{"id":"https://openalex.org/W7214745185","doi":"https://doi.org/10.48550/arxiv.2609.33465","title":"GSM: Efficient Language Modeling with Shared Global State","display_name":"GSM: Efficient Language Modeling with Shared Global State","publication_year":2026,"publication_date":"2026-09-27","ids":{"openalex":"https://openalex.org/W7214745185","doi":"https://doi.org/10.48550/arxiv.2609.33465"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.33465","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33465","pdf_url":"https://arxiv.org/pdf/2609.33465","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,"listed_in":[],"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":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.33465","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101237514","display_name":"Yunao Zheng","orcid":"https://orcid.org/0009-0002-4583-3607","observed_orcids":["https://orcid.org/0009-0002-4583-3607"]},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zheng, Yunao","raw_affiliation_strings":["Beijing University of Posts and Telecommunications"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153533152","display_name":"Bin Wen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4401726859","display_name":"Kuaishou (China)","ror":"https://ror.org/0258as409","country_code":"CN","type":"company","lineage":["https://openalex.org/I4401726859"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wen, Bin","raw_affiliation_strings":["Kuaishou Technology "],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Kuaishou Technology ","institution_ids":["https://openalex.org/I4401726859"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153431385","display_name":"Xiaojie Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Xiaojie","raw_affiliation_strings":["Beijing University of Posts and Telecommunications"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125592949","display_name":"Kaiyu Jiang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiang, Kaiyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120596126","display_name":"Xuanyu Zheng","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Xuanyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125550899","display_name":"Changyi Liu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Changyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140274753","display_name":"Hongyi Fu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fu, Hongyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153471914","display_name":"Jianxiong Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Jianxiong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153411090","display_name":"Tianke Zhang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Tianke","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153516664","display_name":"Haonan Fan","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fan, Haonan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5152513939","display_name":"Yingxin Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Yingxin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130099817","display_name":"Jiankang Chen","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Jiankang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153531033","display_name":"Xu Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I139759216","display_name":"Beijing University of Posts and Telecommunications","ror":"https://ror.org/04w9fbh59","country_code":"CN","type":"education","lineage":["https://openalex.org/I139759216"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Xu","raw_affiliation_strings":["Beijing University of Posts and Telecommunications"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beijing University of Posts and Telecommunications","institution_ids":["https://openalex.org/I139759216"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125518543","display_name":"Tingting Gao","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gao, Tingting","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5153482786","display_name":"Han Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Han","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.5419999957084656,"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":0.5419999957084656,"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":0.41670000553131104,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.008700000122189522,"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/efficient-language-models","display_name":"efficient language models","score":0.8029999732971191},{"id":"https://openalex.org/keywords/long-range-context","display_name":"long-range context","score":0.5360000133514404},{"id":"https://openalex.org/keywords/key-value-cache","display_name":"key-value cache","score":0.6069999933242798},{"id":"https://openalex.org/keywords/computational-efficiency","display_name":"computational efficiency","score":0.4189999997615814}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8504999876022339},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.5756000280380249},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5029000043869019},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.49810001254081726},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.4941999912261963},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.46810001134872437},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.43799999356269836},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.40560001134872437},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3962000012397766},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.3874000012874603},{"id":"https://openalex.org/C183322885","wikidata":"https://www.wikidata.org/wiki/Q17007702","display_name":"Context model","level":3,"score":0.3864000141620636},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.37279999256134033},{"id":"https://openalex.org/C59201141","wikidata":"https://www.wikidata.org/wiki/Q46904","display_name":"GSM","level":2,"score":0.34299999475479126},{"id":"https://openalex.org/C2778751112","wikidata":"https://www.wikidata.org/wiki/Q835016","display_name":"Window (computing)","level":2,"score":0.3018999993801117},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.29260000586509705},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.28519999980926514},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.2809000015258789},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.28060001134872437},{"id":"https://openalex.org/C38556500","wikidata":"https://www.wikidata.org/wiki/Q13404475","display_name":"Cache algorithms","level":4,"score":0.27880001068115234},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.2734000086784363},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2718999981880188},{"id":"https://openalex.org/C37749351","wikidata":"https://www.wikidata.org/wiki/Q3181395","display_name":"Information processor","level":3,"score":0.2709999978542328},{"id":"https://openalex.org/C124304363","wikidata":"https://www.wikidata.org/wiki/Q673661","display_name":"Abstraction","level":2,"score":0.2703999876976013},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.26649999618530273},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.2563999891281128}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.33465","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33465","pdf_url":"https://arxiv.org/pdf/2609.33465","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,"listed_in":[],"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":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.33465","doi":"https://doi.org/10.48550/arxiv.2609.33465","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.33465","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,"listed_in":[],"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":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.33465","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.33465","pdf_url":"https://arxiv.org/pdf/2609.33465","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,"listed_in":[],"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":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214745185.pdf","grobid_xml":"https://content.openalex.org/works/W7214745185.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4415137623","https://openalex.org/W4415936639","https://openalex.org/W7164941922","https://openalex.org/W7092292508","https://openalex.org/W4200634402","https://openalex.org/W4404306933","https://openalex.org/W4403849929","https://openalex.org/W4415798938","https://openalex.org/W4416781483","https://openalex.org/W4394782111"],"abstract_inverted_index":{"Improving":[0],"language":[1,152],"model":[2,105],"efficiency":[3],"requires":[4,19],"not":[5,26],"only":[6],"cheaper":[7],"operations":[8],"but":[9,25],"also":[10,129],"eliminating":[11,143],"redundant":[12],"computation.":[13,54,101],"We":[14],"argue":[15],"that":[16,47,108,142],"effective":[17,148],"modeling":[18,153],"information":[20,50,60,156],"accessibility":[21,157],"and":[22,88,132,135],"computational":[23,159],"depth,":[24],"repeated":[27,80],"historical":[28,49,81],"retrieval":[29,64,126],"across":[30,103],"layers.":[31],"Guided":[32],"by":[33],"this":[34,77],"principle,":[35],"we":[36],"introduce":[37],"the":[38],"Global":[39],"State":[40],"Model":[41],"(GSM),":[42],"a":[43,66],"causal":[44],"encoder--decoder":[45],"architecture":[46],"separates":[48],"acquisition":[51],"from":[52],"deep":[53],"The":[55],"encoder":[56],"progressively":[57],"aggregates":[58],"long-range":[59],"through":[61],"multi-stage":[62],"sparse":[63],"into":[65],"fixed-size":[67],"shared":[68],"state,":[69],"while":[70,95,123],"decoder":[71,86],"layers":[72],"perform":[73],"successive":[74],"transformations":[75],"over":[76],"state":[78],"without":[79,154],"access.":[82],"This":[83],"design":[84],"makes":[85],"attention":[87],"cache":[89,137],"costs":[90],"independent":[91],"of":[92],"history":[93],"length,":[94],"dependency-aware":[96],"pruning":[97],"further":[98],"reduces":[99,136],"prefill":[100],"Experiments":[102],"multiple":[104],"scales":[106],"show":[107],"GSM":[109,128],"achieves":[110],"downstream":[111],"performance":[112],"comparable":[113],"to":[114],"matched":[115],"DeepSeek-V4.1-Flash-based":[116],"baselines,":[117],"with":[118],"improvements":[119],"at":[120],"some":[121],"scales,":[122],"preserving":[124],"long-context":[125],"capabilities.":[127],"improves":[130],"training":[131],"inference":[133],"throughput":[134],"overhead.":[138],"These":[139],"results":[140],"demonstrate":[141],"architectural":[144],"redundancy":[145],"offers":[146],"an":[147],"path":[149],"toward":[150],"efficient":[151],"sacrificing":[155],"or":[158],"depth.":[160]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-30T00:00:00"}