{"id":"https://openalex.org/W7137856705","doi":"https://doi.org/10.1609/aaai.v40i11.37886","title":"Difference Vector Equalization for Robust Fine-tuning of Vision-Language Models","display_name":"Difference Vector Equalization for Robust Fine-tuning of Vision-Language Models","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7137856705","doi":"https://doi.org/10.1609/aaai.v40i11.37886"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i11.37886","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i11.37886","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://doi.org/10.1609/aaai.v40i11.37886","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5027868267","display_name":"Satoshi Suzuki","orcid":"https://orcid.org/0000-0001-5343-4660"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Satoshi Suzuki","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5081715626","display_name":"Shin\u2019ya Yamaguchi","orcid":"https://orcid.org/0000-0001-9113-7405"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shin'ya Yamaguchi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006634833","display_name":"Shoichiro Takeda","orcid":"https://orcid.org/0000-0002-7138-4983"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shoichiro Takeda","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102511251","display_name":"Taiga Yamane","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Taiga Yamane","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082466886","display_name":"Naoki Makishima","orcid":"https://orcid.org/0000-0002-7065-315X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Naoki Makishima","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5016817412","display_name":"Naotaka Kawata","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Naotaka Kawata","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027118468","display_name":"Mana Ihori","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mana Ihori","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129667214","display_name":"Tomohiro Tanaka","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tomohiro Tanaka","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013009514","display_name":"Shota Orihashi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shota Orihashi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5060644399","display_name":"Ryo Masumura","orcid":"https://orcid.org/0000-0002-2415-4149"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ryo Masumura","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":10,"corresponding_author_ids":["https://openalex.org/A5027868267"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.07796853,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"11","first_page":"9278","last_page":"9286"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.4683000147342682,"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"}},"topics":[{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.4683000147342682,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.3723999857902527,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.0203000009059906,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.7699999809265137},{"id":"https://openalex.org/keywords/pairwise-comparison","display_name":"Pairwise comparison","score":0.5210999846458435},{"id":"https://openalex.org/keywords/geometric-shape","display_name":"Geometric shape","score":0.41370001435279846},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.3732999861240387},{"id":"https://openalex.org/keywords/current","display_name":"Current (fluid)","score":0.3603000044822693},{"id":"https://openalex.org/keywords/geometric-modeling","display_name":"Geometric modeling","score":0.3598000109195709},{"id":"https://openalex.org/keywords/image","display_name":"Image (mathematics)","score":0.35670000314712524}],"concepts":[{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.7699999809265137},{"id":"https://openalex.org/C184898388","wikidata":"https://www.wikidata.org/wiki/Q1435712","display_name":"Pairwise comparison","level":2,"score":0.5210999846458435},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.48890000581741333},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.47279998660087585},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4480000138282776},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.44679999351501465},{"id":"https://openalex.org/C7305733","wikidata":"https://www.wikidata.org/wiki/Q207961","display_name":"Geometric shape","level":2,"score":0.41370001435279846},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3732999861240387},{"id":"https://openalex.org/C148043351","wikidata":"https://www.wikidata.org/wiki/Q4456944","display_name":"Current (fluid)","level":2,"score":0.3603000044822693},{"id":"https://openalex.org/C104065381","wikidata":"https://www.wikidata.org/wiki/Q1002535","display_name":"Geometric modeling","level":2,"score":0.3598000109195709},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.35670000314712524},{"id":"https://openalex.org/C136520226","wikidata":"https://www.wikidata.org/wiki/Q302814","display_name":"Geometric data analysis","level":2,"score":0.3447999954223633},{"id":"https://openalex.org/C75755367","wikidata":"https://www.wikidata.org/wiki/Q104531076","display_name":"Equalization (audio)","level":3,"score":0.295199990272522},{"id":"https://openalex.org/C32990609","wikidata":"https://www.wikidata.org/wiki/Q306542","display_name":"Transformation geometry","level":2,"score":0.2865000069141388},{"id":"https://openalex.org/C118965365","wikidata":"https://www.wikidata.org/wiki/Q44528","display_name":"Euclidean vector","level":2,"score":0.26409998536109924},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.26249998807907104},{"id":"https://openalex.org/C39947850","wikidata":"https://www.wikidata.org/wiki/Q729523","display_name":"Geometric distribution","level":3,"score":0.25920000672340393}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i11.37886","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i11.37886","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i11.37886","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i11.37886","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Contrastive":[0],"pre-trained":[1,133],"vision-language":[2,30,87],"models,":[3,88],"such":[4],"as":[5],"CLIP,":[6],"demonstrate":[7,205],"strong":[8,214],"generalization":[9,39,85],"abilities":[10,40],"in":[11,41,61,83,90],"zero-shot":[12,45,94,220],"classification":[13],"by":[14,54,126,180,197],"leveraging":[15],"embeddings":[16,129],"extracted":[17,130],"from":[18,131],"image":[19],"and":[20,44,93,134,169,219],"text":[21],"encoders.":[22],"This":[23],"paper":[24],"aims":[25],"to":[26,117,147,184,187],"robustly":[27],"fine-tune":[28],"these":[29,69],"models":[31,136],"on":[32],"in-distribution":[33],"(ID)":[34],"data":[35,140,152],"without":[36],"compromising":[37],"their":[38,188],"out-of-distribution":[42],"(OOD)":[43],"settings.":[46],"Current":[47],"robust":[48],"fine-tuning":[49,135],"methods":[50,70],"tackle":[51],"this":[52],"challenge":[53],"reusing":[55],"contrastive":[56],"learning,":[57],"which":[58,78,105,123],"was":[59],"used":[60],"pre-training,":[62],"for":[63,137],"fine-tuning.":[64,111],"However,":[65],"we":[66,99,154,161],"found":[67],"that":[68,206],"distort":[71],"the":[72,76,84,107,128,132,138,144,157,176,193,210],"geometric":[73,108,158,177,194,211],"structure":[74,109,178,195],"of":[75,86,122],"embeddings,":[77],"plays":[79],"a":[80,199],"crucial":[81],"role":[82],"resulting":[89],"limited":[91],"OOD":[92],"performance.":[95],"To":[96],"address":[97],"this,":[98],"propose":[100],"Difference":[101],"Vector":[102],"Equalization":[103],"(DiVE),":[104],"preserves":[106,175,192,209],"during":[110],"The":[112],"idea":[113],"behind":[114],"DiVE":[115,207],"is":[116,124],"constrain":[118],"difference":[119,145,182],"vectors,":[120],"each":[121],"obtained":[125],"subtracting":[127],"same":[139],"sample.":[141],"By":[142],"constraining":[143,181],"vectors":[146,183],"be":[148,185],"equal":[149,186],"across":[150,216],"various":[151],"samples,":[153],"effectively":[155,208],"preserve":[156],"structure.":[159],"Therefore,":[160],"introduce":[162],"two":[163],"losses:":[164],"average":[165],"vector":[166,171],"loss":[167,172],"(AVL)":[168],"pairwise":[170],"(PVL).":[173],"AVL":[174],"globally":[179],"weighted":[189],"average.":[190],"PVL":[191],"locally":[196],"ensuring":[198],"consistent":[200],"multimodal":[201],"alignment.":[202],"Our":[203],"experiments":[204],"structure,":[212],"achieving":[213],"results":[215],"ID,":[217],"OOD,":[218],"metrics.":[221]},"counts_by_year":[],"updated_date":"2026-05-21T06:26:12.895304","created_date":"2026-03-18T00:00:00"}