{"id":"https://openalex.org/W7134807015","doi":"https://doi.org/10.48550/arxiv.2603.08309","title":"Concept-Guided Fine-Tuning: Steering ViTs away from Spurious Correlations to Improve Robustness","display_name":"Concept-Guided Fine-Tuning: Steering ViTs away from Spurious Correlations to Improve Robustness","publication_year":2026,"publication_date":"2026-03-09","ids":{"openalex":"https://openalex.org/W7134807015","doi":"https://doi.org/10.48550/arxiv.2603.08309"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2603.08309","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"type":"article","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5128660095","display_name":"Yehonatan Elisha","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Elisha, Yehonatan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077269072","display_name":"Oren Barkan","orcid":"https://orcid.org/0000-0002-5059-0905"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Barkan, Oren","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5083002103","display_name":"Noam Koenigstein","orcid":"https://orcid.org/0000-0001-8219-4512"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Koenigstein, Noam","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":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.27674992,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"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/T11605","display_name":"Visual Attention and Saliency Detection","score":0.2678999900817871,"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/T11605","display_name":"Visual Attention and Saliency Detection","score":0.2678999900817871,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.2085999995470047,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.1266999989748001,"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/spurious-relationship","display_name":"Spurious relationship","score":0.9078999757766724},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.7699999809265137},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.4837999939918518},{"id":"https://openalex.org/keywords/segmentation","display_name":"Segmentation","score":0.47999998927116394},{"id":"https://openalex.org/keywords/relevance","display_name":"Relevance (law)","score":0.3124000132083893},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.3089999854564667},{"id":"https://openalex.org/keywords/path","display_name":"Path (computing)","score":0.289000004529953}],"concepts":[{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.9078999757766724},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.7699999809265137},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7268999814987183},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5097000002861023},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.4837999939918518},{"id":"https://openalex.org/C89600930","wikidata":"https://www.wikidata.org/wiki/Q1423946","display_name":"Segmentation","level":2,"score":0.47999998927116394},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3995000123977661},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3700999915599823},{"id":"https://openalex.org/C158154518","wikidata":"https://www.wikidata.org/wiki/Q7310970","display_name":"Relevance (law)","level":2,"score":0.3124000132083893},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.31040000915527344},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3095000088214874},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.3089999854564667},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.289000004529953},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.2840000092983246},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.28029999136924744},{"id":"https://openalex.org/C124504099","wikidata":"https://www.wikidata.org/wiki/Q56933","display_name":"Image segmentation","level":3,"score":0.2685999870300293},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.26649999618530273},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.26249998807907104},{"id":"https://openalex.org/C2776151529","wikidata":"https://www.wikidata.org/wiki/Q3045304","display_name":"Object detection","level":3,"score":0.25769999623298645}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2603.08309","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2603.08309","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.08309","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,"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":"pmh:doi:10.48550/arxiv.2603.08309","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"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":{"Vision":[0],"Transformers":[1],"(ViTs)":[2],"often":[3],"degrade":[4],"under":[5],"distribution":[6,61],"shifts":[7],"because":[8],"they":[9],"rely":[10],"on":[11,28,132,155],"spurious":[12,133],"correlations,":[13],"such":[14],"as":[15],"background":[16,134],"cues,":[17],"rather":[18],"than":[19,208],"semantically":[20],"meaningful":[21],"features.":[22],"Existing":[23],"regularization":[24],"methods,":[25],"typically":[26],"relying":[27],"simple":[29],"foreground-background":[30],"masks,":[31],"which":[32],"fail":[33],"to":[34,60,88],"capture":[35],"the":[36,83,150,173],"fine-grained":[37],"semantic":[38,181],"concepts":[39,104],"that":[40,73,160,172,198],"define":[41],"an":[42,109],"object":[43,182],"(e.g.,":[44],"``long":[45],"beak''":[46],"and":[47,113,146,191],"``wings''":[48],"for":[49,205],"a":[50,53,69,117,141,185],"``bird'').":[51],"As":[52],"result,":[54],"these":[55,125],"methods":[56],"provide":[57,201],"limited":[58],"robustness":[59,164,207],"shifts.":[62],"To":[63],"address":[64],"this":[65,137],"limitation,":[66],"we":[67,170,196],"introduce":[68],"novel":[70],"finetuning":[71,120],"framework":[72],"steers":[74],"model":[75,206],"reasoning":[76],"toward":[77,188],"concept-level":[78],"semantics.":[79],"Our":[80],"approach":[81],"optimizes":[82],"model's":[84],"internal":[85],"relevance":[86,123,175],"maps":[87,176],"align":[89],"with":[90,124,180],"spatially":[91],"grounded":[92],"concept":[93,126],"masks.":[94],"These":[95],"masks":[96,200],"are":[97,105],"generated":[98],"automatically,":[99],"without":[100],"manual":[101],"annotation:":[102],"class-relevant":[103],"first":[106],"proposed":[107],"using":[108,116],"LLM-based,":[110],"label-free":[111],"method,":[112],"then":[114],"segmented":[115],"VLM.":[118],"The":[119],"objective":[121],"aligns":[122],"regions":[127],"while":[128],"simultaneously":[129],"suppressing":[130],"focus":[131],"areas.":[135],"Notably,":[136],"process":[138],"requires":[139],"only":[140],"minimal":[142],"set":[143],"of":[144,149],"images":[145],"uses":[147],"half":[148],"dataset":[151],"classes.":[152],"Extensive":[153],"experiments":[154],"five":[156],"out-of-distribution":[157],"benchmarks":[158],"demonstrate":[159],"our":[161,213],"method":[162],"improves":[163],"across":[165],"multiple":[166],"ViT-based":[167],"models.":[168,194],"Furthermore,":[169],"show":[171],"resulting":[174],"exhibit":[177],"stronger":[178],"alignment":[179],"parts,":[183],"offering":[184],"scalable":[186],"path":[187],"more":[189,202],"robust":[190],"interpretable":[192],"vision":[193],"Finally,":[195],"confirm":[197],"concept-guided":[199],"effective":[203],"supervision":[204],"conventional":[209],"segmentation":[210],"maps,":[211],"supporting":[212],"central":[214],"hypothesis.":[215]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-03-11T00:00:00"}