{"id":"https://openalex.org/W7140161910","doi":"https://doi.org/10.48550/arxiv.2603.21658","title":"A Comparative Analysis of LLM Memorization at Statistical and Internal Levels: Cross-Model Commonalities and Model-Specific Signatures","display_name":"A Comparative Analysis of LLM Memorization at Statistical and Internal Levels: Cross-Model Commonalities and Model-Specific Signatures","publication_year":2026,"publication_date":"2026-03-23","ids":{"openalex":"https://openalex.org/W7140161910","doi":"https://doi.org/10.48550/arxiv.2603.21658"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.21658","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.21658","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,"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.2603.21658","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5146887848","display_name":"Bowen Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Bowen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5087793573","display_name":"Namgi Han","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Namgi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5130458663","display_name":"Yusuke Miyao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Miyao, Yusuke","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":1,"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/T11577","display_name":"Cognitive Abilities and Testing","score":0.1868000030517578,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T11577","display_name":"Cognitive Abilities and Testing","score":0.1868000030517578,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.12070000171661377,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10429","display_name":"EEG and Brain-Computer Interfaces","score":0.04820000007748604,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/memorization","display_name":"Memorization","score":0.944599986076355},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.536300003528595},{"id":"https://openalex.org/keywords/discriminative-model","display_name":"Discriminative model","score":0.47350001335144043},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.47200000286102295},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.4490000009536743},{"id":"https://openalex.org/keywords/statistical-model","display_name":"Statistical model","score":0.3986999988555908},{"id":"https://openalex.org/keywords/statistical-analysis","display_name":"Statistical analysis","score":0.3837999999523163}],"concepts":[{"id":"https://openalex.org/C30038468","wikidata":"https://www.wikidata.org/wiki/Q4354775","display_name":"Memorization","level":2,"score":0.944599986076355},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6190000176429749},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.536300003528595},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.47350001335144043},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.47200000286102295},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.46389999985694885},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.4490000009536743},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.3986999988555908},{"id":"https://openalex.org/C2986587452","wikidata":"https://www.wikidata.org/wiki/Q938438","display_name":"Statistical analysis","level":2,"score":0.3837999999523163},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.3709999918937683},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.36169999837875366},{"id":"https://openalex.org/C87007009","wikidata":"https://www.wikidata.org/wiki/Q210832","display_name":"Statistical hypothesis testing","level":2,"score":0.350600004196167},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.33570000529289246},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.33500000834465027},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.32100000977516174},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3174000084400177},{"id":"https://openalex.org/C48677424","wikidata":"https://www.wikidata.org/wiki/Q6888088","display_name":"Mode (computer interface)","level":2,"score":0.303600013256073},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.29649999737739563},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.28209999203681946},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.2538999915122986},{"id":"https://openalex.org/C45493050","wikidata":"https://www.wikidata.org/wiki/Q7884934","display_name":"Unified Model","level":2,"score":0.2506999969482422},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2500999867916107}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.21658","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.21658","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,"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.2603.21658","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.21658","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,"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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Memorization":[0],"is":[1],"a":[2,38,116,187,205],"fundamental":[3,208],"component":[4],"of":[5,20,31,179,210],"intelligence":[6],"for":[7,123,174,204],"both":[8,78],"humans":[9],"and":[10,69,81,106,119,160,170,195,207],"LLMs.":[11],"However,":[12,126,176],"while":[13,87,150],"LLM":[14],"performance":[15],"scales":[16,101],"rapidly,":[17],"our":[18],"understanding":[19,209],"memorization":[21,75,99,211],"lags.":[22],"Due":[23],"to":[24,27,43],"limited":[25],"access":[26],"the":[28,79,92,98,134,138,166,177,202],"pre-training":[29],"data":[30],"LLMs,":[32],"most":[33],"previous":[34],"studies":[35],"focus":[36],"on":[37],"single":[39],"model":[40,63,104],"series,":[41,47],"leading":[42],"isolated":[44],"observations":[45,86],"among":[46],"making":[48],"it":[49],"unclear":[50],"which":[51],"findings":[52],"are":[53,153],"general":[54,167],"or":[55,73],"specific.":[56],"In":[57],"this":[58,199],"study,":[59],"we":[60,95,141,164],"collect":[61],"multiple":[62],"series":[64],"(Pythia,":[65],"OpenLLaMa,":[66],"StarCoder,":[67],"OLMo1/2/3)":[68],"analyze":[70],"their":[71],"shared":[72,117,171],"unique":[74,188],"behavior":[76],"at":[77],"statistical":[80,93],"internal":[82,139],"levels,":[83],"connecting":[84],"individual":[85,131],"showing":[88,186],"new":[89,197],"findings.":[90],"At":[91,137],"level,":[94,140],"reveal":[96],"that":[97,143],"rate":[100],"log-linearly":[102],"with":[103],"size,":[105],"memorized":[107,124,151],"sequences":[108,152],"can":[109,145],"be":[110],"further":[111],"compressed.":[112],"Further":[113],"analysis":[114],"demonstrated":[115],"frequency":[118],"domain":[120],"distribution":[121,178],"pattern":[122],"sequences.":[125],"different":[127],"models":[128],"also":[129],"show":[130],"features":[132],"under":[133],"above":[135],"observations.":[136],"find":[142],"LLMs":[144],"remove":[146],"certain":[147],"injected":[148],"perturbations,":[149],"more":[154],"sensitive.":[155],"By":[156],"decoding":[157,168],"middle":[158],"layers":[159],"attention":[161],"head":[162],"ablation,":[163],"revealed":[165],"process":[169],"important":[172,181],"heads":[173,182],"memorization.":[175],"those":[180],"differs":[183],"between":[184],"families,":[185],"family-level":[189],"feature.":[190],"Through":[191],"bridging":[192],"various":[193],"experiments":[194],"revealing":[196],"findings,":[198],"study":[200],"paves":[201],"way":[203],"universal":[206],"in":[212],"LLM.":[213]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-08-12T21:12:35.861297","created_date":"2026-03-25T00:00:00"}