{"id":"https://openalex.org/W4320170046","doi":"https://doi.org/10.48550/arxiv.2206.04817","title":"The Slingshot Mechanism: An Empirical Study of Adaptive Optimizers and the Grokking Phenomenon","display_name":"The Slingshot Mechanism: An Empirical Study of Adaptive Optimizers and the Grokking Phenomenon","publication_year":2022,"publication_date":"2022-06-10","ids":{"openalex":"https://openalex.org/W4320170046","doi":"https://doi.org/10.48550/arxiv.2206.04817"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2206.04817","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2206.04817","pdf_url":"https://arxiv.org/pdf/2206.04817","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2206.04817","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5059333418","display_name":"Vimal Thilak","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Thilak, Vimal","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049929304","display_name":"Etai Littwin","orcid":"https://orcid.org/0009-0001-7396-4658","observed_orcids":["https://orcid.org/0009-0001-7396-4658"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Littwin, Etai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5031329242","display_name":"Shuangfei Zhai","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhai, Shuangfei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011650393","display_name":"Omid Saremi","orcid":"https://orcid.org/0000-0001-6718-093X","observed_orcids":["https://orcid.org/0000-0001-6718-093X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saremi, Omid","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023659200","display_name":"Roni Paiss","orcid":"https://orcid.org/0000-0003-3792-1770","observed_orcids":["https://orcid.org/0000-0003-3792-1770"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Paiss, Roni","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5033404184","display_name":"Joshua M. Susskind","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Susskind, Joshua","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":true,"cited_by_count":4,"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":0.9916999936103821,"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":0.9916999936103821,"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":0.0032999999821186066,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.0010999999940395355,"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/grokking","display_name":"grokking","score":0.8450000286102295},{"id":"https://openalex.org/keywords/overfitting","display_name":"overfitting","score":0.8180000185966492},{"id":"https://openalex.org/keywords/generalization","display_name":"generalization","score":0.7329999804496765},{"id":"https://openalex.org/keywords/inductive-bias","display_name":"inductive bias","score":0.6539999842643738}],"concepts":[{"id":"https://openalex.org/C22019652","wikidata":"https://www.wikidata.org/wiki/Q331309","display_name":"Overfitting","level":3,"score":0.8190503120422363},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.6655328273773193},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6182493567466736},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.5932059288024902},{"id":"https://openalex.org/C50335755","wikidata":"https://www.wikidata.org/wiki/Q483247","display_name":"Phenomenon","level":2,"score":0.5676894783973694},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.4356471002101898},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.4139992594718933},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.40297406911849976},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.18063241243362427},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.16668152809143066},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.08270424604415894},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0},{"id":"https://openalex.org/C199539241","wikidata":"https://www.wikidata.org/wiki/Q7748","display_name":"Law","level":1,"score":0.0},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.0},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2206.04817","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2206.04817","pdf_url":"https://arxiv.org/pdf/2206.04817","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2206.04817","doi":"https://doi.org/10.48550/arxiv.2206.04817","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2206.04817","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2206.04817","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2206.04817","pdf_url":"https://arxiv.org/pdf/2206.04817","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7138140132","https://openalex.org/W7160633226","https://openalex.org/W7166698946","https://openalex.org/W7165496265","https://openalex.org/W7108888519","https://openalex.org/W4410932063","https://openalex.org/W4308670011","https://openalex.org/W7172384682","https://openalex.org/W4415025516","https://openalex.org/W3037278715"],"abstract_inverted_index":{"The":[0],"grokking":[1],"phenomenon":[2],"as":[3,66,116],"reported":[4,117],"by":[5,24,80,96],"Power":[6],"et":[7],"al.":[8],"(":[9,119],"arXiv:2201.02177":[10,120],")":[11,121],"refers":[12],"to":[13,29,37,65,173],"a":[14,17,25,44,174,191],"regime":[15],"where":[16],"long":[18],"period":[19],"of":[20,41,46,62,73,100,103,128,180,187,195],"overfitting":[21],"is":[22,131],"followed":[23],"seemingly":[26],"sudden":[27],"transition":[28],"perfect":[30],"generalization.":[31],"In":[32],"this":[33],"paper,":[34],"we":[35,50,156],"attempt":[36],"reveal":[38],"the":[39,67,74,81,97,101,104,126,144],"underpinnings":[40],"Grokking":[42,115],"via":[43],"series":[45],"empirical":[47],"studies.":[48],"Specifically,":[49],"uncover":[51],"an":[52,166],"optimization":[53,153],"anomaly":[54],"plaguing":[55],"adaptive":[56,181],"optimizers":[57,183],"at":[58,125,184],"extremely":[59],"late":[60,185],"stages":[61,186],"training,":[63,188],"referred":[64],"Slingshot":[68,75,145],"Mechanism.":[69],"A":[70],"prominent":[71],"artifact":[72],"Mechanism":[76,146],"can":[77,92,161],"be":[78,93,162],"measured":[79],"cyclic":[82,98],"phase":[83],"transitions":[84],"between":[85],"stable":[86],"and":[87,91,130,137,160,176],"unstable":[88],"training":[89],"regimes,":[90],"easily":[94,138,163],"monitored":[95],"behavior":[99],"norm":[102],"last":[105],"layers":[106],"weights.":[107],"We":[108],"empirically":[109],"observe":[110],"that":[111,155],"without":[112,133,165],"explicit":[113],"regularization,":[114],"in":[118,140,167],"almost":[122],"exclusively":[123],"happens":[124],"onset":[127],"Slingshots,":[129],"absent":[132],"it.":[134],"While":[135],"common":[136],"reproduced":[139],"more":[141],"general":[142],"settings,":[143],"does":[147],"not":[148],"follow":[149],"from":[150],"any":[151],"known":[152],"theories":[154],"are":[157],"aware":[158],"of,":[159],"overlooked":[164],"depth":[168],"examination.":[169],"Our":[170],"work":[171],"points":[172],"surprising":[175],"useful":[177],"inductive":[178],"bias":[179],"gradient":[182],"calling":[189],"for":[190],"revised":[192],"theoretical":[193],"analysis":[194],"their":[196],"origin.":[197]},"counts_by_year":[{"year":2025,"cited_by_count":2},{"year":2024,"cited_by_count":1},{"year":2023,"cited_by_count":1}],"updated_date":"2026-09-21T07:43:02.447757","created_date":"2023-02-13T00:00:00"}