{"id":"https://openalex.org/W7214898070","doi":"https://doi.org/10.48550/arxiv.2609.37921","title":"Pattern Formation in Transformers","display_name":"Pattern Formation in Transformers","publication_year":2026,"publication_date":"2026-09-29","ids":{"openalex":"https://openalex.org/W7214898070","doi":"https://doi.org/10.48550/arxiv.2609.37921"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.37921","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37921","pdf_url":"https://arxiv.org/pdf/2609.37921","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.37921","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5120055969","display_name":"Erkan Turan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I142476485","display_name":"\u00c9cole Polytechnique","ror":"https://ror.org/05hy3tk52","country_code":"FR","type":"education","lineage":["https://openalex.org/I142476485","https://openalex.org/I4210145102"]},{"id":"https://openalex.org/I4210139461","display_name":"Laboratoire d'Informatique de l'\u00c9cole Polytechnique","ror":"https://ror.org/04afed728","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1294671590","https://openalex.org/I1326498283","https://openalex.org/I142476485","https://openalex.org/I4210139461","https://openalex.org/I4210145102","https://openalex.org/I4210159245"]},{"id":"https://openalex.org/I4210145102","display_name":"Institut Polytechnique de Paris","ror":"https://ror.org/042tfbd02","country_code":"FR","type":"education","lineage":["https://openalex.org/I4210145102"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Turan, Erkan","raw_affiliation_strings":["LIX , \u00c9cole Polytechnique , IP Paris , France"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"LIX , \u00c9cole Polytechnique , IP Paris , France","institution_ids":["https://openalex.org/I142476485","https://openalex.org/I4210139461","https://openalex.org/I4210145102"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119968971","display_name":"Gaspard Abel","orcid":"https://orcid.org/0000-0002-4893-5687","observed_orcids":["https://orcid.org/0000-0002-4893-5687"]},"institutions":[{"id":"https://openalex.org/I1294671590","display_name":"Centre National de la Recherche Scientifique","ror":"https://ror.org/02feahw73","country_code":"FR","type":"government","lineage":["https://openalex.org/I1294671590"]},{"id":"https://openalex.org/I277688954","display_name":"Universit\u00e9 Paris-Saclay","ror":"https://ror.org/03xjwb503","country_code":"FR","type":"education","lineage":["https://openalex.org/I277688954"]},{"id":"https://openalex.org/I4210100624","display_name":"Centre d'Analyse et de Math\u00e9matique Sociales","ror":"https://ror.org/00vb8fc53","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1294671590","https://openalex.org/I4210100624","https://openalex.org/I4210150854","https://openalex.org/I90669466"]},{"id":"https://openalex.org/I4403386768","display_name":"Centre Borelli","ror":"https://ror.org/02hcn4061","country_code":"FR","type":"facility","lineage":["https://openalex.org/I11559806","https://openalex.org/I1294671590","https://openalex.org/I204730241","https://openalex.org/I277688954","https://openalex.org/I4210129973","https://openalex.org/I4210136310","https://openalex.org/I4403386768"]},{"id":"https://openalex.org/I90669466","display_name":"\u00c9cole des hautes \u00e9tudes en sciences sociales","ror":"https://ror.org/02d9dg697","country_code":"FR","type":"facility","lineage":["https://openalex.org/I90669466"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Abel, Gaspard","raw_affiliation_strings":["Centre Borelli , ENS Paris-Saclay , Universit\u00e9 Paris-Saclay , France","Centre d'Analyse et de Math\u00e9matique Sociales , EHESS , CNRS , 75006 Paris , France"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Centre Borelli , ENS Paris-Saclay , Universit\u00e9 Paris-Saclay , France","institution_ids":["https://openalex.org/I277688954","https://openalex.org/I4403386768"]},{"raw_affiliation_string":"Centre d'Analyse et de Math\u00e9matique Sociales , EHESS , CNRS , 75006 Paris , France","institution_ids":["https://openalex.org/I1294671590","https://openalex.org/I4210100624","https://openalex.org/I90669466"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5072368950","display_name":"Maks Ovsjanikov","orcid":"https://orcid.org/0000-0002-5867-4046","observed_orcids":["https://orcid.org/0000-0002-5867-4046"]},"institutions":[{"id":"https://openalex.org/I142476485","display_name":"\u00c9cole Polytechnique","ror":"https://ror.org/05hy3tk52","country_code":"FR","type":"education","lineage":["https://openalex.org/I142476485","https://openalex.org/I4210145102"]},{"id":"https://openalex.org/I4210139461","display_name":"Laboratoire d'Informatique de l'\u00c9cole Polytechnique","ror":"https://ror.org/04afed728","country_code":"FR","type":"facility","lineage":["https://openalex.org/I1294671590","https://openalex.org/I1326498283","https://openalex.org/I142476485","https://openalex.org/I4210139461","https://openalex.org/I4210145102","https://openalex.org/I4210159245"]},{"id":"https://openalex.org/I4210145102","display_name":"Institut Polytechnique de Paris","ror":"https://ror.org/042tfbd02","country_code":"FR","type":"education","lineage":["https://openalex.org/I4210145102"]}],"countries":["FR"],"is_corresponding":false,"raw_author_name":"Ovsjanikov, Maks","raw_affiliation_strings":["LIX , \u00c9cole Polytechnique , IP Paris , France"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"LIX , \u00c9cole Polytechnique , IP Paris , France","institution_ids":["https://openalex.org/I142476485","https://openalex.org/I4210139461","https://openalex.org/I4210145102"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":8,"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":0.48510000109672546,"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":0.48510000109672546,"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/T10320","display_name":"Neural Networks and Applications","score":0.26339998841285706,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.02459999918937683,"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/transformer","display_name":"transformer models","score":0.7950000166893005},{"id":"https://openalex.org/keywords/pattern-formation","display_name":"pattern formation","score":0.8690000176429749},{"id":"https://openalex.org/keywords/inductive-bias","display_name":"inductive bias","score":0.8519999980926514},{"id":"https://openalex.org/keywords/rank-collapse","display_name":"rank collapse","score":0.7129999995231628},{"id":"https://openalex.org/keywords/self-attention","display_name":"self-attention","score":0.8579999804496765},{"id":"https://openalex.org/keywords/positional-encoding","display_name":"positional encoding","score":0.8669999837875366},{"id":"https://openalex.org/keywords/multi-head-attention","display_name":"multi-head attention","score":0.9700000286102295},{"id":"https://openalex.org/keywords/dynamical-systems","display_name":"dynamical systems","score":0.753000020980835},{"id":"https://openalex.org/keywords/traveling-waves","display_name":"traveling waves","score":0.9580000042915344},{"id":"https://openalex.org/keywords/rotating-waves","display_name":"rotating waves","score":0.9620000123977661},{"id":"https://openalex.org/keywords/weight-initialization","display_name":"weight initialization","score":0.9279999732971191},{"id":"https://openalex.org/keywords/data-efficiency","display_name":"data efficiency","score":0.8669999837875366},{"id":"https://openalex.org/keywords/cifar-10","display_name":"CIFAR-10","score":0.8050000071525574}],"concepts":[{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.7738000154495239},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4507000148296356},{"id":"https://openalex.org/C184720557","wikidata":"https://www.wikidata.org/wiki/Q7825049","display_name":"Topology (electrical circuits)","level":2,"score":0.4009999930858612},{"id":"https://openalex.org/C30403606","wikidata":"https://www.wikidata.org/wiki/Q2981904","display_name":"Electromagnetic coil","level":2,"score":0.3946000039577484},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.3896999955177307},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.3578000068664551},{"id":"https://openalex.org/C79379906","wikidata":"https://www.wikidata.org/wiki/Q3174497","display_name":"Dynamical systems theory","level":2,"score":0.296099990606308},{"id":"https://openalex.org/C47446073","wikidata":"https://www.wikidata.org/wiki/Q5165890","display_name":"Control theory (sociology)","level":3,"score":0.29600000381469727},{"id":"https://openalex.org/C24326235","wikidata":"https://www.wikidata.org/wiki/Q126095","display_name":"Electronic engineering","level":1,"score":0.2930000126361847},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.29260000586509705},{"id":"https://openalex.org/C197352929","wikidata":"https://www.wikidata.org/wiki/Q1074074","display_name":"Inductive bias","level":4,"score":0.27379998564720154},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.27230000495910645},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2680000066757202},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.26170000433921814},{"id":"https://openalex.org/C117323899","wikidata":"https://www.wikidata.org/wiki/Q1907989","display_name":"Distribution transformer","level":4,"score":0.2515999972820282}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.37921","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37921","pdf_url":"https://arxiv.org/pdf/2609.37921","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.2609.37921","doi":"https://doi.org/10.48550/arxiv.2609.37921","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.37921","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":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.37921","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37921","pdf_url":"https://arxiv.org/pdf/2609.37921","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":[{"id":"https://openalex.org/F4320323641","display_name":"Vedeck\u00e1 Grantov\u00e1 Agent\u00fara M\u0160VVa\u0160 SR a SAV","ror":"https://ror.org/044gwpv05"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214898070.pdf","grobid_xml":"https://content.openalex.org/works/W7214898070.grobid-xml"},"referenced_works_count":3,"referenced_works":["https://openalex.org/W2001230033","https://openalex.org/W7128041617","https://openalex.org/W7158591772"],"related_works":["https://openalex.org/W7133245089","https://openalex.org/W4407759049","https://openalex.org/W7131389392","https://openalex.org/W7130696399","https://openalex.org/W4417024991","https://openalex.org/W4407185543","https://openalex.org/W7156752337","https://openalex.org/W4376123079","https://openalex.org/W4415317857","https://openalex.org/W4415798446"],"abstract_inverted_index":{"What":[0],"are":[1],"the":[2,13,55,86,131],"inductive":[3,110],"biases":[4],"of":[5,89,118,133],"a":[6,63,68,104,115,158],"Transformer":[7,70,106],"architecture?":[8],"Existing":[9],"theory":[10],"on":[11,47,182,189],"how":[12,74],"forward":[14],"pass":[15],"shapes":[16],"representations":[17],"either":[18],"considers":[19],"whether":[20],"Transformers":[21,90],"escape":[22],"from":[23,39],"rank":[24,72],"collapse":[25],"or":[26,124,147],"demonstrates":[27],"that":[28,85,103,152,162],"self-attention":[29],"drives":[30],"tokens":[31],"toward":[32],"cluster":[33],"patterns.":[34],"The":[35],"latter":[36],"view":[37,88],"arises":[38],"an":[40,109],"elegant":[41],"dynamical":[42,87,160],"systems":[43],"perspective,":[44],"but":[45],"relies":[46],"simplified":[48],"architectural":[49,135],"assumptions,":[50],"and":[51,98,171,179,186],"does":[52,75],"not":[53],"explain":[54],"rich":[56,116],"structures":[57,154],"observed":[58],"in":[59,137],"practice.":[60],"This":[61],"leaves":[62],"major":[64],"open":[65],"question:":[66],"when":[67],"full":[69,105],"escapes":[71],"collapse,":[73],"it":[76],"structure":[77],"token":[78],"representations?":[79],"Using":[80],"pattern-formation":[81],"theory,":[82],"we":[83,150,174],"show":[84,151],"can":[91,155],"account":[92],"for":[93],"Positional":[94],"Encoding,":[95],"Multi-Head":[96],"Attention,":[97],"Output-Value":[99],"geometry.":[100],"We":[101,129],"demonstrate":[102,175],"architecture":[107],"imposes":[108],"prior":[111,161],"by":[112],"selectively":[113],"amplifying":[114],"set":[117],"previously":[119],"unreported":[120],"patterns,":[121],"including":[122],"traveling":[123],"rotating":[125],"waves":[126],"among":[127],"others.":[128],"characterize":[130],"role":[132],"each":[134],"component":[136],"controlling":[138],"which":[139,143],"pattern":[140],"is":[141],"amplified,":[142],"ones":[144],"stabilize,":[145],"compete,":[146],"coexist.":[148],"Finally,":[149],"these":[153],"act":[156],"as":[157],"controllable":[159],"facilitates":[163],"learning.":[164],"By":[165],"choosing":[166],"both":[167],"task-aligned":[168],"positional":[169],"encoding":[170],"weight":[172],"initialization,":[173],"improved":[176],"data":[177],"efficiency":[178],"accelerated":[180],"optimization":[181],"controlled":[183],"sequence":[184],"tasks":[185],"with":[187],"ConViT":[188],"CIFAR-10.":[190]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-01T00:00:00"}
