{"id":"https://openalex.org/W7161746848","doi":"https://doi.org/10.48550/arxiv.2605.17432","title":"DP-SelFT: Differentially Private Selective Fine-Tuning for Large Language Models","display_name":"DP-SelFT: Differentially Private Selective Fine-Tuning for Large Language Models","publication_year":2026,"publication_date":"2026-05-17","ids":{"openalex":"https://openalex.org/W7161746848","doi":"https://doi.org/10.48550/arxiv.2605.17432"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.17432","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.17432","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":null,"license_id":null,"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.2605.17432","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5013148560","display_name":"Haichao Sha","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sha, Haichao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136503742","display_name":"Zihao Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zihao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101635565","display_name":"Yuncheng Wu","orcid":"https://orcid.org/0000-0002-1436-3916"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Yuncheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136503675","display_name":"hong Chen","orcid":"https://orcid.org/0000-0003-2207-5655"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Hong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136473732","display_name":"Wei Dong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong, Wei","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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.7117000222206116,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.7117000222206116,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.03799999877810478,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.032499998807907104,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/differential-privacy","display_name":"Differential privacy","score":0.7178999781608582},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.541700005531311},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.5088000297546387},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5029000043869019},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.48010000586509705},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.4731999933719635},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.4453999996185303},{"id":"https://openalex.org/keywords/synthetic-data","display_name":"Synthetic data","score":0.41679999232292175},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.39419999718666077}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7620000243186951},{"id":"https://openalex.org/C23130292","wikidata":"https://www.wikidata.org/wiki/Q5275358","display_name":"Differential privacy","level":2,"score":0.7178999781608582},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.541700005531311},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.5088000297546387},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5029000043869019},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.48010000586509705},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.4731999933719635},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.47119998931884766},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.4453999996185303},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.44179999828338623},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.4221999943256378},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.41679999232292175},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.39419999718666077},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.35530000925064087},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.34619998931884766},{"id":"https://openalex.org/C99221444","wikidata":"https://www.wikidata.org/wiki/Q1532069","display_name":"Private information retrieval","level":2,"score":0.3425000011920929},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.33559998869895935},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3278000056743622},{"id":"https://openalex.org/C123201435","wikidata":"https://www.wikidata.org/wiki/Q456632","display_name":"Information privacy","level":2,"score":0.32690000534057617},{"id":"https://openalex.org/C2780909371","wikidata":"https://www.wikidata.org/wiki/Q4801092","display_name":"Artificial noise","level":4,"score":0.31700000166893005},{"id":"https://openalex.org/C2781170535","wikidata":"https://www.wikidata.org/wiki/Q30587856","display_name":"Noisy data","level":2,"score":0.314300000667572},{"id":"https://openalex.org/C2776848632","wikidata":"https://www.wikidata.org/wiki/Q853463","display_name":"Clipping (morphology)","level":2,"score":0.3102000057697296},{"id":"https://openalex.org/C2779227376","wikidata":"https://www.wikidata.org/wiki/Q6505497","display_name":"Layer (electronics)","level":2,"score":0.30970001220703125},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.3075000047683716},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.27889999747276306},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2614000141620636},{"id":"https://openalex.org/C29265498","wikidata":"https://www.wikidata.org/wiki/Q7047719","display_name":"Noise measurement","level":3,"score":0.2554999887943268}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.17432","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.17432","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.17432","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.17432","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16","score":0.5187333822250366}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"are":[4,89,210],"commonly":[5],"adapted":[6],"to":[7,48,191,217],"downstream":[8,192],"tasks":[9,224],"through":[10],"fine-tuning,":[11,84,194],"but":[12,214],"fine-tuning":[13,38,64,100,236],"data":[14],"often":[15],"contains":[16],"sensitive":[17],"information":[18],"that":[19,123,209,226],"may":[20],"be":[21],"leaked":[22],"by":[23,59,161],"the":[24,71,148,199,230,239],"resulting":[25],"model.":[26],"Differential":[27],"privacy":[28,113,154,241],"(DP)":[29],"offers":[30],"formal":[31],"protection":[32],"against":[33],"such":[34,66],"leakage,":[35],"yet":[36],"DP":[37,61,136,193,203,235],"of":[39,73,101,198],"LLMs":[40],"still":[41],"suffers":[42],"from":[43],"substantial":[44],"utility":[45],"degradation":[46],"due":[47],"gradient":[49],"clipping":[50],"and":[51,120,128,139,172],"noise":[52],"injection.":[53],"Existing":[54],"work":[55],"improves":[56,229],"this":[57,76,144,181],"trade-off":[58,232],"combining":[60],"with":[62,195],"parameter-efficient":[63],"methods":[65],"as":[67,202],"LoRA,":[68],"which":[69,85],"constrain":[70],"form":[72],"updates.":[74,130,220],"In":[75],"work,":[77],"we":[78],"study":[79],"a":[80,94,134,168,176,187],"complementary":[81],"direction:":[82],"selective":[83,99],"constrains":[86],"where":[87],"updates":[88],"applied.":[90],"We":[91],"propose":[92],"DP-SelFT,":[93],"framework":[95],"for":[96],"differentially":[97],"private":[98,219],"LLMs.":[102],"DP-SelFT":[103,227],"addresses":[104],"three":[105],"DP-specific":[106],"challenges":[107],"in":[108],"parameter":[109],"selection:":[110],"avoiding":[111],"repeated":[112],"cost,":[114],"improving":[115],"stability":[116],"under":[117,126,186,238],"noisy":[118,129,218],"estimates,":[119],"selecting":[121],"parameters":[122],"remain":[124],"useful":[125],"clipped":[127],"It":[131,156],"first":[132],"constructs":[133],"lightweight":[135],"synthetic":[137,145,169,177],"dataset":[138],"performs":[140],"selection":[141,149,160],"only":[142,212],"on":[143,167,175,222],"data,":[146],"so":[147],"stage":[150],"incurs":[151],"no":[152],"additional":[153],"cost.":[155],"then":[157],"conducts":[158],"layer-level":[159],"temporarily":[162],"training":[163,170,183],"candidate":[164],"layer":[165,207],"subsets":[166,208],"split":[171],"evaluating":[173],"them":[174],"validation":[178],"split.":[179],"Crucially,":[180],"temporary":[182],"is":[184],"performed":[185],"perturbation":[188],"regime":[189],"matched":[190],"worst-case":[196],"perturbations":[197],"same":[200,240],"scale":[201],"noise.":[204],"This":[205],"favors":[206],"not":[211],"learnable":[213],"also":[215],"robust":[216],"Experiments":[221],"benchmark":[223],"show":[225],"consistently":[228],"privacy--utility":[231],"over":[233],"existing":[234],"baselines":[237],"guarantees.":[242]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-20T00:00:00"}