{"id":"https://openalex.org/W2997517014","doi":"https://doi.org/10.1162/tacl_a_00353","title":"Efficient Content-Based Sparse Attention with Routing Transformers","display_name":"Efficient Content-Based Sparse Attention with Routing Transformers","publication_year":2021,"publication_date":"2021-02-01","ids":{"openalex":"https://openalex.org/W2997517014","doi":"https://doi.org/10.1162/tacl_a_00353","mag":"2997517014"},"language":"en","primary_location":{"id":"doi:10.1162/tacl_a_00353","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00353","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00353/1923932/tacl_a_00353.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"type":"article","indexed_in":["arxiv","crossref","datacite","doaj"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00353/1923932/tacl_a_00353.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5110616102","display_name":"Aurko Roy","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Aurko Roy","raw_affiliation_strings":["Google Research","Google,,,,,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]},{"raw_affiliation_string":"Google,,,,,","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5038670125","display_name":"Mohammad Saffar","orcid":null},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Mohammad Saffar","raw_affiliation_strings":["Google Research","Google,,,,,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]},{"raw_affiliation_string":"Google,,,,,","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103024730","display_name":"Ashish Vaswani","orcid":"https://orcid.org/0000-0002-7794-2085"},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ashish Vaswani","raw_affiliation_strings":["Google Research","Google,,,,,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]},{"raw_affiliation_string":"Google,,,,,","institution_ids":["https://openalex.org/I1291425158"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5065912572","display_name":"David Grangier","orcid":"https://orcid.org/0000-0002-8847-9532"},"institutions":[{"id":"https://openalex.org/I1291425158","display_name":"Google (United States)","ror":"https://ror.org/00njsd438","country_code":"US","type":"company","lineage":["https://openalex.org/I1291425158","https://openalex.org/I4210128969"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"David Grangier","raw_affiliation_strings":["Google Research","Google,,,,,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Google Research","institution_ids":["https://openalex.org/I1291425158"]},{"raw_affiliation_string":"Google,,,,,","institution_ids":["https://openalex.org/I1291425158"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I1291425158"],"apc_list":null,"apc_paid":null,"fwci":4.4282,"has_fulltext":false,"cited_by_count":49,"citation_normalized_percentile":{"value":0.95140647,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":96,"max":100},"biblio":{"volume":"9","issue":null,"first_page":"53","last_page":"68"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9998000264167786,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9975000023841858,"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/perplexity","display_name":"Perplexity","score":0.9641983509063721},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.663551390171051},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.6047618985176086},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5004851818084717},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.4301314651966095},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4184947609901428},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3526003062725067},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.35165169835090637},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.3394821882247925},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.29465678334236145}],"concepts":[{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.9641983509063721},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.663551390171051},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.6047618985176086},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5004851818084717},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.4301314651966095},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4184947609901428},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3526003062725067},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.35165169835090637},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3394821882247925},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.29465678334236145},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0},{"id":"https://openalex.org/C165801399","wikidata":"https://www.wikidata.org/wiki/Q25428","display_name":"Voltage","level":2,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0}],"mesh":[],"locations_count":5,"locations":[{"id":"doi:10.1162/tacl_a_00353","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00353","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00353/1923932/tacl_a_00353.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},{"id":"pmh:oai:arXiv.org:2003.05997","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2003.05997","pdf_url":"https://arxiv.org/pdf/2003.05997","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"","raw_type":"text"},{"id":"mag:2997517014","is_oa":true,"landing_page_url":"https://arxiv.org/pdf/2003.05997.pdf","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"arXiv (Cornell University)","raw_type":null},{"id":"pmh:oai:doaj.org/article:21cd91f375224260808b3967ff54015d","is_oa":false,"landing_page_url":"https://doaj.org/article/21cd91f375224260808b3967ff54015d","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Transactions of the Association for Computational Linguistics, Vol 9, Pp 53-68 (2021)","raw_type":"article"},{"id":"doi:10.48550/arxiv.2003.05997","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2003.05997","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.1162/tacl_a_00353","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00353","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00353/1923932/tacl_a_00353.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.5699999928474426}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":46,"referenced_works":["https://openalex.org/W97540112","https://openalex.org/W854541894","https://openalex.org/W980713158","https://openalex.org/W1514535095","https://openalex.org/W1593114658","https://openalex.org/W1902237438","https://openalex.org/W2115665694","https://openalex.org/W2149521737","https://openalex.org/W2165141262","https://openalex.org/W2242818861","https://openalex.org/W2405210557","https://openalex.org/W2491408735","https://openalex.org/W2540419089","https://openalex.org/W2571859396","https://openalex.org/W2792376130","https://openalex.org/W2914526845","https://openalex.org/W2940744433","https://openalex.org/W2946567085","https://openalex.org/W2949335953","https://openalex.org/W2950527759","https://openalex.org/W2950635152","https://openalex.org/W2950946978","https://openalex.org/W2952339051","https://openalex.org/W2952406142","https://openalex.org/W2962741254","https://openalex.org/W2962839844","https://openalex.org/W2962911098","https://openalex.org/W2963062480","https://openalex.org/W2963139417","https://openalex.org/W2963341956","https://openalex.org/W2963403868","https://openalex.org/W2963428348","https://openalex.org/W2963494889","https://openalex.org/W2963631907","https://openalex.org/W2963636093","https://openalex.org/W2963729263","https://openalex.org/W2963925437","https://openalex.org/W2964110616","https://openalex.org/W2964121744","https://openalex.org/W2964272710","https://openalex.org/W2964308564","https://openalex.org/W2970597249","https://openalex.org/W2970777192","https://openalex.org/W2996287690","https://openalex.org/W3000514857","https://openalex.org/W3130660608"],"related_works":["https://openalex.org/W2940744433","https://openalex.org/W2963403868","https://openalex.org/W3015468748","https://openalex.org/W2963341956","https://openalex.org/W3131922516","https://openalex.org/W2994673210","https://openalex.org/W2946567085","https://openalex.org/W3045733172","https://openalex.org/W3033529678","https://openalex.org/W3091156754","https://openalex.org/W3085139254","https://openalex.org/W3034573343","https://openalex.org/W2965373594","https://openalex.org/W2964110616","https://openalex.org/W3021293129","https://openalex.org/W2998108143","https://openalex.org/W2986922898","https://openalex.org/W2981852735","https://openalex.org/W2984864519","https://openalex.org/W2194775991"],"abstract_inverted_index":{"Self-attention":[0],"has":[1],"recently":[2],"been":[3],"adopted":[4],"for":[5,141,220],"a":[6,43,116,188,198,204],"wide":[7],"range":[8],"of":[9,46,49,74,82,89,129,201,213],"sequence":[10,27,142],"modeling":[11,87,161],"problems.":[12],"Despite":[13],"its":[14],"effectiveness,":[15],"self-attention":[16,114,183],"suffers":[17],"from":[18,100,136],"quadratic":[19],"computation":[20,63],"and":[21,64,145],"memory":[22,65],"requirements":[23],"with":[24,96,115,203],"respect":[25],"to":[26,31,38,54,66,68,71,131],"length.":[28],"Successful":[29],"approaches":[30,101],"reduce":[32],"this":[33],"complexity":[34,128],"focused":[35],"on":[36,92,103,121,159,162,171,174,191,211],"attending":[37],"local":[39],"sliding":[40],"windows":[41],"or":[42],"small":[44],"set":[45,187],"locations":[47],"independent":[48],"content.":[50],"Our":[51,108],"work":[52,77,91],"proposes":[53],"learn":[55],"dynamic":[56],"sparse":[57,94,106,117,156],"attention":[58,95,130,157],"patterns":[59],"that":[60,151],"avoid":[61],"allocating":[62],"attend":[67],"content":[69],"unrelated":[70],"the":[72,86,97,110,126,192,218],"query":[73],"interest.":[75],"This":[76],"builds":[78],"upon":[79],"two":[80],"lines":[81],"research:":[83],"It":[84],"combines":[85],"flexibility":[88],"prior":[90],"content-based":[93],"efficiency":[98],"gains":[99],"based":[102,120],"local,":[104],"temporal":[105],"attention.":[107],"model,":[109],"Routing":[111,207,221],"Transformer,":[112],"endows":[113],"routing":[118],"module":[119],"online":[122],"k-means":[123],"while":[124,180],"reducing":[125],"overall":[127],"O(":[132,137],"n":[133,138,144],"1.5":[134],"d)":[135,140],"2":[139],"length":[143,214],"hidden":[146],"dimension":[147],"d.":[148],"We":[149,216],"show":[150],"our":[152],"model":[153,209],"outperforms":[154],"comparable":[155],"models":[158],"language":[160],"Wikitext-103":[163],"(15.8":[164],"vs":[165,177],"18.3":[166],"perplexity),":[167],"as":[168,170],"well":[169],"image":[172],"generation":[173],"ImageNet-64":[175],"(3.43":[176],"3.44":[178],"bits/dim)":[179],"using":[181],"fewer":[182],"layers.":[184],"Additionally,":[185],"we":[186],"new":[189],"state-of-the-art":[190],"newly":[193],"released":[194],"PG-19":[195],"data-set,":[196],"obtaining":[197],"test":[199],"perplexity":[200],"33.2":[202],"22":[205],"layer":[206],"Transformer":[208,222],"trained":[210],"sequences":[212],"8192.":[215],"open-source":[217],"code":[219],"in":[223],"Tensorflow.":[224],"1":[225]},"counts_by_year":[{"year":2025,"cited_by_count":3},{"year":2024,"cited_by_count":5},{"year":2023,"cited_by_count":7},{"year":2022,"cited_by_count":5},{"year":2021,"cited_by_count":17},{"year":2020,"cited_by_count":12}],"updated_date":"2026-08-01T09:00:35.917206","created_date":"2025-10-10T00:00:00"}