{"id":"https://openalex.org/W7137892271","doi":"https://doi.org/10.1609/aaai.v40i11.37822","title":"FlashKAT: Understanding and Addressing Performance Bottlenecks in the Kolmogorov-Arnold Transformer","display_name":"FlashKAT: Understanding and Addressing Performance Bottlenecks in the Kolmogorov-Arnold Transformer","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7137892271","doi":"https://doi.org/10.1609/aaai.v40i11.37822"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i11.37822","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i11.37822","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://doi.org/10.1609/aaai.v40i11.37822","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Matthew Raffel","orcid":null},"institutions":[{"id":"https://openalex.org/I131249849","display_name":"Oregon State University","ror":"https://ror.org/00ysfqy60","country_code":"US","type":"education","lineage":["https://openalex.org/I131249849"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Matthew Raffel","raw_affiliation_strings":["Oregon State University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Oregon State University","institution_ids":["https://openalex.org/I131249849"]}]},{"author_position":"last","author":{"id":null,"display_name":"Lizhong Chen","orcid":null},"institutions":[{"id":"https://openalex.org/I131249849","display_name":"Oregon State University","ror":"https://ror.org/00ysfqy60","country_code":"US","type":"education","lineage":["https://openalex.org/I131249849"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Lizhong Chen","raw_affiliation_strings":["Oregon State University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Oregon State University","institution_ids":["https://openalex.org/I131249849"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I131249849"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"11","first_page":"8694","last_page":"8702"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12611","display_name":"Neural Networks and Reservoir Computing","score":0.25110000371932983,"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/T12611","display_name":"Neural Networks and Reservoir Computing","score":0.25110000371932983,"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/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.18850000202655792,"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"}},{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.1582999974489212,"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/speedup","display_name":"Speedup","score":0.6851000189781189},{"id":"https://openalex.org/keywords/limiting","display_name":"Limiting","score":0.6333000063896179},{"id":"https://openalex.org/keywords/slowdown","display_name":"Slowdown","score":0.5503000020980835},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.5260000228881836},{"id":"https://openalex.org/keywords/flops","display_name":"FLOPS","score":0.49149999022483826},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.34200000762939453},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.337799996137619},{"id":"https://openalex.org/keywords/root-cause","display_name":"Root cause","score":0.32670000195503235}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7099000215530396},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6851000189781189},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.6333000063896179},{"id":"https://openalex.org/C2777242750","wikidata":"https://www.wikidata.org/wiki/Q31351","display_name":"Slowdown","level":2,"score":0.5503000020980835},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5260000228881836},{"id":"https://openalex.org/C3826847","wikidata":"https://www.wikidata.org/wiki/Q188768","display_name":"FLOPS","level":2,"score":0.49149999022483826},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.35100001096725464},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.34200000762939453},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.337799996137619},{"id":"https://openalex.org/C84945661","wikidata":"https://www.wikidata.org/wiki/Q7366567","display_name":"Root cause","level":2,"score":0.32670000195503235},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.319599986076355},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.31459999084472656},{"id":"https://openalex.org/C123593499","wikidata":"https://www.wikidata.org/wiki/Q6008583","display_name":"In-Memory Processing","level":5,"score":0.3052000105381012},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.29420000314712524},{"id":"https://openalex.org/C136625980","wikidata":"https://www.wikidata.org/wiki/Q663208","display_name":"Rounding","level":2,"score":0.27390000224113464},{"id":"https://openalex.org/C2986394398","wikidata":"https://www.wikidata.org/wiki/Q1805477","display_name":"Long memory","level":3,"score":0.26600000262260437},{"id":"https://openalex.org/C70153297","wikidata":"https://www.wikidata.org/wiki/Q5591907","display_name":"Gradient boosting","level":3,"score":0.2630999982357025},{"id":"https://openalex.org/C3017489831","wikidata":"https://www.wikidata.org/wiki/Q2393193","display_name":"Running time","level":2,"score":0.26159998774528503},{"id":"https://openalex.org/C5465570","wikidata":"https://www.wikidata.org/wiki/Q5326898","display_name":"Early stopping","level":3,"score":0.25029999017715454},{"id":"https://openalex.org/C149810388","wikidata":"https://www.wikidata.org/wiki/Q5374873","display_name":"Emulation","level":2,"score":0.2502000033855438}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i11.37822","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i11.37822","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i11.37822","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i11.37822","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"Kolmogorov-Arnold":[1,50],"Network":[2],"(KAN)":[3],"has":[4,53],"been":[5,54],"gaining":[6],"popularity":[7],"as":[8],"an":[9,168],"alternative":[10],"to":[11,17,37,45,59,103,121,127,148,167],"the":[12,49,72,105,109,116,132,152],"multilayer":[13],"perceptron":[14],"(MLP)":[15],"due":[16,36],"its":[18,38,43],"greater":[19],"expressiveness":[20],"and":[21,30,151],"interpretability.":[22],"Even":[23],"so,":[24],"KAN":[25,68],"suffers":[26],"from":[27],"training":[28,170],"instability":[29],"being":[31],"orders":[32],"of":[33,101,108,135,154],"magnitude":[34],"slower":[35,82],"increased":[39],"computational":[40],"cost,":[41],"limiting":[42],"applicability":[44],"large-scale":[46],"tasks.":[47],"Recently,":[48],"Transformer":[51,61],"(KAT)":[52],"proposed,":[55],"achieving":[56],"FLOPs":[57],"comparable":[58,73],"traditional":[60],"models":[62],"with":[63],"MLPs":[64],"by":[65],"leveraging":[66],"Group-Rational":[67],"(GR-KAN).":[69],"Unfortunately,":[70],"despite":[71],"FLOPs,":[74],"our":[75],"testing":[76],"shows":[77],"that":[78,86,115,163],"KAT":[79,174],"remains":[80],"123x":[81],"during":[83],"training,":[84],"indicating":[85],"there":[87],"are":[88],"other":[89],"performance":[90],"bottlenecks":[91],"beyond":[92],"FLOPs.":[93],"In":[94],"this":[95,139],"paper,":[96],"we":[97,142],"conduct":[98],"a":[99,158],"series":[100],"experiments":[102],"understand":[104],"root":[106],"cause":[107],"slowdown":[110,117],"in":[111,131,179],"KAT.":[112],"We":[113],"uncover":[114],"can":[118],"be":[119],"isolated":[120],"memory":[122,140,150],"stalls,":[123],"linked":[124],"more":[125],"specifically":[126],"inefficient":[128],"gradient":[129,180],"accumulations":[130],"backward":[133],"pass":[134],"GR-KAN.":[136],"To":[137],"address":[138],"bottleneck,":[141],"propose":[143],"FlashKAT,":[144],"which":[145],"minimizes":[146],"accesses":[147],"slow":[149],"usage":[153],"atomic":[155],"adds":[156],"through":[157],"restructured":[159],"kernel.":[160],"Evaluations":[161],"show":[162],"FlashKAT":[164],"achieves":[165],"up":[166],"86.5x":[169],"speedup":[171],"over":[172],"state-of-the-art":[173],"while":[175],"reducing":[176],"rounding":[177],"errors":[178],"computation.":[181]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-02-12T00:00:00"}