{"id":"https://openalex.org/W7224956891","doi":"https://doi.org/10.48550/arxiv.2610.39455","title":"Sparse Fine-Tuning for Parameter-Efficient Adversarial Training","display_name":"Sparse Fine-Tuning for Parameter-Efficient Adversarial Training","publication_year":2026,"publication_date":"2026-10-29","ids":{"openalex":"https://openalex.org/W7224956891","doi":"https://doi.org/10.48550/arxiv.2610.39455"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.39455","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39455","pdf_url":"https://arxiv.org/pdf/2610.39455","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/2610.39455","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5160234208","display_name":"Zhaoxin Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Zhaoxin","raw_affiliation_strings":["School of Artificial Intelligence Xidian University Xi'an , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence Xidian University Xi'an , China","institution_ids":["https://openalex.org/I149594827"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160221133","display_name":"Zihang Ding","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ding, Zihang","raw_affiliation_strings":["School of Artificial Intelligence Xidian University Xi'an , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence Xidian University Xi'an , China","institution_ids":["https://openalex.org/I149594827"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5160197477","display_name":"Handing Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I149594827","display_name":"Xidian University","ror":"https://ror.org/05s92vm98","country_code":"CN","type":"education","lineage":["https://openalex.org/I149594827"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Handing","raw_affiliation_strings":["School of Artificial Intelligence Xidian University Xi'an , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Artificial Intelligence Xidian University Xi'an , China","institution_ids":["https://openalex.org/I149594827"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I149594827"],"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":1.0,"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":1.0,"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/T11424","display_name":"Security and Verification in Computing","score":0.0,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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":0.0,"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/adversarial-training","display_name":"adversarial training","score":0.8740000128746033},{"id":"https://openalex.org/keywords/parameter-efficient-fine-tuning","display_name":"parameter-efficient fine-tuning","score":0.9900000095367432},{"id":"https://openalex.org/keywords/accuracy-robustness-trade-off","display_name":"robustness-accuracy trade-off","score":0.6890000104904175},{"id":"https://openalex.org/keywords/lora","display_name":"LoRa","score":0.609000027179718},{"id":"https://openalex.org/keywords/adversarial-perturbations","display_name":"adversarial perturbations","score":0.7089999914169312},{"id":"https://openalex.org/keywords/neural-network-robustness","display_name":"neural network robustness","score":0.6029999852180481},{"id":"https://openalex.org/keywords/memory-efficient-training","display_name":"memory-efficient training","score":0.5680000185966492}],"concepts":[{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.9358999729156494},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.8047000169754028},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7037000060081482},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.6725999712944031},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6496999859809875},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5460000038146973},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.413100004196167},{"id":"https://openalex.org/C2776036281","wikidata":"https://www.wikidata.org/wiki/Q48769818","display_name":"Constraint (computer-aided design)","level":2,"score":0.413100004196167},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.38370001316070557},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.34380000829696655},{"id":"https://openalex.org/C193254401","wikidata":"https://www.wikidata.org/wiki/Q2160088","display_name":"Robust optimization","level":2,"score":0.3343999981880188},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.29760000109672546},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2872999906539917},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2743000090122223},{"id":"https://openalex.org/C73586568","wikidata":"https://www.wikidata.org/wiki/Q2600211","display_name":"Parameter space","level":2,"score":0.27160000801086426},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2581999897956848},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.2567000091075897}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.39455","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39455","pdf_url":"https://arxiv.org/pdf/2610.39455","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.2610.39455","doi":"https://doi.org/10.48550/arxiv.2610.39455","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.39455","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:2610.39455","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.39455","pdf_url":"https://arxiv.org/pdf/2610.39455","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":[{"id":"https://openalex.org/G754480656","display_name":null,"funder_award_id":"62376202","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7224956891.pdf","grobid_xml":"https://content.openalex.org/works/W7224956891.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4402952703","https://openalex.org/W7154509587","https://openalex.org/W4387210943","https://openalex.org/W4392008385","https://openalex.org/W7164885886","https://openalex.org/W3118168379","https://openalex.org/W7154951325","https://openalex.org/W7166863014","https://openalex.org/W4402352843","https://openalex.org/W2979524290"],"abstract_inverted_index":{"Deep":[0],"neural":[1],"networks":[2],"achieve":[3],"strong":[4],"performance":[5],"across":[6],"many":[7],"tasks":[8],"but":[9,51],"remain":[10],"vulnerable":[11],"to":[12,60],"adversarial":[13,69,82,122],"perturbations.":[14],"Adversarial":[15,115],"training":[16],"(AT)":[17],"is":[18,84,102],"one":[19],"of":[20,91,169],"the":[21,36,135],"most":[22],"effective":[23],"defenses,":[24],"yet":[25],"it":[26],"suffers":[27],"from":[28,71],"high":[29],"memory":[30],"cost":[31],"on":[32,87,109],"large":[33],"models":[34],"and":[35,125,139,148,155],"persistent":[37],"clean-robust":[38],"accuracy":[39,150],"trade-off.":[40],"Recent":[41],"work":[42],"introduces":[43],"parameter-efficient":[44,68],"fine-tuning,":[45],"such":[46],"as":[47],"LoRA,":[48],"into":[49],"AT,":[50],"this":[52,64,110],"constraint":[53],"induces":[54],"a":[55,72,88,105],"notable":[56],"robustness":[57,101,163],"gap":[58],"relative":[59],"full-parameter":[61],"AT.":[62],"In":[63],"work,":[65],"we":[66,79,94,112],"revisit":[67],"fine-tuning":[70],"parameter":[73,160],"space":[74],"perspective.":[75],"Through":[76],"gradient":[77,123],"analysis,":[78],"find":[80],"that":[81,100],"optimization":[83],"highly":[85],"concentrated":[86],"small":[89],"subset":[90],"parameters,":[92,154],"which":[93,118],"call":[95],"Robustness-Critical":[96],"Parameters":[97],"(RCPs),":[98],"suggesting":[99],"encoded":[103],"in":[104,145,162],"sparse":[106],"subspace.":[107],"Building":[108],"observation,":[111],"propose":[113],"Sparse":[114],"Fine-Tuning":[116],"(SAFT),":[117],"identifies":[119],"RCPs":[120],"via":[121],"saliency":[124],"performs":[126],"AT":[127,161],"by":[128],"updating":[129,165],"only":[130,166],"these":[131],"parameters":[132],"while":[133,164],"freezing":[134],"rest.":[136],"Across":[137],"architectures":[138],"datasets,":[140],"SAFT":[141],"outperforms":[142],"LoRA-based":[143],"methods":[144],"both":[146],"clean":[147],"robust":[149],"with":[151],"fewer":[152],"trainable":[153],"approaches":[156],"or":[157],"surpasses":[158],"full":[159],"about":[167],"5\\%":[168],"parameters.":[170]},"counts_by_year":[],"updated_date":"2026-11-03T10:50:19.383937","created_date":"2026-10-31T00:00:00"}