{"id":"https://openalex.org/W7140682661","doi":"https://doi.org/10.48550/arxiv.2603.23860","title":"Why the Maximum Second Derivative of Activations Matters for Adversarial Robustness","display_name":"Why the Maximum Second Derivative of Activations Matters for Adversarial Robustness","publication_year":2026,"publication_date":"2026-03-25","ids":{"openalex":"https://openalex.org/W7140682661","doi":"https://doi.org/10.48550/arxiv.2603.23860"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.23860","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.23860","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.2603.23860","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5078590799","display_name":"Yunrui Yu","orcid":"https://orcid.org/0009-0001-1717-8687"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Yunrui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130632284","display_name":"Hang Su","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Su, Hang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5130713291","display_name":"Jun Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Jun","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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9775999784469604,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9775999784469604,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.003599999938160181,"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":0.0032999999821186066,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/hessian-matrix","display_name":"Hessian matrix","score":0.9162999987602234},{"id":"https://openalex.org/keywords/curvature","display_name":"Curvature","score":0.7784000039100647},{"id":"https://openalex.org/keywords/adversarial-system","display_name":"Adversarial system","score":0.696399986743927},{"id":"https://openalex.org/keywords/diagonal","display_name":"Diagonal","score":0.6942999958992004},{"id":"https://openalex.org/keywords/maxima-and-minima","display_name":"Maxima and minima","score":0.6841999888420105},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.6604999899864197},{"id":"https://openalex.org/keywords/norm","display_name":"Norm (philosophy)","score":0.46070000529289246},{"id":"https://openalex.org/keywords/second-derivative","display_name":"Second derivative","score":0.3903000056743622}],"concepts":[{"id":"https://openalex.org/C203616005","wikidata":"https://www.wikidata.org/wiki/Q620495","display_name":"Hessian matrix","level":2,"score":0.9162999987602234},{"id":"https://openalex.org/C195065555","wikidata":"https://www.wikidata.org/wiki/Q214881","display_name":"Curvature","level":2,"score":0.7784000039100647},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.696399986743927},{"id":"https://openalex.org/C130367717","wikidata":"https://www.wikidata.org/wiki/Q189791","display_name":"Diagonal","level":2,"score":0.6942999958992004},{"id":"https://openalex.org/C186633575","wikidata":"https://www.wikidata.org/wiki/Q845060","display_name":"Maxima and minima","level":2,"score":0.6841999888420105},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.6604999899864197},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5414000153541565},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.46630001068115234},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.46070000529289246},{"id":"https://openalex.org/C33257320","wikidata":"https://www.wikidata.org/wiki/Q1932744","display_name":"Second derivative","level":2,"score":0.3903000056743622},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3806999921798706},{"id":"https://openalex.org/C47446073","wikidata":"https://www.wikidata.org/wiki/Q5165890","display_name":"Control theory (sociology)","level":3,"score":0.36739999055862427},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.3666999936103821},{"id":"https://openalex.org/C60761296","wikidata":"https://www.wikidata.org/wiki/Q651480","display_name":"Main diagonal","level":3,"score":0.32670000195503235},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.3224000036716461},{"id":"https://openalex.org/C17685861","wikidata":"https://www.wikidata.org/wiki/Q6960532","display_name":"Time derivative","level":2,"score":0.30379998683929443},{"id":"https://openalex.org/C200331156","wikidata":"https://www.wikidata.org/wiki/Q506041","display_name":"Jacobian matrix and determinant","level":2,"score":0.2870999872684479},{"id":"https://openalex.org/C184720557","wikidata":"https://www.wikidata.org/wiki/Q7825049","display_name":"Topology (electrical circuits)","level":2,"score":0.2802000045776367},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2800999879837036},{"id":"https://openalex.org/C31531917","wikidata":"https://www.wikidata.org/wiki/Q915157","display_name":"Robust control","level":3,"score":0.26829999685287476},{"id":"https://openalex.org/C92207270","wikidata":"https://www.wikidata.org/wiki/Q939253","display_name":"Matrix norm","level":3,"score":0.2621000111103058}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.23860","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.23860","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.2603.23860","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.23860","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":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"This":[0,76],"work":[1],"investigates":[2],"the":[3,13,23,60,66,117,121,125,131,146,152],"critical":[4],"role":[5],"of":[6,65,120,124],"activation":[7,114],"function":[8],"curvature":[9,34,52,58,115],"--":[10,18],"quantified":[11],"by":[12],"maximum":[14],"second":[15],"derivative":[16],"$\\max|\u03c3''|$":[17,89],"in":[19,78],"adversarial":[20,84,105],"robustness.":[21],"Using":[22],"Recursive":[24],"Curvature-Tunable":[25],"Activation":[26],"Family":[27],"(RCT-AF),":[28],"which":[29],"enables":[30],"precise":[31],"control":[32],"over":[33],"through":[35],"parameters":[36],"$\u03b1$":[37],"and":[38,104,127],"$\u03b2$,":[39],"we":[40],"systematically":[41],"analyze":[42],"this":[43],"relationship.":[44],"Our":[45],"study":[46],"reveals":[47],"a":[48,79,95,137],"fundamental":[49],"trade-off:":[50],"insufficient":[51],"limits":[53],"model":[54],"expressivity,":[55],"while":[56],"excessive":[57],"amplifies":[59],"normalized":[61,132],"Hessian":[62,133],"diagonal":[63,118,134],"norm":[64,135],"loss,":[67,126],"leading":[68],"to":[69,93],"sharper":[70],"minima":[71],"that":[72,97,130],"hinder":[73],"robust":[74],"generalization.":[75],"results":[77],"non-monotonic":[80],"relationship":[81],"where":[82],"optimal":[83,147],"robustness":[85,148],"consistently":[86],"occurs":[87],"when":[88],"falls":[90],"within":[91,145],"4":[92],"10,":[94],"finding":[96],"holds":[98],"across":[99],"diverse":[100],"network":[101],"architectures,":[102],"datasets,":[103],"training":[106],"methods.":[107],"We":[108],"provide":[109],"theoretical":[110],"insights":[111],"into":[112],"how":[113],"affects":[116],"elements":[119],"hessian":[122],"matrix":[123],"experimentally":[128],"demonstrate":[129],"exhibits":[136],"U-shaped":[138],"dependence":[139],"on":[140],"$\\max|\u03c3''|$,":[141],"with":[142],"its":[143],"minimum":[144],"range,":[149],"thereby":[150],"validating":[151],"proposed":[153],"mechanism.":[154]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-27T00:00:00"}