{"id":"https://openalex.org/W7204238824","doi":"https://doi.org/10.48550/arxiv.2608.24007","title":"Revenge of Monosemanticity: Neuron Specialization as a New Form of Feature Learning in MLPs","display_name":"Revenge of Monosemanticity: Neuron Specialization as a New Form of Feature Learning in MLPs","publication_year":2026,"publication_date":"2026-08-25","ids":{"openalex":"https://openalex.org/W7204238824","doi":"https://doi.org/10.48550/arxiv.2608.24007"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.24007","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24007","pdf_url":"https://arxiv.org/pdf/2608.24007","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/2608.24007","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5023506063","display_name":"Amirhesam Abedsoltan","orcid":"https://orcid.org/0009-0002-3816-8856","observed_orcids":["https://orcid.org/0009-0002-3816-8856"]},"institutions":[{"id":"https://openalex.org/I36258959","display_name":"University of California San Diego","ror":"https://ror.org/0168r3w48","country_code":"US","type":"education","lineage":["https://openalex.org/I36258959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Abedsoltan, Amirhesam","raw_affiliation_strings":["Department of Computer Science and Engineering , UC San Diego"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering , UC San Diego","institution_ids":["https://openalex.org/I36258959"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5091031925","display_name":"Enric Boix-Adser\u00e0","orcid":"https://orcid.org/0000-0003-0635-9703","observed_orcids":["https://orcid.org/0000-0003-0635-9703"]},"institutions":[{"id":"https://openalex.org/I79576946","display_name":"University of Pennsylvania","ror":"https://ror.org/00b30xv10","country_code":"US","type":"education","lineage":["https://openalex.org/I79576946"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Boix-Adsera, Enric","raw_affiliation_strings":["Department of Statistics and Data Science , The Wharton School , University of Pennsylvania"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Statistics and Data Science , The Wharton School , University of Pennsylvania","institution_ids":["https://openalex.org/I79576946"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5069416974","display_name":"Fivos Kalogiannis","orcid":"https://orcid.org/0009-0002-5235-6116","observed_orcids":["https://orcid.org/0009-0002-5235-6116"]},"institutions":[{"id":"https://openalex.org/I36258959","display_name":"University of California San Diego","ror":"https://ror.org/0168r3w48","country_code":"US","type":"education","lineage":["https://openalex.org/I36258959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Kalogiannis, Fivos","raw_affiliation_strings":["Department of Computer Science and Engineering , UC San Diego"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering , UC San Diego","institution_ids":["https://openalex.org/I36258959"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5122989276","display_name":"Mikhail Belkin","orcid":"https://orcid.org/0009-0005-8633-3926","observed_orcids":["https://orcid.org/0009-0005-8633-3926"]},"institutions":[{"id":"https://openalex.org/I36258959","display_name":"University of California San Diego","ror":"https://ror.org/0168r3w48","country_code":"US","type":"education","lineage":["https://openalex.org/I36258959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Belkin, Mikhail","raw_affiliation_strings":["Department of Computer Science and Engineering , UC San Diego","Hal\u0131c\u0131oglu Data Science Institute , UC San Diego"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering , UC San Diego","institution_ids":["https://openalex.org/I36258959"]},{"raw_affiliation_string":"Hal\u0131c\u0131oglu Data Science Institute , UC San Diego","institution_ids":["https://openalex.org/I36258959"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10320","display_name":"Neural Networks and Applications","score":0.9883000254631042,"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/T10320","display_name":"Neural Networks and Applications","score":0.9883000254631042,"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":0.005400000140070915,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.002400000113993883,"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/representation-learning","display_name":"representation learning","score":0.902999997138977},{"id":"https://openalex.org/keywords/multilayer-perceptron","display_name":"multilayer perceptron","score":0.9959999918937683},{"id":"https://openalex.org/keywords/clustered-data","display_name":"clustered data","score":0.7789999842643738},{"id":"https://openalex.org/keywords/regression-analysis","display_name":"regression analysis","score":0.6710000038146973},{"id":"https://openalex.org/keywords/data-efficiency","display_name":"data efficiency","score":0.9309999942779541},{"id":"https://openalex.org/keywords/neural-network-interpretability","display_name":"neural network interpretability","score":0.5040000081062317}],"concepts":[{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6987000107765198},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6945000290870667},{"id":"https://openalex.org/C60908668","wikidata":"https://www.wikidata.org/wiki/Q690207","display_name":"Perceptron","level":3,"score":0.6402000188827515},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5795999765396118},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5767999887466431},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.5281999707221985},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.34769999980926514},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.3003999888896942},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.2849999964237213},{"id":"https://openalex.org/C179717631","wikidata":"https://www.wikidata.org/wiki/Q2991667","display_name":"Multilayer perceptron","level":3,"score":0.2757999897003174},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.26739999651908875},{"id":"https://openalex.org/C52622490","wikidata":"https://www.wikidata.org/wiki/Q1026626","display_name":"Feature extraction","level":2,"score":0.257099986076355},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.25519999861717224}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.24007","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24007","pdf_url":"https://arxiv.org/pdf/2608.24007","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.2608.24007","doi":"https://doi.org/10.48550/arxiv.2608.24007","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.24007","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:2608.24007","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24007","pdf_url":"https://arxiv.org/pdf/2608.24007","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/G1453181110","display_name":null,"funder_award_id":"#2138286","funder_id":"https://openalex.org/F4320337377","funder_display_name":"Office of Advanced Cyberinfrastructure"},{"id":"https://openalex.org/G1912064254","display_name":"Track 4:  Advanced CI Coordination Ecosystem: Monitoring and Measurement Services","funder_award_id":"2137603","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G3562386289","display_name":null,"funder_award_id":"#2138296","funder_id":"https://openalex.org/F4320337377","funder_display_name":"Office of Advanced Cyberinfrastructure"},{"id":"https://openalex.org/G3650812695","display_name":null,"funder_award_id":"HR001125CE020","funder_id":"https://openalex.org/F4320332180","funder_display_name":"Defense Advanced Research Projects Agency"},{"id":"https://openalex.org/G3744848179","display_name":"AI Institute for Learning-Enabled Optimization at Scale (TILOS)","funder_award_id":"2112665","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G4006316252","display_name":"Track 1: ACCESS Resource Allocations Marketplace and Platform Services (RAMPS)","funder_award_id":"2138259","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G4010157413","display_name":null,"funder_award_id":"2138259","funder_id":"https://openalex.org/F4320337377","funder_display_name":"Office of Advanced Cyberinfrastructure"},{"id":"https://openalex.org/G4676224565","display_name":"ACO: An Open CI Ecosystem to Advance Scientific Discovery (OpenCI)","funder_award_id":"2138296","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G509285653","display_name":"Track 3: COre National Ecosystem for CyberinfrasTructure (CONECT)","funder_award_id":"2138307","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G5367127336","display_name":null,"funder_award_id":"#2138307","funder_id":"https://openalex.org/F4320337377","funder_display_name":"Office of Advanced Cyberinfrastructure"},{"id":"https://openalex.org/G5464919783","display_name":null,"funder_award_id":"2137603","funder_id":"https://openalex.org/F4320337377","funder_display_name":"Office of Advanced Cyberinfrastructure"},{"id":"https://openalex.org/G6311202229","display_name":"Collaborative Research: MFAI: Emergence of features in modern machine learning","funder_award_id":"2502258","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G748459979","display_name":null,"funder_award_id":"N000142412631","funder_id":"https://openalex.org/F4320337345","funder_display_name":"Office of Naval Research"},{"id":"https://openalex.org/G8300691027","display_name":"Track 2: Customized Multi-tier Assistance, Training, and Computational Help (MATCH) for End User ACCESS to CI","funder_award_id":"2138286","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/F4320332180","display_name":"Defense Advanced Research Projects Agency","ror":"https://ror.org/02caytj08"},{"id":"https://openalex.org/F4320332815","display_name":"Advanced Research Projects Agency","ror":"https://ror.org/02caytj08"},{"id":"https://openalex.org/F4320337345","display_name":"Office of Naval Research","ror":"https://ror.org/00rk2pe57"},{"id":"https://openalex.org/F4320337377","display_name":"Office of Advanced Cyberinfrastructure","ror":"https://ror.org/04nh1dc89"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204238824.pdf","grobid_xml":"https://content.openalex.org/works/W7204238824.grobid-xml"},"referenced_works_count":11,"referenced_works":["https://openalex.org/W2899663614","https://openalex.org/W2981298641","https://openalex.org/W2988294134","https://openalex.org/W3006439205","https://openalex.org/W4315588185","https://openalex.org/W4316135772","https://openalex.org/W4389520786","https://openalex.org/W4392544739","https://openalex.org/W4408928873","https://openalex.org/W4417095868","https://openalex.org/W7130689667"],"related_works":["https://openalex.org/W7095614770","https://openalex.org/W3215521796","https://openalex.org/W7042549285","https://openalex.org/W1992222410","https://openalex.org/W6929423013","https://openalex.org/W2025944916","https://openalex.org/W1489162984","https://openalex.org/W4320086284","https://openalex.org/W3103242287","https://openalex.org/W1970024203"],"abstract_inverted_index":{"Understanding":[0],"how":[1],"neural":[2],"networks":[3],"learn":[4,96],"and":[5],"organize":[6],"features":[7,59],"is":[8,35],"central":[9],"to":[10,61,109],"understanding":[11],"their":[12],"behavior.":[13],"Much":[14],"existing":[15],"theory":[16],"of":[17,25,65,72,99],"feature":[18],"learning":[19,89],"has":[20],"focused":[21],"on":[22,80,121],"the":[23,66,83],"emergence":[24],"a":[26,62,90,97,113,122],"global":[27,92,123],"low-dimensional":[28,93,101,124],"representation.":[29,125],"We":[30],"show":[31,105],"that":[32,45,106],"this":[33,107],"picture":[34],"incomplete.":[36],"In":[37],"regression":[38],"problems":[39],"with":[40,57],"clustered":[41],"data,":[42],"we":[43,104],"demonstrate":[44],"multilayer":[46],"perceptrons":[47],"(MLPs)":[48],"develop":[49],"specialized":[50],"neurons:":[51],"individual":[52],"neurons":[53,74],"become":[54],"strongly":[55],"aligned":[56],"specific":[58],"relevant":[60],"particular":[63],"region":[64],"input":[67],"space.":[68],"A":[69],"substantial":[70],"fraction":[71],"these":[73],"exhibit":[75],"monosemanticity:":[76],"they":[77],"activate":[78],"predominantly":[79],"inputs":[81],"from":[82],"corresponding":[84],"cluster.":[85],"Thus,":[86],"rather":[87],"than":[88],"single":[91],"representation,":[94],"MLPs":[95,112],"collection":[98],"local":[100],"representations.":[102],"Furthermore,":[103],"ability":[108],"specialize":[110],"gives":[111],"provable":[114],"data-efficiency":[115],"advantage":[116],"over":[117],"feature-learning":[118],"methods":[119],"based":[120]},"counts_by_year":[],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2026-08-27T00:00:00"}