{"id":"https://openalex.org/W4416965622","doi":"https://doi.org/10.1109/tcsi.2025.3638448","title":"TinyFormer: Efficient Sparse Transformer Design and Deployment on Tiny Devices","display_name":"TinyFormer: Efficient Sparse Transformer Design and Deployment on Tiny Devices","publication_year":2025,"publication_date":"2025-12-03","ids":{"openalex":"https://openalex.org/W4416965622","doi":"https://doi.org/10.1109/tcsi.2025.3638448"},"language":null,"primary_location":{"id":"doi:10.1109/tcsi.2025.3638448","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tcsi.2025.3638448","pdf_url":null,"source":{"id":"https://openalex.org/S116977442","display_name":"IEEE Transactions on Circuits and Systems I Regular Papers","issn_l":"1549-8328","issn":["1549-8328","1558-0806"],"is_oa":false,"is_in_doaj":false,"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":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Circuits and Systems I: Regular Papers","raw_type":"journal-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/A5053303853","display_name":"Jianlei Yang","orcid":"https://orcid.org/0000-0001-8424-7040"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Jianlei Yang","raw_affiliation_strings":["School of Computer Science and Engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-8424-7040","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5053082222","display_name":"Jiacheng Liao","orcid":"https://orcid.org/0000-0002-4037-9022"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiacheng Liao","raw_affiliation_strings":["School of Computer Science and Engineering, Beihang University, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101264329","display_name":"Fanding Lei","orcid":null},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Fanding Lei","raw_affiliation_strings":["School of Computer Science and Engineering, Beihang University, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100672863","display_name":"Meichen Liu","orcid":"https://orcid.org/0000-0002-3242-783X"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Meichen Liu","raw_affiliation_strings":["School of Computer Science and Engineering, Beihang University, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101291748","display_name":"Lingkun Long","orcid":null},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Lingkun Long","raw_affiliation_strings":["School of Computer Science and Engineering, Beihang University, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Junyi Chen","orcid":"https://orcid.org/0009-0003-7397-6311"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Junyi Chen","raw_affiliation_strings":["School of Computer Science and Engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0003-7397-6311","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101545018","display_name":"Han Wan","orcid":"https://orcid.org/0000-0002-9238-451X"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Han Wan","raw_affiliation_strings":["School of Computer Science and Engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0002-9238-451X","affiliations":[{"raw_affiliation_string":"School of Computer Science and Engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5051340429","display_name":"Bei Yu","orcid":"https://orcid.org/0000-0001-6406-4810"},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Bei Yu","raw_affiliation_strings":["Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong, SAR, China"],"raw_orcid":"https://orcid.org/0000-0001-6406-4810","affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong, SAR, China","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5066473925","display_name":"Weisheng Zhao","orcid":"https://orcid.org/0000-0001-8088-0404"},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Weisheng Zhao","raw_affiliation_strings":["Fert Beijing Research Institute, School of Integrated Circuit Science and Engineering, Beihang University, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-8088-0404","affiliations":[{"raw_affiliation_string":"Fert Beijing Research Institute, School of Integrated Circuit Science and Engineering, Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":9,"corresponding_author_ids":["https://openalex.org/A5053303853"],"corresponding_institution_ids":["https://openalex.org/I82880672"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.40732406,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"73","issue":"5","first_page":"3350","last_page":"3362"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.22499999403953552,"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"}},"topics":[{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.22499999403953552,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.19589999318122864,"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/T10363","display_name":"Low-power high-performance VLSI design","score":0.08959999680519104,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.6811000108718872},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.6456000208854675},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6258000135421753},{"id":"https://openalex.org/keywords/microcontroller","display_name":"Microcontroller","score":0.5220000147819519},{"id":"https://openalex.org/keywords/deep-learning","display_name":"Deep learning","score":0.40939998626708984}],"concepts":[{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6811000108718872},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6769999861717224},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.6456000208854675},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6258000135421753},{"id":"https://openalex.org/C173018170","wikidata":"https://www.wikidata.org/wiki/Q165678","display_name":"Microcontroller","level":2,"score":0.5220000147819519},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.44780001044273376},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.4341999888420105},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.42250001430511475},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.40939998626708984},{"id":"https://openalex.org/C46743427","wikidata":"https://www.wikidata.org/wiki/Q1341685","display_name":"Inference engine","level":3,"score":0.33739998936653137},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.3260999917984009},{"id":"https://openalex.org/C195672273","wikidata":"https://www.wikidata.org/wiki/Q6888132","display_name":"Model-based design","level":2,"score":0.3127000033855438},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3093000054359436},{"id":"https://openalex.org/C24326235","wikidata":"https://www.wikidata.org/wiki/Q126095","display_name":"Electronic engineering","level":1,"score":0.3046000003814697},{"id":"https://openalex.org/C56372850","wikidata":"https://www.wikidata.org/wiki/Q1050404","display_name":"Sparse matrix","level":3,"score":0.25699999928474426}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tcsi.2025.3638448","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tcsi.2025.3638448","pdf_url":null,"source":{"id":"https://openalex.org/S116977442","display_name":"IEEE Transactions on Circuits and Systems I Regular Papers","issn_l":"1549-8328","issn":["1549-8328","1558-0806"],"is_oa":false,"is_in_doaj":false,"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":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Circuits and Systems I: Regular Papers","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G437365050","display_name":null,"funder_award_id":"62572036","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G4759830254","display_name":null,"funder_award_id":"L243031","funder_id":"https://openalex.org/F4320334977","funder_display_name":"Beijing Municipal Natural Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320334977","display_name":"Beijing Municipal Natural Science Foundation","ror":null}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":19,"referenced_works":["https://openalex.org/W1677182931","https://openalex.org/W1994211684","https://openalex.org/W2183341477","https://openalex.org/W2194775991","https://openalex.org/W2963163009","https://openalex.org/W2967733054","https://openalex.org/W3013186616","https://openalex.org/W3016142271","https://openalex.org/W3092357178","https://openalex.org/W3096533519","https://openalex.org/W3097777922","https://openalex.org/W3134579744","https://openalex.org/W3167210698","https://openalex.org/W3197527602","https://openalex.org/W3213154195","https://openalex.org/W4312742569","https://openalex.org/W4385245566","https://openalex.org/W4386764393","https://openalex.org/W4401568395"],"related_works":[],"abstract_inverted_index":{"Developing":[0],"deep":[1,204],"learning":[2,205],"models":[3,31,62,111,133],"on":[4,34,63,134,138],"tiny":[5,35],"devices":[6,36],"(e.g.":[7,32],"Microcontroller":[8],"units,":[9],"MCUs)":[10],"has":[11],"attracted":[12],"much":[13],"attention":[14],"in":[15,170],"various":[16],"embedded":[17],"IoT":[18],"applications.":[19,206],"However,":[20],"it":[21],"is":[22,121,188],"challenging":[23],"to":[24,38,56,81,156,174,183,190,198],"efficiently":[25,106],"design":[26,146],"and":[27,58,74,162,197],"deploy":[28,59],"recent":[29],"advanced":[30],"transformers)":[33],"due":[37],"their":[39],"severe":[40],"hardware":[41,157],"resource":[42],"constraints.":[43],"In":[44],"this":[45],"work,":[46],"we":[47],"propose":[48],"<italic":[49,68,71,75],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[50,69,72,76,176],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">TinyFormer</i>,":[51],"a":[52,88],"framework":[53,125],"specifically":[54],"designed":[55],"develop":[57],"resource-efficient":[60],"transformer":[61,98,132],"MCUs.":[64,113,135],"TinyFormer":[65,144,166,187],"consists":[66],"of":[67,117,127,130,152,159,203],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">SuperNAS</i>,":[70],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">SparseNAS</i>,":[73],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">SparseEngine</i>.":[77],"Separately,":[78],"SuperNAS":[79],"aims":[80],"search":[82,90],"for":[83],"an":[84,150],"appropriate":[85],"supernet":[86],"from":[87,100],"vast":[89],"space.":[91],"SparseNAS":[92],"evaluates":[93],"the":[94,101,108,115,122,139,184,201],"best":[95,116],"sparse":[96,110,131,171],"single-path":[97],"model":[99],"identified":[102],"supernet.":[103],"Finally,":[104],"SparseEngine":[105,120],"deploys":[107],"searched":[109],"onto":[112],"To":[114],"our":[118],"knowledge,":[119],"first":[123],"deployment":[124],"capable":[126],"performing":[128],"inference":[129],"Evaluation":[136],"results":[137],"CIFAR-10":[140],"dataset":[141],"demonstrate":[142],"that":[143],"can":[145],"efficient":[147],"transformers":[148,193],"with":[149],"accuracy":[151],"96.1%":[153],"while":[154],"adhering":[155],"constraints":[158],"1MB":[160],"storage":[161],"320KB":[163],"memory.":[164],"Additionally,":[165],"achieves":[167],"significant":[168],"speedups":[169],"inference,":[172],"up":[173],"<inline-formula":[175],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">":[177],"<tex-math":[178],"notation=\"LaTeX\">$12.2\\times":[179],"$</tex-math>":[180],"</inline-formula>":[181],"comparing":[182],"CMSIS-NN":[185],"library.":[186],"believed":[189],"bring":[191],"powerful":[192],"into":[194],"TinyML":[195],"scenarios":[196],"greatly":[199],"expand":[200],"scope":[202]},"counts_by_year":[],"updated_date":"2026-05-05T06:06:40.768181","created_date":"2025-12-03T00:00:00"}