{"id":"https://openalex.org/W7161764536","doi":"https://doi.org/10.48550/arxiv.2605.19956","title":"Towards Fine-Grained Robustness: Attention-Guided Test-Time Prompt Tuning for Vision-Language Models","display_name":"Towards Fine-Grained Robustness: Attention-Guided Test-Time Prompt Tuning for Vision-Language Models","publication_year":2026,"publication_date":"2026-05-19","ids":{"openalex":"https://openalex.org/W7161764536","doi":"https://doi.org/10.48550/arxiv.2605.19956"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.19956","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19956","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.2605.19956","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136546138","display_name":"Jia-Wei Hai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hai, Jia-Wei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136558963","display_name":"Yijun Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yijun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136540311","display_name":"Xiu-Shen Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Xiu-Shen","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":0.7153000235557556,"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":0.7153000235557556,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.1160999983549118,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.06719999760389328,"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/adversarial-system","display_name":"Adversarial system","score":0.791700005531311},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.738099992275238},{"id":"https://openalex.org/keywords/discriminative-model","display_name":"Discriminative model","score":0.6550999879837036},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6388999819755554},{"id":"https://openalex.org/keywords/adaptation","display_name":"Adaptation (eye)","score":0.5153999924659729},{"id":"https://openalex.org/keywords/domain-adaptation","display_name":"Domain adaptation","score":0.4490000009536743}],"concepts":[{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.791700005531311},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.753600001335144},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.738099992275238},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.6550999879837036},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6388999819755554},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.5153999924659729},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49129998683929443},{"id":"https://openalex.org/C2776434776","wikidata":"https://www.wikidata.org/wiki/Q19246213","display_name":"Domain adaptation","level":3,"score":0.4490000009536743},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.436599999666214},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.415800005197525},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.39719998836517334},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.28220000863075256},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.26420000195503235}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.19956","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19956","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.2605.19956","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19956","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":[{"id":"https://metadata.un.org/sdg/10","score":0.6510270237922668,"display_name":"Reduced inequalities"}],"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],"Models":[1],"(VLMs),":[2],"such":[3],"as":[4],"CLIP,":[5],"have":[6,22],"achieved":[7],"significant":[8],"zero-shot":[9],"performance":[10],"on":[11,48,136],"downstream":[12],"tasks":[13],"with":[14],"various":[15,53],"fine-tuning":[16,54],"adaptation":[17,44,134],"methods.":[18],"However,":[19],"recent":[20],"studies":[21],"proven":[23],"that":[24,129],"adversarial":[25,105,138],"attacks":[26],"can":[27],"significantly":[28],"degrade":[29],"the":[30,93,113],"inference":[31],"ability":[32],"of":[33],"VLMs,":[34],"posing":[35],"substantial":[36],"risks":[37],"to":[38,51,58,65,98,111],"their":[39],"practical":[40],"applications.":[41],"Prevalent":[42],"test-time":[43,88,133],"methods":[45,135],"typically":[46],"rely":[47],"multi-view":[49,119],"augmentation":[50,116],"implement":[52],"strategies,":[55],"which":[56],"struggle":[57],"identify":[59,99],"semantic":[60],"information":[61],"and":[62,118,124,139],"are":[63,143],"prone":[64],"destroying":[66],"discriminative":[67],"regions":[68,102],"in":[69],"fine-grained":[70],"scenarios.":[71],"To":[72],"address":[73],"these":[74],"limitations,":[75],"we":[76,108],"propose":[77],"Attention-Guided":[78],"Test-Time":[79],"Prompt":[80],"Tuning":[81],"(A-TPT),":[82],"a":[83],"semantics-preserving":[84],"method":[85],"designed":[86],"for":[87,121],"adaptation.":[89],"We":[90],"first":[91],"refine":[92],"gradient":[94],"attention":[95],"rollout":[96],"mechanism":[97],"semantically":[100],"meaningful":[101],"surviving":[103],"under":[104],"attacks.":[106],"Furthermore,":[107],"leverage":[109],"them":[110],"guide":[112],"spatially":[114],"varying":[115],"intensities":[117],"ensemble":[120],"prompt":[122],"tuning":[123],"inference.":[125],"Extensive":[126],"experiments":[127],"demonstrate":[128],"A-TPT":[130],"outperforms":[131],"existing":[132],"both":[137],"clean":[140],"data.":[141],"Codes":[142],"available":[144],"at":[145],"https://github.com/SEU-VIPGroup/A-TPT":[146],".":[147]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-21T00:00:00"}