{"id":"https://openalex.org/W7169542570","doi":"https://doi.org/10.48550/arxiv.2607.14576","title":"Certifying Residual Architectures from Their Primitives: Growth Exponents and Criticality","display_name":"Certifying Residual Architectures from Their Primitives: Growth Exponents and Criticality","publication_year":2026,"publication_date":"2026-07-16","ids":{"openalex":"https://openalex.org/W7169542570","doi":"https://doi.org/10.48550/arxiv.2607.14576"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2607.14576","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2607.14576","pdf_url":"https://arxiv.org/pdf/2607.14576","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2607.14576","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5141074681","display_name":"Hyemin Gu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1298353152","display_name":"SRI International","ror":"https://ror.org/05s570m15","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I1298353152"]},{"id":"https://openalex.org/I24603500","display_name":"University of Massachusetts Amherst","ror":"https://ror.org/0072zz521","country_code":"US","type":"education","lineage":["https://openalex.org/I24603500"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Gu, Hyemin","raw_affiliation_strings":["University of Massachusetts Amherst SRI International"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Massachusetts Amherst SRI International","institution_ids":["https://openalex.org/I1298353152","https://openalex.org/I24603500"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136098054","display_name":"Michael Tyrrell","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1298353152","display_name":"SRI International","ror":"https://ror.org/05s570m15","country_code":"US","type":"nonprofit","lineage":["https://openalex.org/I1298353152"]},{"id":"https://openalex.org/I24603500","display_name":"University of Massachusetts Amherst","ror":"https://ror.org/0072zz521","country_code":"US","type":"education","lineage":["https://openalex.org/I24603500"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tyrrell, Michael","raw_affiliation_strings":["University of Massachusetts Amherst SRI International"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Massachusetts Amherst SRI International","institution_ids":["https://openalex.org/I1298353152","https://openalex.org/I24603500"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5057249914","display_name":"Tuhin Sahai","orcid":"https://orcid.org/0000-0003-1896-8768","observed_orcids":["https://orcid.org/0000-0003-1896-8768"]},"institutions":[{"id":"https://openalex.org/I24603500","display_name":"University of Massachusetts Amherst","ror":"https://ror.org/0072zz521","country_code":"US","type":"education","lineage":["https://openalex.org/I24603500"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Sahai, Tuhin","raw_affiliation_strings":["SRI International University of Massachusetts Amherst"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SRI International University of Massachusetts Amherst","institution_ids":["https://openalex.org/I24603500"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5045079747","display_name":"Markos A. Katsoulakis","orcid":"https://orcid.org/0000-0003-4354-1766","observed_orcids":["https://orcid.org/0000-0003-4354-1766"]},"institutions":[{"id":"https://openalex.org/I24603500","display_name":"University of Massachusetts Amherst","ror":"https://ror.org/0072zz521","country_code":"US","type":"education","lineage":["https://openalex.org/I24603500"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Katsoulakis, Markos A.","raw_affiliation_strings":["SRI International University of Massachusetts Amherst"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"SRI International University of Massachusetts Amherst","institution_ids":["https://openalex.org/I24603500"]}]}],"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":0.5746999979019165,"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":0.5746999979019165,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.15760000050067902,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.07739999890327454,"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/floating-point-precision","display_name":"floating-point precision","score":0.6679999828338623},{"id":"https://openalex.org/keywords/backward-stability","display_name":"backward stability","score":0.7459999918937683},{"id":"https://openalex.org/keywords/normalization","display_name":"normalization","score":0.5220000147819519},{"id":"https://openalex.org/keywords/phase-transitions","display_name":"phase transitions","score":0.6380000114440918},{"id":"https://openalex.org/keywords/mamba","display_name":"Mamba","score":0.6299999952316284},{"id":"https://openalex.org/keywords/toto","display_name":"Toto","score":0.5289999842643738},{"id":"https://openalex.org/keywords/gpt-2","display_name":"GPT-2","score":0.7960000038146973},{"id":"https://openalex.org/keywords/operator-learning","display_name":"operator learning","score":0.6660000085830688},{"id":"https://openalex.org/keywords/forecasting","display_name":"forecasting","score":0.34200000762939453}],"concepts":[{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.710099995136261},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.6744999885559082},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.6162999868392944},{"id":"https://openalex.org/C34862557","wikidata":"https://www.wikidata.org/wiki/Q178985","display_name":"Ode","level":2,"score":0.5120000243186951},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.4544000029563904},{"id":"https://openalex.org/C2780388253","wikidata":"https://www.wikidata.org/wiki/Q5421508","display_name":"Exponent","level":2,"score":0.4440999925136566},{"id":"https://openalex.org/C47446073","wikidata":"https://www.wikidata.org/wiki/Q5165890","display_name":"Control theory (sociology)","level":3,"score":0.41819998621940613},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.4120999872684479},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.390500009059906},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3714999854564667},{"id":"https://openalex.org/C38349280","wikidata":"https://www.wikidata.org/wiki/Q1434290","display_name":"Flow (mathematics)","level":2,"score":0.36890000104904175},{"id":"https://openalex.org/C62354387","wikidata":"https://www.wikidata.org/wiki/Q875399","display_name":"Boundary (topology)","level":2,"score":0.3450999855995178},{"id":"https://openalex.org/C134306372","wikidata":"https://www.wikidata.org/wiki/Q7754","display_name":"Mathematical analysis","level":1,"score":0.3440999984741211},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.3337000012397766},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.3310000002384186},{"id":"https://openalex.org/C2778925954","wikidata":"https://www.wikidata.org/wiki/Q16883749","display_name":"Stability criterion","level":3,"score":0.3305000066757202},{"id":"https://openalex.org/C45347329","wikidata":"https://www.wikidata.org/wiki/Q5166604","display_name":"Convolution (computer science)","level":3,"score":0.295199990272522},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.2903999984264374},{"id":"https://openalex.org/C167964875","wikidata":"https://www.wikidata.org/wiki/Q17011487","display_name":"Exponential stability","level":3,"score":0.2883000075817108},{"id":"https://openalex.org/C46304622","wikidata":"https://www.wikidata.org/wiki/Q374814","display_name":"Certification","level":2,"score":0.2825999855995178},{"id":"https://openalex.org/C143170015","wikidata":"https://www.wikidata.org/wiki/Q17007850","display_name":"Stability conditions","level":3,"score":0.28110000491142273},{"id":"https://openalex.org/C182310444","wikidata":"https://www.wikidata.org/wiki/Q1332643","display_name":"Boundary value problem","level":2,"score":0.27630001306533813},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.27559998631477356},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.26840001344680786},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.2563999891281128}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2607.14576","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2607.14576","pdf_url":"https://arxiv.org/pdf/2607.14576","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.2607.14576","doi":"https://doi.org/10.48550/arxiv.2607.14576","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2607.14576","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:2607.14576","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2607.14576","pdf_url":"https://arxiv.org/pdf/2607.14576","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":[{"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"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7169542570.pdf","grobid_xml":"https://content.openalex.org/works/W7169542570.grobid-xml"},"referenced_works_count":1,"referenced_works":["https://openalex.org/W3037932933"],"related_works":["https://openalex.org/W7203674978","https://openalex.org/W4289345622","https://openalex.org/W2789439684","https://openalex.org/W7142114282","https://openalex.org/W7221025593","https://openalex.org/W7140305772","https://openalex.org/W2986677756","https://openalex.org/W3198312958","https://openalex.org/W4303644632","https://openalex.org/W2804262948"],"abstract_inverted_index":{"Whether":[0],"a":[1,30,42,86,159,202],"deep":[2],"residual":[3,27,140],"architecture":[4],"avoids":[5],"numerical":[6],"blow-up":[7],"in":[8,103,117,172,194],"its":[9],"forward":[10,55],"and":[11,26,59,70,100,108,142,158,174,196,219],"backward":[12,126],"dynamics":[13],"is":[14],"typically":[15],"determined":[16],"through":[17],"costly":[18],"trial-and-error":[19,218],"training.":[20,183,225],"Existing":[21],"stabilization":[22],"strategies,":[23],"including":[24,138],"normalization":[25],"scaling,":[28],"lack":[29],"common":[31],"framework":[32,45,206],"for":[33,89],"predicting":[34],"their":[35],"effects":[36],"across":[37],"architectures.":[38],"We":[39],"develop":[40],"an":[41],"priori":[43],"certification":[44,134],"based":[46],"on":[47,153],"two":[48],"exponents":[49,66],"derived":[50],"from":[51,135],"architectural":[52],"primitives:":[53],"the":[54,60,78,104,163,215],"growth":[56,169],"exponent":[57,63,122],"$q$":[58],"Jacobian":[61],"sensitivity":[62,121,179],"$\u03c3$.":[64],"These":[65],"characterize":[67],"stability":[68,165],"regimes":[69,105],"determine":[71],"how":[72],"certified":[73,91],"allowable":[74],"depth":[75,92],"scales":[76,93],"with":[77,211],"floating-point":[79],"limit":[80],"$M$.":[81],"In":[82],"particular,":[83],"$q=1$":[84,187],"marks":[85],"sharp":[87],"threshold:":[88],"$\u03c3>0$,":[90],"as":[94],"$\\log":[95],"M/\\log\\log":[96],"M$,":[97],"$\\sqrt{\\log":[98],"M}$,":[99],"$\\log_q\\log":[101],"M$":[102],"$q<1$,":[106],"$q=1$,":[107],"$q>1$,":[109],"respectively,":[110],"exhibiting":[111],"changes":[112],"reminiscent":[113],"of":[114,209,217],"phase":[115],"transitions":[116],"statistical":[118],"physics.":[119],"The":[120],"$\u03c3$":[123],"further":[124],"constrains":[125],"stability,":[127,213],"even":[128],"when":[129],"$q\\le1$.":[130],"Exponent":[131],"arithmetic":[132],"enables":[133],"computation":[136],"graphs,":[137],"heterogeneous":[139],"stacks,":[141],"identifies":[143,207],"targeted":[144],"modifications":[145],"to":[146],"existing":[147],"backbones":[148],"without":[149],"uniform":[150],"normalization.":[151],"Experiments":[152],"Mamba,":[154],"PatchTST,":[155],"FNO,":[156],"GPT-2,":[157],"Toto~2.0":[160],"proxy":[161],"support":[162],"predicted":[164],"distinctions:":[166],"controlling":[167],"superlinear":[168],"prevents":[170],"divergence":[171],"Mamba":[173],"Toto~2.0,":[175],"while":[176],"reducing":[177],"attention":[178],"restores":[180],"stable":[181,185],"GPT-2":[182],"Among":[184],"designs,":[186],"architectures":[188,210],"can":[189],"outperform":[190],"normalized":[191],"$q=0$":[192],"counterparts":[193],"forecasting":[195],"operator-learning":[197],"extrapolation.":[198],"Rather":[199],"than":[200],"prescribing":[201],"single":[203],"architecture,":[204],"our":[205],"families":[208],"certifiable":[212],"narrowing":[214],"scope":[216],"enabling":[220],"systematic,":[221],"task-dependent":[222],"exploration":[223],"before":[224]},"counts_by_year":[],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2026-07-18T00:00:00"}