{"id":"https://openalex.org/W7165672336","doi":"https://doi.org/10.48550/arxiv.2606.23313","title":"Uncertainty-based Debiasing and Unlearning for Decontamination","display_name":"Uncertainty-based Debiasing and Unlearning for Decontamination","publication_year":2026,"publication_date":"2026-06-22","ids":{"openalex":"https://openalex.org/W7165672336","doi":"https://doi.org/10.48550/arxiv.2606.23313"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.23313","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.23313","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.2606.23313","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5027163093","display_name":"G . X Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Guangzhi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5097391710","display_name":"\u6653\u6167 \u6218","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhan, Xiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5121270832","display_name":"Mark Gales","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gales, Mark","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/T10028","display_name":"Topic Modeling","score":0.3725999891757965,"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/T10028","display_name":"Topic Modeling","score":0.3725999891757965,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.1200999990105629,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.08219999819993973,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/debiasing","display_name":"Debiasing","score":0.8533999919891357},{"id":"https://openalex.org/keywords/human-decontamination","display_name":"Human decontamination","score":0.7476999759674072},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.6690999865531921},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.4465999901294708},{"id":"https://openalex.org/keywords/contamination","display_name":"Contamination","score":0.3598000109195709},{"id":"https://openalex.org/keywords/prior-probability","display_name":"Prior probability","score":0.3465999960899353}],"concepts":[{"id":"https://openalex.org/C2779458634","wikidata":"https://www.wikidata.org/wiki/Q24963715","display_name":"Debiasing","level":2,"score":0.8533999919891357},{"id":"https://openalex.org/C107211472","wikidata":"https://www.wikidata.org/wiki/Q284486","display_name":"Human decontamination","level":2,"score":0.7476999759674072},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.6690999865531921},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6272000074386597},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.4465999901294708},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3626999855041504},{"id":"https://openalex.org/C112570922","wikidata":"https://www.wikidata.org/wiki/Q60528603","display_name":"Contamination","level":2,"score":0.3598000109195709},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.3465999960899353},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.3409000039100647},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.3359000086784363},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.30799999833106995},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.29760000109672546},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.27810001373291016},{"id":"https://openalex.org/C177212765","wikidata":"https://www.wikidata.org/wiki/Q627335","display_name":"Workflow","level":2,"score":0.2540999948978424},{"id":"https://openalex.org/C200601418","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability engineering","level":1,"score":0.25360000133514404}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.23313","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.23313","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.2606.23313","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.23313","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Benchmark-based":[0],"evaluation":[1,56],"is":[2,132,154],"the":[3,75,103,142],"dominant":[4],"paradigm":[5],"for":[6,58,170],"assessing":[7],"large":[8],"language":[9],"model":[10,39,73,81,105,114,197,205],"(LLM)":[11],"capabilities,":[12],"yet":[13],"data":[14],"contamination":[15],"inflates":[16],"reported":[17],"performance":[18,206],"and":[19,41,177],"undermines":[20],"fair":[21],"comparison.":[22],"Existing":[23],"decontamination":[24,59],"methods":[25,97],"are":[26,120],"evaluated":[27],"solely":[28],"through":[29],"aggregate":[30],"accuracy,":[31],"which":[32,118,131],"can":[33],"obscure":[34],"substantial":[35],"differences":[36],"in":[37],"per-sample":[38,108,125,187],"behaviour,":[40],"many":[42],"require":[43],"access":[44],"to":[45,106,134,192],"an":[46,79,195],"uncontaminated":[47,80,113,196,208],"model.":[48],"In":[49],"this":[50,87],"paper,":[51],"we":[52,89],"propose":[53],"a":[54,71,94,112,124,136,159,166],"sample-level":[55],"framework":[57],"that":[60,98,140,184],"complements":[61],"accuracy-based":[62],"assessment":[63],"with":[64],"distributional":[65],"distance":[66],"metrics,":[67],"measuring":[68],"how":[69],"closely":[70],"decontaminated":[72],"recovers":[74],"output":[76,161,188],"distribution":[77,139],"of":[78,96,102,117,194],"on":[82,86,146,175,207],"each":[83],"sample.":[84],"Building":[85],"framework,":[88],"introduce":[90],"Uncertainty-Based":[91],"Decontamination":[92],"(UBD),":[93],"family":[95],"leverage":[99],"deep":[100],"ensembles":[101],"contaminated":[104],"estimate":[107],"memorization":[109],"without":[110],"requiring":[111],"or":[115,164,200],"knowledge":[116],"samples":[119],"contaminated.":[121],"UBD":[122,185],"estimates":[123],"correction":[126,162],"scalar":[127],"from":[128],"ensemble":[129],"uncertainty,":[130],"used":[133,156],"construct":[135],"debiased":[137],"target":[138,153],"suppresses":[141],"inflated":[143],"probability":[144],"mass":[145],"correct":[147],"answers":[148],"induced":[149],"by":[150],"contamination.":[151],"This":[152],"then":[155],"either":[157],"as":[158,165],"post-hoc":[160],"(debiasing)":[163],"soft":[167],"training":[168],"signal":[169],"parameter":[171],"update":[172],"(unlearning).":[173],"Experiments":[174],"MMLU-Pro":[176],"MATH-MCQA":[178],"across":[179],"multiple":[180],"LLM":[181],"backbones":[182],"demonstrate":[183],"produces":[186],"distributions":[189],"substantially":[190],"closer":[191],"those":[193],"than":[198],"paraphrasing":[199],"choice-permutation":[201],"baselines,":[202],"while":[203],"preserving":[204],"data.":[209]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-24T00:00:00"}