{"id":"https://openalex.org/W7143480655","doi":"https://doi.org/10.48550/arxiv.2603.26554","title":"Sharp Capacity Scaling of Spectral Optimizers in Learning Associative Memory","display_name":"Sharp Capacity Scaling of Spectral Optimizers in Learning Associative Memory","publication_year":2026,"publication_date":"2026-03-27","ids":{"openalex":"https://openalex.org/W7143480655","doi":"https://doi.org/10.48550/arxiv.2603.26554"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.26554","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.26554","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.26554","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5130981999","display_name":"Juno Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kim, Juno","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5004602170","display_name":"Eshaan Nichani","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nichani, Eshaan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047033147","display_name":"Denny Wu","orcid":"https://orcid.org/0000-0002-8448-9610"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Denny","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088570356","display_name":"Alberto Bietti","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bietti, Alberto","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5130957467","display_name":"Jason Sang Hun Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Jason D.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.10180000215768814,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.10180000215768814,"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/T14347","display_name":"Big Data and Digital Economy","score":0.0729999989271164,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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":0.0575999990105629,"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/scaling","display_name":"Scaling","score":0.6496999859809875},{"id":"https://openalex.org/keywords/muon","display_name":"Muon","score":0.5916000008583069},{"id":"https://openalex.org/keywords/limit","display_name":"Limit (mathematics)","score":0.5785999894142151},{"id":"https://openalex.org/keywords/content-addressable-memory","display_name":"Content-addressable memory","score":0.5127000212669373},{"id":"https://openalex.org/keywords/gaussian","display_name":"Gaussian","score":0.4717999994754791},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.4537999927997589},{"id":"https://openalex.org/keywords/scaling-limit","display_name":"Scaling limit","score":0.4065000116825104},{"id":"https://openalex.org/keywords/block","display_name":"Block (permutation group theory)","score":0.3856000006198883},{"id":"https://openalex.org/keywords/associative-property","display_name":"Associative property","score":0.38100001215934753}],"concepts":[{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.6496999859809875},{"id":"https://openalex.org/C205334942","wikidata":"https://www.wikidata.org/wiki/Q3151","display_name":"Muon","level":2,"score":0.5916000008583069},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5861999988555908},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.5785999894142151},{"id":"https://openalex.org/C53442348","wikidata":"https://www.wikidata.org/wiki/Q745101","display_name":"Content-addressable memory","level":3,"score":0.5127000212669373},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.4717999994754791},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.4537999927997589},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.44179999828338623},{"id":"https://openalex.org/C24084028","wikidata":"https://www.wikidata.org/wiki/Q7429781","display_name":"Scaling limit","level":3,"score":0.4065000116825104},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.39079999923706055},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.3856000006198883},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3828999996185303},{"id":"https://openalex.org/C159423971","wikidata":"https://www.wikidata.org/wiki/Q177251","display_name":"Associative property","level":2,"score":0.38100001215934753},{"id":"https://openalex.org/C2780388253","wikidata":"https://www.wikidata.org/wiki/Q5421508","display_name":"Exponent","level":2,"score":0.36579999327659607},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.36489999294281006},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.35190001130104065},{"id":"https://openalex.org/C163258240","wikidata":"https://www.wikidata.org/wiki/Q25342","display_name":"Power (physics)","level":2,"score":0.33880001306533813},{"id":"https://openalex.org/C168110828","wikidata":"https://www.wikidata.org/wiki/Q1331626","display_name":"Spectral density","level":2,"score":0.3257000148296356},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3224000036716461},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.3156999945640564},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3027999997138977},{"id":"https://openalex.org/C61224824","wikidata":"https://www.wikidata.org/wiki/Q2260434","display_name":"Mixture model","level":2,"score":0.29789999127388},{"id":"https://openalex.org/C61326573","wikidata":"https://www.wikidata.org/wiki/Q1496376","display_name":"Gaussian process","level":3,"score":0.29440000653266907},{"id":"https://openalex.org/C2988430800","wikidata":"https://www.wikidata.org/wiki/Q428971","display_name":"Scaling law","level":3,"score":0.2930000126361847},{"id":"https://openalex.org/C87040749","wikidata":"https://www.wikidata.org/wiki/Q428971","display_name":"Power law","level":2,"score":0.2863999903202057},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.2696000039577484},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.26899999380111694},{"id":"https://openalex.org/C166785042","wikidata":"https://www.wikidata.org/wiki/Q190391","display_name":"Central limit theorem","level":2,"score":0.257999986410141},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.2563999891281128},{"id":"https://openalex.org/C83546350","wikidata":"https://www.wikidata.org/wiki/Q1139051","display_name":"Regression","level":2,"score":0.25540000200271606}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.26554","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.26554","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.26554","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.26554","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16","score":0.5862152576446533}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Spectral":[0],"optimizers":[1],"such":[2],"as":[3],"Muon":[4,107,124,146],"have":[5],"recently":[6],"shown":[7],"strong":[8],"empirical":[9],"performance":[10],"in":[11,43],"large-scale":[12],"language":[13,201],"model":[14,39],"training,":[15],"but":[16],"the":[17,32,61,69,77,90,103,135,162,173,184,192],"source":[18],"and":[19,53,57,86,113,143,190,204],"extent":[20],"of":[21,63,80,83,106,111,183,187],"their":[22],"advantage":[23],"remain":[24],"poorly":[25],"understood.":[26],"We":[27,100,132],"study":[28],"this":[29],"question":[30],"through":[31],"linear":[33],"associative":[34],"memory":[35],"problem,":[36],"a":[37,95,127,139,148,180],"tractable":[38],"for":[40,194],"factual":[41],"recall":[42],"transformer-based":[44],"models.":[45],"In":[46],"particular,":[47],"we":[48],"go":[49],"beyond":[50],"orthogonal":[51],"embeddings":[52],"consider":[54],"Gaussian":[55],"inputs":[56],"outputs,":[58],"which":[59],"allows":[60],"number":[62],"stored":[64],"associations":[65],"to":[66,161],"greatly":[67],"exceed":[68],"embedding":[70],"dimension.":[71],"Our":[72,177],"main":[73],"result":[74],"sharply":[75],"characterizes":[76],"recovery":[78,152],"rates":[79],"one":[81],"step":[82],"Muon,":[84],"SGD,":[85,112,155],"Newton's":[87,116],"method":[88,117],"on":[89,169],"logistic":[91],"regression":[92],"loss":[93],"under":[94,138],"power":[96],"law":[97],"frequency":[98],"distribution.":[99],"show":[101,144],"that":[102,110,145],"storage":[104],"capacity":[105],"significantly":[108],"exceeds":[109],"even":[114],"matches":[115],"while":[118,156],"only":[119],"using":[120],"first-order":[121],"information.":[122],"Moreover,":[123],"saturates":[125],"at":[126,165],"larger":[128],"critical":[129],"batch":[130],"size.":[131],"further":[133],"analyze":[134],"multi-step":[136],"dynamics":[137],"thresholded":[140],"gradient":[141],"approximation":[142],"achieves":[147],"substantially":[149],"faster":[150],"initial":[151],"rate":[153],"than":[154],"both":[157],"methods":[158],"eventually":[159],"converge":[160],"information-theoretic":[163],"limit":[164],"comparable":[166],"speeds.":[167],"Experiments":[168],"synthetic":[170],"tasks":[171,203],"validate":[172],"predicted":[174],"scaling":[175,196],"laws.":[176],"analysis":[178],"provides":[179],"quantitative":[181],"understanding":[182],"signal":[185],"amplification":[186],"spectral":[188],"preconditioners":[189],"lays":[191],"groundwork":[193],"establishing":[195],"laws":[197],"across":[198],"more":[199],"practical":[200],"modeling":[202],"optimizers.":[205]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-31T00:00:00"}