{"id":"https://openalex.org/W7162788398","doi":"https://doi.org/10.48550/arxiv.2605.29548","title":"Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention","display_name":"Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention","publication_year":2026,"publication_date":"2026-05-28","ids":{"openalex":"https://openalex.org/W7162788398","doi":"https://doi.org/10.48550/arxiv.2605.29548"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.29548","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29548","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.2605.29548","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137339851","display_name":"Jing Huang","orcid":"https://orcid.org/0000-0001-8812-2691"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Jing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137361294","display_name":"Daniel Wurgaft","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wurgaft, Daniel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137360370","display_name":"Rachit Bansal","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bansal, Rachit","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137319905","display_name":"Laura Ruis","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ruis, Laura","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137369316","display_name":"Naomi Saphra","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saphra, Naomi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137315451","display_name":"David Alvarez-Melis","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Alvarez-Melis, David","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030015839","display_name":"Andrew K. Lampinen","orcid":"https://orcid.org/0000-0002-6988-8437"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lampinen, Andrew Kyle","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137369036","display_name":"Christopher Potts","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Potts, Christopher","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137390134","display_name":"Ekdeep Singh Lubana","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lubana, Ekdeep Singh","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/T10581","display_name":"Neural dynamics and brain function","score":0.17550000548362732,"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"}},"topics":[{"id":"https://openalex.org/T10581","display_name":"Neural dynamics and brain function","score":0.17550000548362732,"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"}},{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.10209999978542328,"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/T11094","display_name":"Face Recognition and Perception","score":0.10010000318288803,"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/task","display_name":"Task (project management)","score":0.7434999942779541},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.6783999800682068},{"id":"https://openalex.org/keywords/interference","display_name":"Interference (communication)","score":0.600600004196167},{"id":"https://openalex.org/keywords/monotonic-function","display_name":"Monotonic function","score":0.5976999998092651},{"id":"https://openalex.org/keywords/simple","display_name":"Simple (philosophy)","score":0.5353999733924866},{"id":"https://openalex.org/keywords/point","display_name":"Point (geometry)","score":0.5271999835968018},{"id":"https://openalex.org/keywords/computational-model","display_name":"Computational model","score":0.4165000021457672},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.36890000104904175}],"concepts":[{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.7434999942779541},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.6783999800682068},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6437000036239624},{"id":"https://openalex.org/C32022120","wikidata":"https://www.wikidata.org/wiki/Q797225","display_name":"Interference (communication)","level":3,"score":0.600600004196167},{"id":"https://openalex.org/C72169020","wikidata":"https://www.wikidata.org/wiki/Q194404","display_name":"Monotonic function","level":2,"score":0.5976999998092651},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.5353999733924866},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.5271999835968018},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4952000081539154},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.4165000021457672},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.36890000104904175},{"id":"https://openalex.org/C98184364","wikidata":"https://www.wikidata.org/wiki/Q1780131","display_name":"Argument (complex analysis)","level":2,"score":0.35359999537467957},{"id":"https://openalex.org/C91682802","wikidata":"https://www.wikidata.org/wiki/Q620538","display_name":"Multidimensional scaling","level":2,"score":0.3456999957561493},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.33820000290870667},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3199999928474426},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.31139999628067017},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.31040000915527344},{"id":"https://openalex.org/C76969082","wikidata":"https://www.wikidata.org/wiki/Q486902","display_name":"Mathematical model","level":2,"score":0.3034999966621399},{"id":"https://openalex.org/C2779569378","wikidata":"https://www.wikidata.org/wiki/Q25303768","display_name":"Phenomenological model","level":2,"score":0.2705000042915344},{"id":"https://openalex.org/C21080849","wikidata":"https://www.wikidata.org/wiki/Q13611879","display_name":"Data point","level":2,"score":0.26919999718666077},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.26829999685287476},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.26429998874664307},{"id":"https://openalex.org/C2779714256","wikidata":"https://www.wikidata.org/wiki/Q25305062","display_name":"Multiple Models","level":2,"score":0.25270000100135803}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.29548","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29548","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.2605.29548","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29548","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Larger":[0],"models":[1,5,92,150,189,215,225,249,254,262],"learn":[2,31,106,216,250],"tasks":[3,74,157,164,196,251],"smaller":[4,40,91,253],"do":[6,171],"not.":[7],"What":[8],"drives":[9],"this":[10,52,117,137],"phenomenon?":[11],"We":[12,129],"develop":[13],"a":[14,24,32,39,66,71,84,133,145,243],"simple":[15],"phenomenological":[16],"argument":[17],"that":[18,23,38,75,108,141,158,169,252],"power-law":[19],"scaling":[20,64,78],"already":[21],"suggests":[22],"larger":[25,134,149,213,224,248,261],"model":[26,41,63,135,274],"will":[27],"be":[28],"able":[29],"to":[30,43,83,96,144,155,181,191],"part":[33],"of":[34,62,70,73,123,197,246],"the":[35,60,125,159,212,217],"data":[36,209,278],"distribution":[37],"fails":[42],"learn,":[44],"even":[45,119],"with":[46],"infinite":[47],"training":[48,277],"data.":[49],"To":[50],"validate":[51,183],"claim":[53],"and":[54,103,113,200,219,222,233,267,276],"identify":[55],"its":[56],"causes,":[57],"we":[58,186,241],"study":[59],"effects":[61],"on":[65,111,194],"synthetic":[67,208],"setup":[68],"consisting":[69],"mixture":[72],"show":[76,234],"monotonic":[77],"curves.":[79],"The":[80,202],"results":[81,203],"point":[82],"data-induced":[85],"competition":[86],"over":[87],"resources":[88,154],"(neurons).":[89],"Specifically,":[90],"allocate":[93,152],"their":[94,231],"neurons":[95],"high":[97],"frequency":[98,199],"or":[99],"low":[100],"complexity":[101],"tasks,":[102,221],"so":[104],"they":[105,170,177],"solutions":[107,121],"perform":[109],"poorly":[110],"rare":[112],"complex":[114,220],"tasks.":[115,239],"Moreover,":[116],"happens":[118],"when":[120],"capable":[122],"expressing":[124],"desired":[126],"task":[127,228],"exist.":[128],"then":[130],"assess":[131],"how":[132],"circumvents":[136],"data-centric":[138,244],"bottleneck,":[139],"finding":[140],"it":[142,268],"traces":[143],"reduced":[146],"interference":[147,237],"mechanism:":[148],"can":[151,269],"enough":[153],"common":[156],"gradient":[160,236],"updates":[161],"for":[162],"those":[163,205],"become":[165],"weak,":[166],"which":[167],"means":[168],"not":[172],"overwrite":[173],"rare-task":[174],"features":[175,229],"as":[176],"slowly":[178],"accumulate.":[179],"Finally,":[180],"further":[182],"these":[184,223],"claims,":[185],"pretrain":[187],"OLMo":[188,214],"(4M":[190],"4B":[192],"parameters)":[193],"novel":[195],"varying":[198],"complexity.":[201],"mirror":[204],"from":[206],"our":[207],"experiments:":[210],"only":[211],"infrequent":[218],"embed":[226],"more":[227],"in":[230,265],"representations":[232],"less":[235],"between":[238],"Overall,":[240],"offer":[242],"account":[245],"why":[247,260],"fail":[255],"to.":[256],"This":[257],"helps":[258],"explain":[259],"are":[263],"better":[264],"practice,":[266],"inform":[270],"practical":[271],"questions":[272],"concerning":[273],"sizing":[275],"mixtures.":[279]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-30T00:00:00"}