{"id":"https://openalex.org/W7162663277","doi":"https://doi.org/10.48550/arxiv.2605.28780","title":"Bias Leaves a Gradient Trail: Label-Free Bias Identification via Gradient Probes on Concept Decompositions","display_name":"Bias Leaves a Gradient Trail: Label-Free Bias Identification via Gradient Probes on Concept Decompositions","publication_year":2026,"publication_date":"2026-05-27","ids":{"openalex":"https://openalex.org/W7162663277","doi":"https://doi.org/10.48550/arxiv.2605.28780"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.28780","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.28780","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.2605.28780","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137240884","display_name":"Thomas Vitry","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Vitry, Thomas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137305051","display_name":"Kieran Edgeworth","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Edgeworth, Kieran","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137199761","display_name":"Stefan Wermter","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wermter, Stefan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137282494","display_name":"Jae Hee Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Jae Hee","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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.6941999793052673,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.6941999793052673,"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/T11652","display_name":"Imbalanced Data Classification Techniques","score":0.07339999824762344,"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":0.03280000016093254,"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/debiasing","display_name":"Debiasing","score":0.8791000247001648},{"id":"https://openalex.org/keywords/spurious-relationship","display_name":"Spurious relationship","score":0.8669999837875366},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5580999851226807},{"id":"https://openalex.org/keywords/identification","display_name":"Identification (biology)","score":0.5411999821662903},{"id":"https://openalex.org/keywords/class","display_name":"Class (philosophy)","score":0.517799973487854},{"id":"https://openalex.org/keywords/estimator","display_name":"Estimator","score":0.5149000287055969},{"id":"https://openalex.org/keywords/filter","display_name":"Filter (signal processing)","score":0.4733000099658966},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.39320001006126404},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.38449999690055847}],"concepts":[{"id":"https://openalex.org/C2779458634","wikidata":"https://www.wikidata.org/wiki/Q24963715","display_name":"Debiasing","level":2,"score":0.8791000247001648},{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.8669999837875366},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6123999953269958},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5580999851226807},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5566999912261963},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.5411999821662903},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.517799973487854},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.5149000287055969},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.4733000099658966},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4075999855995178},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.39320001006126404},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.38449999690055847},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3555000126361847},{"id":"https://openalex.org/C55166926","wikidata":"https://www.wikidata.org/wiki/Q2892946","display_name":"Oracle","level":2,"score":0.3472000062465668},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.34689998626708984},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.33889999985694885},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.33500000834465027},{"id":"https://openalex.org/C75553542","wikidata":"https://www.wikidata.org/wiki/Q178161","display_name":"A priori and a posteriori","level":2,"score":0.32989999651908875},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.3264999985694885},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3240000009536743},{"id":"https://openalex.org/C187834632","wikidata":"https://www.wikidata.org/wiki/Q188804","display_name":"Factorization","level":2,"score":0.322299987077713},{"id":"https://openalex.org/C2780695315","wikidata":"https://www.wikidata.org/wiki/Q3799040","display_name":"Unobservable","level":2,"score":0.3102000057697296},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.298799991607666},{"id":"https://openalex.org/C106487976","wikidata":"https://www.wikidata.org/wiki/Q685816","display_name":"Matrix (chemical analysis)","level":2,"score":0.29750001430511475},{"id":"https://openalex.org/C2777402240","wikidata":"https://www.wikidata.org/wiki/Q6783436","display_name":"Masking (illustration)","level":2,"score":0.2973000109195709},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.2883000075817108},{"id":"https://openalex.org/C75917345","wikidata":"https://www.wikidata.org/wiki/Q2725298","display_name":"Sampling bias","level":3,"score":0.27140000462532043},{"id":"https://openalex.org/C40423286","wikidata":"https://www.wikidata.org/wiki/Q284172","display_name":"Selection bias","level":2,"score":0.25699999928474426},{"id":"https://openalex.org/C136536468","wikidata":"https://www.wikidata.org/wiki/Q1225894","display_name":"Undersampling","level":2,"score":0.2554999887943268},{"id":"https://openalex.org/C190502265","wikidata":"https://www.wikidata.org/wiki/Q17069496","display_name":"MNIST database","level":3,"score":0.2547999918460846}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.28780","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.28780","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.2605.28780","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.28780","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/5","display_name":"Gender equality","score":0.7401597499847412}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Vision":[0],"classifiers":[1],"can":[2],"exploit":[3],"spurious":[4,56,150,200],"correlations,":[5],"achieving":[6],"high":[7],"in-distribution":[8],"accuracy":[9,177],"yet":[10],"failing":[11],"under":[12],"distribution":[13],"shift.":[14],"Existing":[15],"approaches":[16],"to":[17,91,94,124,180],"bias":[18,45,109,121],"mitigation":[19],"and":[20,86,131,140,152,186,215],"analysis":[21],"often":[22],"depend":[23],"on":[24,64,118,153,184,188],"curated":[25],"datasets,":[26],"spurious-attribute":[27],"or":[28,31,42,193],"group":[29],"labels,":[30],"retraining,":[32],"which":[33],"may":[34],"be":[35],"infeasible":[36],"once":[37],"a":[38,50,69,96,108],"model":[39],"is":[40,46,225],"deployed":[41],"the":[43,142,148,164,169],"relevant":[44],"unknown.":[47],"We":[48],"present":[49],"bias-label-free,":[51],"post-hoc":[52],"method":[53,143,197],"for":[54,220],"identifying":[55],"concepts":[57,103,122,145,171],"in":[58],"frozen":[59,221],"vision":[60,222],"models,":[61],"relying":[62],"only":[63,160],"standard":[65],"class":[66,85],"labels":[67],"from":[68,80,112],"held-out":[70],"audit":[71],"dataset.":[72],"For":[73],"each":[74],"target":[75],"class,":[76],"we":[77],"collect":[78],"patches":[79],"inputs":[81],"predicted":[82],"as":[83],"that":[84,159,202],"apply":[87],"non-negative":[88],"matrix":[89],"factorization":[90],"intermediate":[92],"activations":[93],"obtain":[95],"bank":[97],"of":[98],"interpretable":[99,212],"concept":[100],"vectors.":[101],"Candidate":[102],"are":[104],"then":[105],"ranked":[106],"with":[107,115,147,163,206],"estimator":[110],"derived":[111],"their":[113],"interaction":[114],"backpropagated":[116],"gradients":[117],"misclassified":[119],"examples:":[120],"tend":[123],"get":[125],"activated":[126],"when":[127,133],"correcting":[128,134],"false":[129,135],"negatives":[130],"suppressed":[132],"positives.":[136],"On":[137],"Colored":[138],"MNIST":[139],"Waterbirds":[141,185],"recovers":[144],"aligned":[146],"known":[149],"cue,":[151],"CelebA":[154,189],"it":[155],"surfaces":[156],"decision-relevant":[157,199],"directions":[158,201],"partially":[161],"coincide":[162,205],"annotated":[165,207],"gender":[166],"attribute;":[167],"suppressing":[168],"top-ranked":[170],"at":[172,227],"inference":[173],"time":[174],"improves":[175],"worst-group":[176],"by":[178],"up":[179],"17.9":[181],"percentage":[182],"points":[183],"10.4":[187],"without":[190],"any":[191],"retraining":[192],"parameter":[194],"updates.":[195],"Our":[196],"identifies":[198],"need":[203],"not":[204],"ones,":[208],"providing":[209],"both":[210],"an":[211,216],"auditing":[213],"tool":[214],"actionable":[217],"debiasing":[218],"handle":[219],"models.":[223],"Code":[224],"available":[226],"https://github.com/vitryt/label-free-bias-identification.":[228]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-29T00:00:00"}