{"id":"https://openalex.org/W7160984472","doi":"https://doi.org/10.48550/arxiv.2605.12394","title":"Detecting overfitting in Neural Networks during long-horizon grokking using Random Matrix Theory","display_name":"Detecting overfitting in Neural Networks during long-horizon grokking using Random Matrix Theory","publication_year":2026,"publication_date":"2026-05-12","ids":{"openalex":"https://openalex.org/W7160984472","doi":"https://doi.org/10.48550/arxiv.2605.12394"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.12394","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.12394","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.12394","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136060368","display_name":"Hari K. Prakash","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Prakash, Hari K.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136088502","display_name":"Charles H Martin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Martin, Charles H","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":2,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.7092000246047974,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.7092000246047974,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.031599998474121094,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.016499999910593033,"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/overfitting","display_name":"Overfitting","score":0.972599983215332},{"id":"https://openalex.org/keywords/outlier","display_name":"Outlier","score":0.5989999771118164},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.5088000297546387},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.46970000863075256},{"id":"https://openalex.org/keywords/correlation","display_name":"Correlation","score":0.4691999852657318},{"id":"https://openalex.org/keywords/divergence","display_name":"Divergence (linguistics)","score":0.45190000534057617},{"id":"https://openalex.org/keywords/covariance-matrix","display_name":"Covariance matrix","score":0.4180999994277954},{"id":"https://openalex.org/keywords/bayesian-probability","display_name":"Bayesian probability","score":0.40049999952316284}],"concepts":[{"id":"https://openalex.org/C22019652","wikidata":"https://www.wikidata.org/wiki/Q331309","display_name":"Overfitting","level":3,"score":0.972599983215332},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6669999957084656},{"id":"https://openalex.org/C79337645","wikidata":"https://www.wikidata.org/wiki/Q779824","display_name":"Outlier","level":2,"score":0.5989999771118164},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5701000094413757},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5145999789237976},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.5088000297546387},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.46970000863075256},{"id":"https://openalex.org/C117220453","wikidata":"https://www.wikidata.org/wiki/Q5172842","display_name":"Correlation","level":2,"score":0.4691999852657318},{"id":"https://openalex.org/C207390915","wikidata":"https://www.wikidata.org/wiki/Q1230525","display_name":"Divergence (linguistics)","level":2,"score":0.45190000534057617},{"id":"https://openalex.org/C185142706","wikidata":"https://www.wikidata.org/wiki/Q1134404","display_name":"Covariance matrix","level":2,"score":0.4180999994277954},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.40049999952316284},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.34709998965263367},{"id":"https://openalex.org/C169258074","wikidata":"https://www.wikidata.org/wiki/Q245748","display_name":"Random forest","level":2,"score":0.34150001406669617},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.32109999656677246},{"id":"https://openalex.org/C64812099","wikidata":"https://www.wikidata.org/wiki/Q176604","display_name":"Random matrix","level":3,"score":0.3160000145435333},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.31529998779296875},{"id":"https://openalex.org/C106487976","wikidata":"https://www.wikidata.org/wiki/Q685816","display_name":"Matrix (chemical analysis)","level":2,"score":0.30889999866485596},{"id":"https://openalex.org/C16910744","wikidata":"https://www.wikidata.org/wiki/Q7705759","display_name":"Test data","level":2,"score":0.2935999929904938},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2825999855995178},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.2822999954223633},{"id":"https://openalex.org/C87007009","wikidata":"https://www.wikidata.org/wiki/Q210832","display_name":"Statistical hypothesis testing","level":2,"score":0.2815000116825104},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.28040000796318054},{"id":"https://openalex.org/C2780092901","wikidata":"https://www.wikidata.org/wiki/Q3433612","display_name":"Correlation coefficient","level":2,"score":0.26600000262260437},{"id":"https://openalex.org/C81363708","wikidata":"https://www.wikidata.org/wiki/Q17084460","display_name":"Convolutional neural network","level":2,"score":0.25999999046325684}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.12394","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.12394","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.12394","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.12394","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":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Training":[0],"Neural":[1],"Networks":[2],"(NNs)":[3],"without":[4,33],"overfitting":[5,10,28],"is":[6,11,146],"difficult;":[7],"detecting":[8],"that":[9,23,67,144,171],"difficult":[12],"as":[13,99],"well.":[14],"We":[15,70],"present":[16],"a":[17,60],"novel":[18],"Random":[19],"Matrix":[20],"Theory":[21],"method":[22],"detects":[24],"the":[25,54,77,84,130,135,176],"onset":[26,78],"of":[27,79,138],"in":[29,87,95],"deep":[30],"learning":[31],"models":[32],"access":[34],"to":[35,121],"train":[36,104,153],"or":[37,112],"test":[38,100,157],"data.":[39],"For":[40],"each":[41,46],"model":[42,132],"layer,":[43],"we":[44,82,116,169],"randomize":[45],"weight":[47],"matrix":[48],"element-wise,":[49],"$\\mathbf{W}":[50],"\\to":[51],"\\mathbf{W}^{\\mathrm{rand}}$,":[52],"fit":[53],"randomized":[55],"empirical":[56,119],"spectral":[57],"distribution":[58],"with":[59,151],"Marchenko-Pastur":[61],"distribution,":[62],"and":[63,93,97,133,155],"identify":[64],"large":[65],"outliers":[66,73],"violate":[68],"self-averaging.":[69],"call":[71,83],"these":[72],"Correlation":[74,90,165,178],"Traps.":[75,166],"During":[76],"overfitting,":[80],"which":[81],"\"anti-grokking\"":[85],"phase":[86,150],"long-horizon":[88],"grokking,":[89],"Traps":[91,108],"form":[92],"grow":[94],"number":[96],"scale":[98],"accuracy":[101,105,154],"decreases":[102],"while":[103],"remains":[106],"high.":[107],"may":[109,113],"be":[110],"benign":[111],"harm":[114],"generalization;":[115],"provide":[117],"an":[118,147],"approach":[120],"distinguish":[122],"between":[123],"them":[124],"by":[125],"passing":[126],"random":[127],"data":[128],"through":[129,163],"trained":[131],"evaluating":[134],"JS":[136],"divergence":[137],"output":[139],"logits.":[140],"Our":[141],"findings":[142],"show":[143],"anti-grokking":[145],"additional":[148],"grokking":[149],"high":[152],"decreasing":[156],"accuracy,":[158],"structurally":[159],"distinct":[160],"from":[161],"pre-grokking":[162],"its":[164],"More":[167],"broadly,":[168],"find":[170],"some":[172],"foundation-scale":[173],"LLMs":[174],"exhibit":[175],"same":[177],"Traps,":[179],"indicating":[180],"potentially":[181],"harmful":[182],"overfitting.":[183]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-05-14T00:00:00"}