{"id":"https://openalex.org/W7143583446","doi":null,"title":"Dynamical Transformer: Attentional Selective State Space Model","display_name":"Dynamical Transformer: Attentional Selective State Space Model","publication_year":2026,"publication_date":"2026-03-23","ids":{"openalex":"https://openalex.org/W7143583446"},"language":"en","primary_location":{"id":"pmh:oai:HAL:hal-05568648v1","doi":null,"is_oa":true,"landing_page_url":"https://hal.science/hal-05568648","pdf_url":"https://hal.science/hal-05568648/document","source":{"id":"https://openalex.org/S4306402512","display_name":"HAL","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I4405252976","host_organization_name":"Centre pour la Communication Scientifique Directe","host_organization_lineage":["https://openalex.org/I4405252976"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2026","raw_type":"info:eu-repo/semantics/preprint"},"parent_works":[],"lineage":["https://openalex.org/W7143583446"],"child_works_count":0,"type":"preprint","is_traditional":true,"indexed_in":[],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://hal.science/hal-05568648/document","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5114415465","display_name":"Yannis Bendi-Ouis","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405260129","display_name":"MNEMOSYNE: Mnemonic Synergy","ror":"https://ror.org/039zcz976","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1326498283","https://openalex.org/I4210131512","https://openalex.org/I4405260129"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Bendi-Ouis, Yannis","raw_affiliation_strings":["Mnemosyne - Mnemonic Synergy (200, avenue de la Vieille Tour, 33405 Talence cedex / B\u00e2timent Neurocampus \u2013 146 rue L\u00e9o Saignat \u2013 CS 61292 33076 Bordeaux cedex - France)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Mnemosyne - Mnemonic Synergy (200, avenue de la Vieille Tour, 33405 Talence cedex / B\u00e2timent Neurocampus \u2013 146 rue L\u00e9o Saignat \u2013 CS 61292 33076 Bordeaux cedex - France)","institution_ids":["https://openalex.org/I4405260129"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5051036759","display_name":"Xavier Hinaut","orcid":"https://orcid.org/0000-0002-1924-1184","observed_orcids":["https://orcid.org/0000-0002-1924-1184"]},"institutions":[{"id":"https://openalex.org/I4405260129","display_name":"MNEMOSYNE: Mnemonic Synergy","ror":"https://ror.org/039zcz976","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1326498283","https://openalex.org/I4210131512","https://openalex.org/I4405260129"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Hinaut, Xavier","raw_affiliation_strings":["Mnemosyne - Mnemonic Synergy (200, avenue de la Vieille Tour, 33405 Talence cedex / B\u00e2timent Neurocampus \u2013 146 rue L\u00e9o Saignat \u2013 CS 61292 33076 Bordeaux cedex - France)"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Mnemosyne - Mnemonic Synergy (200, avenue de la Vieille Tour, 33405 Talence cedex / B\u00e2timent Neurocampus \u2013 146 rue L\u00e9o Saignat \u2013 CS 61292 33076 Bordeaux cedex - France)","institution_ids":["https://openalex.org/I4405260129"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I4405260129"],"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/T12611","display_name":"Neural Networks and Reservoir Computing","score":0.7738999724388123,"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":0.7738999724388123,"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/T10028","display_name":"Topic Modeling","score":0.09759999811649323,"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":0.01080000028014183,"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/selective-state-space-models","display_name":"selective state space models","score":0.8270000219345093},{"id":"https://openalex.org/keywords/computational-cost-reduction","display_name":"computational complexity reduction","score":0.5400000214576721},{"id":"https://openalex.org/keywords/matrix-diagonalization","display_name":"diagonalization","score":0.734000027179718},{"id":"https://openalex.org/keywords/recurrent-neural-networks","display_name":"recurrent neural networks","score":0.5379999876022339},{"id":"https://openalex.org/keywords/sequence-modeling","display_name":"sequence modeling","score":0.3490000069141388}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6776999831199646},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.616100013256073},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.5999000072479248},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.4927999973297119},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.48500001430511475},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4702000021934509},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4456000030040741},{"id":"https://openalex.org/C53442348","wikidata":"https://www.wikidata.org/wiki/Q745101","display_name":"Content-addressable memory","level":3,"score":0.4410000145435333},{"id":"https://openalex.org/C2778618852","wikidata":"https://www.wikidata.org/wiki/Q1128613","display_name":"Content-addressable storage","level":4,"score":0.4120999872684479},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.39430001378059387},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.3856000006198883},{"id":"https://openalex.org/C135796866","wikidata":"https://www.wikidata.org/wiki/Q7315328","display_name":"Reservoir computing","level":4,"score":0.38190001249313354},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.38179999589920044},{"id":"https://openalex.org/C147168706","wikidata":"https://www.wikidata.org/wiki/Q1457734","display_name":"Recurrent neural network","level":3,"score":0.37599998712539673},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.34310001134872437},{"id":"https://openalex.org/C12186640","wikidata":"https://www.wikidata.org/wiki/Q6815743","display_name":"Memory model","level":3,"score":0.3407999873161316},{"id":"https://openalex.org/C148047603","wikidata":"https://www.wikidata.org/wiki/Q1014612","display_name":"Parallelizable manifold","level":2,"score":0.33390000462532043},{"id":"https://openalex.org/C129844170","wikidata":"https://www.wikidata.org/wiki/Q41299","display_name":"Quadratic equation","level":2,"score":0.3334999978542328},{"id":"https://openalex.org/C72434380","wikidata":"https://www.wikidata.org/wiki/Q230930","display_name":"State space","level":2,"score":0.33309999108314514},{"id":"https://openalex.org/C3832189","wikidata":"https://www.wikidata.org/wiki/Q8588916","display_name":"Models of neural computation","level":3,"score":0.3296000063419342},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.32580000162124634},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.3018999993801117},{"id":"https://openalex.org/C82687282","wikidata":"https://www.wikidata.org/wiki/Q66221","display_name":"Auxiliary memory","level":2,"score":0.29989999532699585},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.29919999837875366},{"id":"https://openalex.org/C159423971","wikidata":"https://www.wikidata.org/wiki/Q177251","display_name":"Associative property","level":2,"score":0.29580000042915344},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.29440000653266907},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.2930999994277954},{"id":"https://openalex.org/C79470037","wikidata":"https://www.wikidata.org/wiki/Q279748","display_name":"Out-of-core algorithm","level":2,"score":0.28790000081062317},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.28369998931884766},{"id":"https://openalex.org/C167822520","wikidata":"https://www.wikidata.org/wiki/Q176452","display_name":"Finite-state machine","level":2,"score":0.2736000120639801},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.2711000144481659},{"id":"https://openalex.org/C30390489","wikidata":"https://www.wikidata.org/wiki/Q4680748","display_name":"Adaptive memory","level":3,"score":0.26669999957084656},{"id":"https://openalex.org/C77246614","wikidata":"https://www.wikidata.org/wiki/Q1409400","display_name":"Gramian matrix","level":3,"score":0.2621999979019165},{"id":"https://openalex.org/C2781357197","wikidata":"https://www.wikidata.org/wiki/Q5757597","display_name":"High memory","level":2,"score":0.2590999901294708},{"id":"https://openalex.org/C187834632","wikidata":"https://www.wikidata.org/wiki/Q188804","display_name":"Factorization","level":2,"score":0.25859999656677246}],"mesh":[],"locations_count":1,"locations":[{"id":"pmh:oai:HAL:hal-05568648v1","doi":null,"is_oa":true,"landing_page_url":"https://hal.science/hal-05568648","pdf_url":"https://hal.science/hal-05568648/document","source":{"id":"https://openalex.org/S4306402512","display_name":"HAL","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I4405252976","host_organization_name":"Centre pour la Communication Scientifique Directe","host_organization_lineage":["https://openalex.org/I4405252976"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2026","raw_type":"info:eu-repo/semantics/preprint"}],"best_oa_location":{"id":"pmh:oai:HAL:hal-05568648v1","doi":null,"is_oa":true,"landing_page_url":"https://hal.science/hal-05568648","pdf_url":"https://hal.science/hal-05568648/document","source":{"id":"https://openalex.org/S4306402512","display_name":"HAL","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I4405252976","host_organization_name":"Centre pour la Communication Scientifique Directe","host_organization_lineage":["https://openalex.org/I4405252976"],"host_organization_lineage_names":[],"type":"repository"},"license":"other-oa","license_id":"https://openalex.org/licenses/other-oa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"2026","raw_type":"info:eu-repo/semantics/preprint"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7143583446.pdf","grobid_xml":"https://content.openalex.org/works/W7143583446.grobid-xml"},"referenced_works_count":2,"referenced_works":["https://openalex.org/W4414625730","https://openalex.org/W7131390409"],"related_works":["https://openalex.org/W4414625730","https://openalex.org/W4404783458","https://openalex.org/W7168027269","https://openalex.org/W7119188126","https://openalex.org/W4416372109","https://openalex.org/W4389326242","https://openalex.org/W7214525449","https://openalex.org/W7162686326","https://openalex.org/W3093260394","https://openalex.org/W7155044684"],"abstract_inverted_index":{"The":[0,90],"standard":[1,100],"Transformer":[2,160],"architecture":[3],"has":[4],"revolutionized":[5],"sequence":[6,28,107,140,165,204],"modeling,":[7],"yet":[8],"it":[9],"is":[10,97],"fundamentally":[11],"limited":[12],"by":[13,49,65],"its":[14,71,227,230],"quadratic":[15],"computational":[16],"complexity":[17,201],"and":[18,74,168,189,246,255],"infinite":[19],"memory":[20,96,117,133,145,220],"capacity":[21,41],"requirement,":[22],"which":[23,35],"scales":[24],"proportionally":[25],"with":[26,77,175],"the":[27,78,99,105,122,130,139,144,152,164,222,243,250,256],"length.":[29,141],"This":[30],"contradicts":[31],"biological":[32],"working":[33],"memory,":[34],"compresses":[36],"past":[37],"information":[38],"into":[39],"finite":[40],"representations.":[42],"We":[43,210,241],"propose":[44],"to":[45,224],"address":[46],"this":[47,94],"bottleneck":[48],"introducing":[50],"a":[51,86,159,170,176,190,203,235],"Memory":[52],"Component":[53],"that":[54,98],"integrates":[55],"several":[56],"dynamical":[57],"systems":[58],"implemented":[59],"as":[60,129,158,169],"linear":[61],"reservoir":[62],"units":[63,134],"inspired":[64],"Reservoir":[66],"Computing.":[67],"Each":[68],"unit":[69,221],"possesses":[70],"own":[72],"learned":[73],"adaptive":[75,214],"dynamics,":[76],"objective":[79],"of":[80,108,132,138],"selectively":[81],"storing":[82],"previous":[83],"inputs":[84],"in":[85],"compact":[87],"internal":[88],"state.":[89],"core":[91],"intuition":[92],"behind":[93],"multi-state-vector":[95],"attention":[101,123],"mechanism":[102],"applied":[103,113],"across":[104,114,163],"temporal":[106],"tokens":[109],"can":[110,154],"instead":[111],"be":[112,155],"these":[115],"parallel":[116],"state-vectors.":[118],"By":[119,181],"doing":[120],"so,":[121],"computation":[124],"becomes":[125],"constant":[126],"per":[127],"timestep,":[128],"number":[131],"remains":[135],"fixed":[136],"regardless":[137],"Furthermore,":[142],"because":[143],"implementation":[146],"relies":[147],"strictly":[148],"on":[149,249],"linearized":[150],"reservoirs,":[151],"model":[153,196],"interpreted":[156],"simultaneously":[157],"(fully":[161],"parallelizable":[162],"during":[166,179],"training)":[167],"Recurrent":[171],"Neural":[172],"Network":[173],"(operating":[174],"fixed-size":[177],"state":[178],"inference).":[180],"leveraging":[182],"matrix":[183],"diagonalization,":[184],"an":[185,213],"associative":[186],"scan":[187],"algorithm,":[188],"custom":[191],"Triton":[192],"CUDA":[193],"kernel,":[194],"our":[195],"yields":[197],"O(log":[198],"T)":[199],"training":[200],"over":[202],"while":[205],"fully":[206],"exploiting":[207],"high-performance":[208],"GPUs.":[209],"additionally":[211],"introduce":[212],"leak":[215],"rate":[216],"mechanism,":[217],"granting":[218],"each":[219],"ability":[223],"dynamically":[225],"modulates":[226],"retention":[228],"around":[229],"preferred":[231],"timescale,":[232],"effectively":[233],"instantiating":[234],"Selective":[236],"State":[237],"Space":[238],"Model":[239],"(SSM).":[240],"confirm":[242],"architecture's":[244],"scalability":[245],"modeling":[247],"capabilities":[248],"Shakespeare":[251],"dataset":[252,258],"(1M":[253],"tokens)":[254],"OpenWebText":[257],"(9B":[259],"tokens).":[260]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-03-31T00:00:00"}