management","level":1,"score":0.0},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.0},{"id":"https://openalex.org/C76155785","wikidata":"https://www.wikidata.org/wiki/Q418","display_name":"Telecommunications","level":1,"score":0.0},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3579371.3589040","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3579371.3589040","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 50th Annual International Symposium on Computer Architecture","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7","score":0.8799999952316284}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":23,"referenced_works":["https://openalex.org/W1516184288","https://openalex.org/W2135046866","https://openalex.org/W2165698076","https://openalex.org/W2285660444","https://openalex.org/W2606722458","https://openalex.org/W2794952988","https://openalex.org/W2978670439","https://openalex.org/W2979310060","https://openalex.org/W2979439447","https://openalex.org/W3016542674","https://openalex.org/W3016735325","https://openalex.org/W3016832937","https://openalex.org/W3047848469","https://openalex.org/W3100985894","https://openalex.org/W3103168911","https://openalex.org/W3155922894","https://openalex.org/W3206453033","https://openalex.org/W3207622241","https://openalex.org/W4214686755","https://openalex.org/W4224267386","https://openalex.org/W4240168186","https://openalex.org/W4303684596","https://openalex.org/W4313156423"],"related_works":["https://openalex.org/W17155033","https://openalex.org/W3207760230","https://openalex.org/W1496222301","https://openalex.org/W1590307681","https://openalex.org/W2536018345","https://openalex.org/W4312814274","https://openalex.org/W4285370786","https://openalex.org/W2296488620","https://openalex.org/W2358353312","https://openalex.org/W3203561214"],"abstract_inverted_index":{"The":[0],"combination":[1],"of":[2,30,82,129,234,257],"pre-trained":[3],"models":[4,32,67],"and":[5,27,74,111,131,142,165,220,268,276],"task-specific":[6],"fine-tuning":[7],"schemes,":[8],"such":[9,31],"as":[10],"BERT,":[11],"has":[12],"achieved":[13],"great":[14],"success":[15],"in":[16,39,43,162,242],"various":[17],"natural":[18],"language":[19],"processing":[20],"(NLP)":[21],"tasks.":[22,83,172],"However,":[23],"the":[24,72,80,97,117,197,232,255,265],"large":[25],"memory":[26,75,223],"computation":[28,108],"costs":[29],"make":[33],"it":[34],"challenging":[35],"to":[36,52,56,95,139,167,182,195,208,216],"deploy":[37],"them":[38],"edge":[40],"devices.":[41],"Moreover,":[42],"real-world":[44],"applications":[45],"like":[46],"chatbots,":[47],"multiple":[48,62,106],"NLP":[49,63,244,288],"tasks":[50,64,115],"need":[51],"be":[53,134],"processed":[54],"together":[55],"achieve":[57,273],"higher":[58],"response":[59],"credibility.":[60],"Running":[61],"with":[65,79,212,245],"specialized":[66],"for":[68,144,286],"each":[69],"task":[70,253],"increases":[71],"latency":[73,77,143],"cost":[76,141,256],"linearly":[78],"number":[81,233],"Though":[84],"there":[85],"have":[86],"been":[87],"recent":[88],"works":[89],"on":[90,179],"parameter-shared":[91],"tuning":[92],"that":[93,125,158,228],"aim":[94],"reduce":[96,140,221,231],"total":[98],"parameter":[99,261],"size":[100,262],"by":[101,239],"partially":[102],"sharing":[103,170],"weights":[104,132,164],"among":[105,136,171],"tasks,":[107,138],"remains":[109],"intensive":[110],"redundant":[112],"despite":[113],"different":[114,137],"using":[116,282],"same":[118],"input.":[119],"In":[120],"this":[121],"work,":[122],"we":[123,149],"identify":[124],"a":[126,204,251],"significant":[127],"portion":[128],"activations":[130,166],"can":[133,230,272],"reused":[135],"efficient":[145,153],"multi-task":[146,210,243,287],"NLP.":[147],"Specifically,":[148],"propose":[150],"TaskFusion,":[151],"an":[152,213],"transfer":[154],"learning":[155],"software-hardware":[156],"co-design":[157],"exploits":[159],"delta":[160,180,199],"sparsity":[161],"both":[163],"boost":[168],"data":[169,185],"For":[173],"training,":[174],"TaskFusion":[175,229],"uses":[176],"\u21131":[177],"regularization":[178],"activation":[181],"learn":[183],"inter-task":[184],"redundancies.":[186],"A":[187],"novel":[188],"hardware-aware":[189],"sub-task":[190],"inference":[191,211],"algorithm":[192],"is":[193],"proposed":[194,266],"exploit":[196],"dual":[198],"sparsity.":[200],"We":[201],"then":[202],"designed":[203],"dedicated":[205],"heterogeneous":[206],"architecture":[207,267],"accelerate":[209],"optimized":[214,269],"scheduling":[215],"increase":[217],"hardware":[218],"utilization":[219],"off-chip":[222],"access.":[224],"Extensive":[225],"experiments":[226],"demonstrate":[227],"floating":[235],"point":[236],"operations":[237],"(FLOPs)":[238],"over":[240],"73%":[241],"negligible":[246],"accuracy":[247],"loss,":[248],"while":[249],"adding":[250],"new":[252],"at":[254],"only":[258],"<":[259],"2%":[260],"increase.":[263],"With":[264],"scheduling,":[270],"Task-Fusion":[271],"1.48--2.43\u00d7":[274],"performance":[275],"1.62--3.77\u00d7":[277],"energy":[278],"efficiency":[279],"than":[280],"those":[281],"state-of-the-art":[283],"single-task":[284],"accelerators":[285],"applications.":[289]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":6},{"year":2023,"cited_by_count":1}],"updated_date":"2025-11-06T03:46:38.306776","created_date":"2025-10-10T00:00:00"}