{"id":"https://openalex.org/W7166809168","doi":"https://doi.org/10.18653/v1/2026.acl-long.1778","title":"PHOTON: Hierarchical Autoregressive Modeling for Lightspeed and Memory-Efficient Language Generation","display_name":"PHOTON: Hierarchical Autoregressive Modeling for Lightspeed and Memory-Efficient Language Generation","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166809168","doi":"https://doi.org/10.18653/v1/2026.acl-long.1778"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.acl-long.1778","doi":"https://doi.org/10.18653/v1/2026.acl-long.1778","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1778","pdf_url":"https://aclanthology.org/2026.acl-long.1778.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2026.acl-long.1778.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5075950477","display_name":"Yuma Ichikawa","orcid":"https://orcid.org/0009-0004-4216-7017","observed_orcids":["https://orcid.org/0009-0004-4216-7017"]},"institutions":[{"id":"https://openalex.org/I4210110652","display_name":"RIKEN","ror":"https://ror.org/01sjwvz98","country_code":"JP","type":"facility","lineage":["https://openalex.org/I4210110652"]},{"id":"https://openalex.org/I4210126580","display_name":"RIKEN Center for Advanced Intelligence Project","ror":"https://ror.org/03ckxwf91","country_code":"JP","type":"facility","lineage":["https://openalex.org/I4210110652","https://openalex.org/I4210126580"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Yuma Ichikawa","raw_affiliation_strings":["Fujitsu Limited ,","RIKEN Center for AIP ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fujitsu Limited ,","institution_ids":[]},{"raw_affiliation_string":"RIKEN Center for AIP ,","institution_ids":["https://openalex.org/I4210110652","https://openalex.org/I4210126580"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121325134","display_name":"Naoya Takagi","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Naoya Takagi","raw_affiliation_strings":["Fujitsu Limited ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fujitsu Limited ,","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121355845","display_name":"Takumi Nakagawa","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4400009020","display_name":"Institute of Science Tokyo","ror":"https://ror.org/05dqf9946","country_code":"JP","type":"education","lineage":["https://openalex.org/I4400009020"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Takumi Nakagawa","raw_affiliation_strings":["Fujitsu Limited ,","Institute of Science Tokyo ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fujitsu Limited ,","institution_ids":[]},{"raw_affiliation_string":"Institute of Science Tokyo ,","institution_ids":["https://openalex.org/I4400009020"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011586391","display_name":"Yuzi Kanazawa","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuzi Kanazawa","raw_affiliation_strings":["Fujitsu Limited ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fujitsu Limited ,","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5052455821","display_name":"Akira Sakai","orcid":"https://orcid.org/0000-0003-0943-7842","observed_orcids":["https://orcid.org/0000-0003-0943-7842"]},"institutions":[{"id":"https://openalex.org/I1314466530","display_name":"Tokai University","ror":"https://ror.org/01p7qe739","country_code":"JP","type":"education","lineage":["https://openalex.org/I1314466530"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Akira Sakai","raw_affiliation_strings":["Fujitsu Limited ,","Tokai University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fujitsu Limited ,","institution_ids":[]},{"raw_affiliation_string":"Tokai University","institution_ids":["https://openalex.org/I1314466530"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.43609258,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"38365","last_page":"38383"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.5092999935150146,"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":0.5092999935150146,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.45969998836517334,"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":0.007799999788403511,"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/language-generation","display_name":"language generation","score":0.6340000033378601},{"id":"https://openalex.org/keywords/long-context-decoding","display_name":"long-context decoding","score":0.9190000295639038},{"id":"https://openalex.org/keywords/recursive-generation","display_name":"recursive generation","score":0.8939999938011169}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6588000059127808},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.45910000801086426},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3968999981880188},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3937000036239624},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.36820000410079956},{"id":"https://openalex.org/C2776187449","wikidata":"https://www.wikidata.org/wiki/Q1513879","display_name":"Natural language generation","level":3,"score":0.3208000063896179},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.31610000133514404},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3021000027656555},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.24729999899864197},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.24719999730587006}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.1778","doi":"https://doi.org/10.18653/v1/2026.acl-long.1778","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1778","pdf_url":"https://aclanthology.org/2026.acl-long.1778.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.1778","doi":"https://doi.org/10.18653/v1/2026.acl-long.1778","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1778","pdf_url":"https://aclanthology.org/2026.acl-long.1778.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320335125","display_name":"RIKEN","ror":"https://ror.org/01sjwvz98"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166809168.pdf","grobid_xml":"https://content.openalex.org/works/W7166809168.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4398156940","https://openalex.org/W4399759157","https://openalex.org/W4406880887","https://openalex.org/W7103751997","https://openalex.org/W7139126395","https://openalex.org/W3209873929","https://openalex.org/W7161172251","https://openalex.org/W4412042427","https://openalex.org/W4403580165","https://openalex.org/W7223019693"],"abstract_inverted_index":{"Transformers":[0],"operate":[1],"as":[2,29],"horizontal":[3,54],"token-bytoken":[4],"scanners;":[5],"at":[6],"each":[7],"generation":[8,89],"step,":[9],"attending":[10],"to":[11,107,131],"an":[12],"ever-growing":[13],"sequence":[14],"of":[15,64],"tokenlevel":[16],"states.This":[17],"access":[18],"pattern":[19],"increases":[20],"prefill":[21],"latency":[22],"and":[23,32,97,120],"makes":[24],"long-context":[25,119],"decoding":[26],"more":[27],"memory-bound,":[28],"KV-cache":[30,127],"reads":[31],"writes":[33],"dominate":[34],"inference":[35],"time":[36],"over":[37],"arithmetic":[38],"operations.We":[39],"propose":[40],"Parallel":[41],"Hierarchical":[42],"Operation":[43],"for":[44],"TOp-down":[45],"Networks":[46],"(PHOTON),":[47],"a":[48,62,67],"hierarchical":[49],"autoregressive":[50],"model":[51],"that":[52,90,103],"replaces":[53],"scanning":[55],"with":[56],"vertical,":[57],"multi-resolution":[58],"context":[59],"scanning.PHOTON":[60],"maintains":[61],"hierarchy":[63],"latent":[65,95],"streams:":[66],"bottom-up":[68,99],"encoder":[69],"compresses":[70],"tokens":[71],"into":[72],"low-rate":[73],"contextual":[74],"states,":[75],"while":[76],"lightweight":[77],"top-down":[78],"decoders":[79],"reconstruct":[80],"fine-grained":[81],"token":[82],"representations":[83],"in":[84,118],"parallel.We":[85],"further":[86],"introduce":[87],"recursive":[88],"updates":[91],"only":[92],"the":[93,113],"coarsest":[94],"stream":[96],"eliminates":[98],"re-encoding.Experimental":[100],"results":[101],"show":[102],"PHOTON":[104],"is":[105],"superior":[106],"competitive":[108],"Transformer-based":[109],"language":[110],"models":[111],"regarding":[112],"throughput-quality":[114],"tradeoff,":[115],"providing":[116],"advantages":[117],"multi-query":[121],"tasks.In":[122],"particular,":[123],"this":[124],"reduces":[125],"decode-time":[126],"traffic,":[128],"yielding":[129],"up":[130],"10":[132],"3":[133],"\u00d7":[134],"higher":[135],"throughput":[136],"per":[137],"unit":[138],"memory.":[139]},"counts_by_year":[],"updated_date":"2026-10-29T09:56:32.521200","created_date":"2026-07-02T00:00:00"}