{"id":"https://openalex.org/W7160930630","doi":"https://doi.org/10.48550/arxiv.2605.10931","title":"Quantifying Concentration Phenomena of Mean-Field Transformers in the Low-Temperature Regime","display_name":"Quantifying Concentration Phenomena of Mean-Field Transformers in the Low-Temperature Regime","publication_year":2026,"publication_date":"2026-05-11","ids":{"openalex":"https://openalex.org/W7160930630","doi":"https://doi.org/10.48550/arxiv.2605.10931"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2605.10931","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.10931","pdf_url":"https://arxiv.org/pdf/2605.10931","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","license_id":"https://openalex.org/licenses/cc-by","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/2605.10931","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100030234","display_name":"Albert Alcalde","orcid":"https://orcid.org/0009-0000-5147-0406","observed_orcids":["https://orcid.org/0009-0000-5147-0406"]},"institutions":[{"id":"https://openalex.org/I181369854","display_name":"Friedrich-Alexander-Universit\u00e4t Erlangen-N\u00fcrnberg","ror":"https://ror.org/00f7hpc57","country_code":"DE","type":"education","lineage":["https://openalex.org/I181369854"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Alcalde, Albert","raw_affiliation_strings":["Department of Mathematics Friedrich-Alexander-University Erlangen-N\u00fcrnberg"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Mathematics Friedrich-Alexander-University Erlangen-N\u00fcrnberg","institution_ids":["https://openalex.org/I181369854"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035045595","display_name":"Leon Bungert","orcid":"https://orcid.org/0000-0002-6554-9892","observed_orcids":["https://orcid.org/0000-0002-6554-9892"]},"institutions":[{"id":"https://openalex.org/I25974101","display_name":"University of W\u00fcrzburg","ror":"https://ror.org/00fbnyb24","country_code":"DE","type":"education","lineage":["https://openalex.org/I25974101"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Bungert, Leon","raw_affiliation_strings":["Institute of Mathematics CAIDAS University of W\u00fcrzburg"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Mathematics CAIDAS University of W\u00fcrzburg","institution_ids":["https://openalex.org/I25974101"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088292594","display_name":"Konstantin Riedl","orcid":"https://orcid.org/0000-0002-2206-4334","observed_orcids":["https://orcid.org/0000-0002-2206-4334"]},"institutions":[{"id":"https://openalex.org/I40120149","display_name":"University of Oxford","ror":"https://ror.org/052gg0110","country_code":"GB","type":"education","lineage":["https://openalex.org/I40120149"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Riedl, Konstantin","raw_affiliation_strings":["Mathematical Institute University of Oxford Reuben College"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Mathematical Institute University of Oxford Reuben College","institution_ids":["https://openalex.org/I40120149"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5051931414","display_name":"Tim Roith","orcid":"https://orcid.org/0000-0001-8440-2928","observed_orcids":["https://orcid.org/0000-0001-8440-2928"]},"institutions":[{"id":"https://openalex.org/I62916508","display_name":"Technical University of Munich","ror":"https://ror.org/02kkvpp62","country_code":"DE","type":"education","lineage":["https://openalex.org/I62916508"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Roith, Tim","raw_affiliation_strings":["CIT School Technical University of Munich Munich","Center for Machine Learning"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CIT School Technical University of Munich Munich","institution_ids":["https://openalex.org/I62916508"]},{"raw_affiliation_string":"Center for Machine Learning","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":7,"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":0.25929999351501465,"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":0.25929999351501465,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.17550000548362732,"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":0.09449999779462814,"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/self-attention","display_name":"self-attention","score":0.9089999794960022},{"id":"https://openalex.org/keywords/low-temperature-limit","display_name":"low-temperature limit","score":0.6349999904632568},{"id":"https://openalex.org/keywords/wasserstein-distance","display_name":"Wasserstein distance","score":0.6769999861717224},{"id":"https://openalex.org/keywords/metastability","display_name":"metastability","score":0.8569999933242798},{"id":"https://openalex.org/keywords/encoder-only-transformer","display_name":"encoder-only transformer","score":0.7799999713897705},{"id":"https://openalex.org/keywords/interacting-particle-systems","display_name":"interacting particle systems","score":0.5559999942779541}],"concepts":[{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5099999904632568},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.483599990606308},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.44190001487731934},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.43560001254081726},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.4255000054836273},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.4239000082015991},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.3968000113964081},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.37779998779296875},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.34619998931884766},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.34439998865127563},{"id":"https://openalex.org/C97937538","wikidata":"https://www.wikidata.org/wiki/Q199691","display_name":"Laplace transform","level":2,"score":0.34439998865127563},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3231000006198883},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.2930999994277954},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.28929999470710754},{"id":"https://openalex.org/C73586568","wikidata":"https://www.wikidata.org/wiki/Q2600211","display_name":"Parameter space","level":2,"score":0.2847999930381775},{"id":"https://openalex.org/C57493831","wikidata":"https://www.wikidata.org/wiki/Q3134666","display_name":"Projection (relational algebra)","level":2,"score":0.2822999954223633},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.2809999883174896},{"id":"https://openalex.org/C8272713","wikidata":"https://www.wikidata.org/wiki/Q176737","display_name":"Stochastic process","level":2,"score":0.27309998869895935},{"id":"https://openalex.org/C17829176","wikidata":"https://www.wikidata.org/wiki/Q179043","display_name":"Electrical impedance","level":2,"score":0.26930001378059387},{"id":"https://openalex.org/C183057437","wikidata":"https://www.wikidata.org/wiki/Q671617","display_name":"Laplace distribution","level":3,"score":0.26739999651908875},{"id":"https://openalex.org/C166785042","wikidata":"https://www.wikidata.org/wiki/Q190391","display_name":"Central limit theorem","level":2,"score":0.26249998807907104},{"id":"https://openalex.org/C112313634","wikidata":"https://www.wikidata.org/wiki/Q7886648","display_name":"Complement (music)","level":5,"score":0.2621999979019165},{"id":"https://openalex.org/C62799726","wikidata":"https://www.wikidata.org/wiki/Q190056","display_name":"Hilbert space","level":2,"score":0.26170000433921814},{"id":"https://openalex.org/C176321772","wikidata":"https://www.wikidata.org/wiki/Q1430640","display_name":"Numerical stability","level":3,"score":0.25679999589920044}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2605.10931","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.10931","pdf_url":"https://arxiv.org/pdf/2605.10931","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","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2605.10931","doi":"https://doi.org/10.48550/arxiv.2605.10931","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.10931","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","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2605.10931","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2605.10931","pdf_url":"https://arxiv.org/pdf/2605.10931","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","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G1289050001","display_name":null,"funder_award_id":"DFG-SPP 2298","funder_id":"https://openalex.org/F4320320879","funder_display_name":"Deutsche Forschungsgemeinschaft"},{"id":"https://openalex.org/G1511613426","display_name":null,"funder_award_id":"544579844","funder_id":"https://openalex.org/F4320320879","funder_display_name":"Deutsche Forschungsgemeinschaft"},{"id":"https://openalex.org/G2972077997","display_name":null,"funder_award_id":"101073558","funder_id":"https://openalex.org/F4320334322","funder_display_name":"HORIZON EUROPE Framework Programme"},{"id":"https://openalex.org/G442688124","display_name":null,"funder_award_id":"SPP2298","funder_id":"https://openalex.org/F4320320879","funder_display_name":"Deutsche Forschungsgemeinschaft"},{"id":"https://openalex.org/G5350531418","display_name":null,"funder_award_id":"101073558","funder_id":"https://openalex.org/F4320320300","funder_display_name":"European Commission"},{"id":"https://openalex.org/G6056722365","display_name":"DMS-EPSRC: Asymptotic Analysis of Online Training Algorithms in Machine Learning: Recurrent, Graphical, and Deep Neural Networks","funder_award_id":"2311500","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320320300","display_name":"European Commission","ror":"https://ror.org/00k4n6c32"},{"id":"https://openalex.org/F4320320879","display_name":"Deutsche Forschungsgemeinschaft","ror":"https://ror.org/018mejw64"},{"id":"https://openalex.org/F4320334322","display_name":"HORIZON EUROPE Framework Programme","ror":null},{"id":"https://openalex.org/F4320334627","display_name":"Engineering and Physical Sciences Research Council","ror":"https://ror.org/0439y7842"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7160930630.pdf","grobid_xml":"https://content.openalex.org/works/W7160930630.grobid-xml"},"referenced_works_count":19,"referenced_works":["https://openalex.org/W3195577433","https://openalex.org/W4322718191","https://openalex.org/W4390041933","https://openalex.org/W4401306886","https://openalex.org/W4403345493","https://openalex.org/W4405057535","https://openalex.org/W4407012314","https://openalex.org/W4414977149","https://openalex.org/W4415329210","https://openalex.org/W4416245076","https://openalex.org/W4416968619","https://openalex.org/W4417025187","https://openalex.org/W4417521252","https://openalex.org/W6948287557","https://openalex.org/W7101688532","https://openalex.org/W7126250235","https://openalex.org/W7134280024","https://openalex.org/W7148748946","https://openalex.org/W7158837292"],"related_works":["https://openalex.org/W4415337592","https://openalex.org/W6999270437","https://openalex.org/W4416968619","https://openalex.org/W7106580379","https://openalex.org/W7204436530","https://openalex.org/W4376123079","https://openalex.org/W7222093125","https://openalex.org/W7161074523","https://openalex.org/W4417521252","https://openalex.org/W7160829335"],"abstract_inverted_index":{"Transformers":[0],"with":[1,58,148],"self-attention":[2],"modules":[3],"as":[4,137],"their":[5],"core":[6],"components":[7],"have":[8],"become":[9],"an":[10],"integral":[11],"architecture":[12],"in":[13,29,39,110,144],"modern":[14],"large":[15,206],"language":[16],"and":[17,87,90,118,139,205],"foundation":[18],"models.":[19],"In":[20],"this":[21,192],"paper,":[22],"we":[23,63,97,126],"study":[24],"the":[25,40,51,66,72,75,84,100,104,113,124,131,155,171,180,185,208,217,220],"evolution":[26],"of":[27,54,74,103,112,165,219],"tokens":[28],"deep":[30],"encoder-only":[31],"transformers":[32],"at":[33,184],"inference":[34,119],"time":[35,120,163],"which":[36,168],"is":[37],"described":[38],"large-token":[41],"limit":[42,136],"by":[43,83,199,216],"a":[44,79,141,149,211],"mean-field":[45],"continuity":[46],"equation.":[47],"Leveraging":[48],"ideas":[49],"from":[50],"convergence":[52],"analysis":[53],"interacting":[55],"multi-particle":[56],"systems,":[57],"particles":[59],"corresponding":[60],"to":[61,153,170],"tokens,":[62],"prove":[64],"that":[65,99,161,201],"token":[67,181],"distribution":[68,77,182],"rapidly":[69],"concentrates":[70,183],"onto":[71],"push-forward":[73],"initial":[76],"under":[78],"projection":[80],"map":[81],"induced":[82],"key,":[85],"query,":[86],"value":[88,221],"matrices,":[89],"remains":[91],"metastable":[92],"for":[93,130,162,176,202],"moderate":[94],"times.":[95],"Specifically,":[96],"show":[98],"Wasserstein":[101,145],"distance":[102],"two":[105,156],"distributions":[106],"scales":[107,164],"like":[108],"$\\sqrt{{\\log(\u03b2+1)}/\u03b2}\\exp(Ct)+\\exp(-ct)$":[109],"terms":[111],"temperature":[114],"parameter":[115],"$\u03b2^{-1}\\to":[116],"0$":[117],"$t\\geq":[121],"0$.":[122],"For":[123],"proof,":[125],"establish":[127],"Lyapunov-type":[128],"estimates":[129],"zero-temperature":[132],"equation,":[133],"identify":[134],"its":[135],"$t\\to\\infty$,":[138],"employ":[140],"stability":[142],"estimate":[143],"space":[146],"together":[147],"quantitative":[150],"Laplace":[151],"principle":[152],"couple":[154],"equations.":[157],"Our":[158],"result":[159],"implies":[160],"order":[166],"$\\log\u03b2$,":[167],"correspond":[169],"early-phase":[172],"behavior":[173],"most":[174],"relevant":[175],"practical":[177],"finite-depth":[178],"models,":[179],"identified":[186],"limiting":[187],"distribution.":[188],"Numerical":[189],"experiments":[190],"confirm":[191],"and,":[193],"beyond":[194],"that,":[195],"complement":[196],"our":[197],"theory":[198],"showing":[200],"finite":[203],"$\u03b2$":[204],"$t$":[207],"dynamics":[209],"enter":[210],"different":[212],"terminal":[213],"phase,":[214],"dominated":[215],"spectrum":[218],"matrix.":[222]},"counts_by_year":[{"year":2026,"cited_by_count":6},{"year":2025,"cited_by_count":1}],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2026-05-13T00:00:00"}