{"id":"https://openalex.org/W7165746058","doi":"https://doi.org/10.48550/arxiv.2606.24774","title":"Revealing Training Data Exposure in Vision Language Large Models via Parameter Gradients","display_name":"Revealing Training Data Exposure in Vision Language Large Models via Parameter Gradients","publication_year":2026,"publication_date":"2026-06-23","ids":{"openalex":"https://openalex.org/W7165746058","doi":"https://doi.org/10.48550/arxiv.2606.24774"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.24774","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24774","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.2606.24774","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139284049","display_name":"Zhihao Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Zhihao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067831144","display_name":"Hongyi Tang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tang, Hongyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139265003","display_name":"Yi Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Yi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5016943900","display_name":"Ahmed Abbasi","orcid":"https://orcid.org/0000-0001-7698-7794"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Abbasi, Ahmed","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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.15919999778270721,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.15919999778270721,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.09610000252723694,"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":0.08190000057220459,"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/discriminative-model","display_name":"Discriminative model","score":0.7236999869346619},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.5996999740600586},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.567799985408783},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.4796000123023987},{"id":"https://openalex.org/keywords/modality","display_name":"Modality (human\u2013computer interaction)","score":0.4641999900341034},{"id":"https://openalex.org/keywords/modalities","display_name":"Modalities","score":0.40860000252723694},{"id":"https://openalex.org/keywords/data-modeling","display_name":"Data modeling","score":0.3792000114917755},{"id":"https://openalex.org/keywords/audit","display_name":"Audit","score":0.37599998712539673}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7549999952316284},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.7236999869346619},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6176000237464905},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.5996999740600586},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.567799985408783},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.4796000123023987},{"id":"https://openalex.org/C2780226545","wikidata":"https://www.wikidata.org/wiki/Q6888030","display_name":"Modality (human\u2013computer interaction)","level":2,"score":0.4641999900341034},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.45899999141693115},{"id":"https://openalex.org/C2779903281","wikidata":"https://www.wikidata.org/wiki/Q6888026","display_name":"Modalities","level":2,"score":0.40860000252723694},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.3792000114917755},{"id":"https://openalex.org/C199521495","wikidata":"https://www.wikidata.org/wiki/Q181487","display_name":"Audit","level":2,"score":0.37599998712539673},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.361299991607666},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.31929999589920044},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.29179999232292175},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2842999994754791},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2786000072956085},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2732999920845032},{"id":"https://openalex.org/C123201435","wikidata":"https://www.wikidata.org/wiki/Q456632","display_name":"Information privacy","level":2,"score":0.2590000033378601},{"id":"https://openalex.org/C145912823","wikidata":"https://www.wikidata.org/wiki/Q113558","display_name":"Dynamics (music)","level":2,"score":0.25380000472068787},{"id":"https://openalex.org/C2776145971","wikidata":"https://www.wikidata.org/wiki/Q30673951","display_name":"Labeled data","level":2,"score":0.25119999051094055},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.25}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.24774","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24774","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.2606.24774","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.24774","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":[{"id":"https://metadata.un.org/sdg/10","display_name":"Reduced inequalities","score":0.7375853657722473}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Vision-Language":[0],"Large":[1],"Models":[2],"(VLLMs)":[3],"trained":[4],"on":[5,47,77,88,97],"massive":[6],"crawled":[7],"corpora":[8],"raise":[9],"pressing":[10],"copyright":[11],"and":[12,93,102,113,130,141,180],"data-provenance":[13],"concerns.":[14],"These":[15],"concerns":[16],"are":[17],"particularly":[18],"acute":[19],"in":[20,42,138],"healthcare,":[21],"where":[22,86],"patient":[23],"medical":[24,129],"images":[25],"paired":[26],"with":[27,51],"clinical":[28],"reports":[29],"demand":[30],"rigorous":[31],"privacy":[32],"safeguards.":[33],"However,":[34],"existing":[35,154],"training":[36,89,155],"data":[37,156,166],"detection":[38,157],"methods":[39,158],"either":[40],"fail":[41],"cross-modal":[43],"scenarios":[44],"or":[45],"rely":[46],"superficial":[48],"output":[49],"signals":[50],"insufficient":[52],"discriminative":[53],"power.":[54],"We":[55],"introduce":[56],"GradAudit,":[57],"a":[58,78,145],"gradient-based":[59],"auditing":[60],"framework":[61],"that":[62,153,169],"examines":[63],"internal":[64],"optimization":[65],"dynamics":[66],"rather":[67],"than":[68],"treating":[69],"VLLMs":[70],"as":[71,175],"black":[72],"boxes.":[73],"Our":[74],"approach":[75],"builds":[76],"key":[79],"observation:":[80],"model":[81],"parameters":[82],"converge":[83],"to":[84],"regions":[85],"gradients":[87,96],"samples":[90,99],"become":[91,177],"stable":[92],"well-aligned,":[94],"whereas":[95],"non-training":[98],"remain":[100],"noisy":[101],"inconsistent.":[103],"By":[104],"analyzing":[105],"these":[106],"gradient":[107],"signatures,":[108],"GradAudit":[109,133],"achieves":[110],"strong":[111],"separability":[112],"detects":[114],"genuine":[115],"image-text":[116],"associations":[117],"learned":[118],"during":[119],"training,":[120],"not":[121,159],"merely":[122],"individual":[123],"modality":[124],"membership.":[125],"Empirically,":[126],"across":[127],"both":[128,139],"general-domain":[131],"datasets,":[132],"substantially":[134],"outperforms":[135],"state-of-the-art":[136],"baselines":[137],"pretraining":[140],"fine-tuning":[142],"VLLMs.":[143],"In":[144],"case":[146],"study":[147],"employing":[148],"copyrighted":[149],"content,":[150],"we":[151],"show":[152],"only":[160],"underestimate":[161],"the":[162],"extent":[163],"of":[164],"unauthorized":[165],"usage,":[167],"but":[168],"this":[170],"underestimation":[171],"becomes":[172],"more":[173,178,181],"pronounced":[174],"models":[176],"recent":[179],"advanced.":[182]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-25T00:00:00"}