{"id":"https://openalex.org/W7163333503","doi":"https://doi.org/10.48550/arxiv.2606.02920","title":"Fast Unlearning at Scale via Margin Self-Correction","display_name":"Fast Unlearning at Scale via Margin Self-Correction","publication_year":2026,"publication_date":"2026-06-01","ids":{"openalex":"https://openalex.org/W7163333503","doi":"https://doi.org/10.48550/arxiv.2606.02920"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.02920","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.02920","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":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.2606.02920","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137757810","display_name":"Federico Di Gennaro","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Di Gennaro, Federico","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137762941","display_name":"Alexander Shevchenko","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shevchenko, Alexander","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5113814271","display_name":"Fanny Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Fanny","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":0.47450000047683716,"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":0.47450000047683716,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.1753000020980835,"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":0.05040000006556511,"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/security-token","display_name":"Security token","score":0.8325999975204468},{"id":"https://openalex.org/keywords/margin","display_name":"Margin (machine learning)","score":0.8222000002861023},{"id":"https://openalex.org/keywords/sequence","display_name":"Sequence (biology)","score":0.6313999891281128},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.6291000247001648},{"id":"https://openalex.org/keywords/downstream","display_name":"Downstream (manufacturing)","score":0.6259999871253967},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.5212000012397766},{"id":"https://openalex.org/keywords/fraction","display_name":"Fraction (chemistry)","score":0.48980000615119934},{"id":"https://openalex.org/keywords/limit","display_name":"Limit (mathematics)","score":0.48809999227523804}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.8325999975204468},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.8222000002861023},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7609000205993652},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.6313999891281128},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6291000247001648},{"id":"https://openalex.org/C2776207758","wikidata":"https://www.wikidata.org/wiki/Q5303302","display_name":"Downstream (manufacturing)","level":2,"score":0.6259999871253967},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.5212000012397766},{"id":"https://openalex.org/C149629883","wikidata":"https://www.wikidata.org/wiki/Q660926","display_name":"Fraction (chemistry)","level":2,"score":0.48980000615119934},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.48809999227523804},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.4162999987602234},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.4101000130176544},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37310001254081726},{"id":"https://openalex.org/C2779664074","wikidata":"https://www.wikidata.org/wiki/Q3518405","display_name":"Terminal (telecommunication)","level":2,"score":0.3546999990940094},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3400000035762787},{"id":"https://openalex.org/C2777851325","wikidata":"https://www.wikidata.org/wiki/Q7094102","display_name":"Online model","level":2,"score":0.32499998807907104},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.30329999327659607},{"id":"https://openalex.org/C43364308","wikidata":"https://www.wikidata.org/wiki/Q8799","display_name":"Byte","level":2,"score":0.30140000581741333},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2750000059604645},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.27000001072883606},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.2671999931335449},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.2669000029563904},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2574999928474426},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.2508000135421753}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.02920","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.02920","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":"doi:10.48550/arxiv.2606.02920","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.02920","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":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":{"Language-model":[0],"unlearning":[1,85],"updates":[2],"a":[3,32,99,123,134,154,159],"trained":[4],"model":[5,30,40,125,175],"to":[6,102],"behave":[7],"as":[8,172],"if":[9],"it":[10],"had":[11],"not":[12,94],"seen":[13],"selected":[14],"training":[15,34,58],"examples,":[16],"while":[17,195],"preserving":[18],"utility":[19,197],"and":[20,36,64,71,116,149,187],"avoiding":[21],"costly":[22],"retraining.":[23],"Existing":[24],"approaches":[25],"typically":[26],"fine-tune":[27],"the":[28,38,60,108,112,117,162,181,190],"pretrained":[29],"with":[31,87],"fixed":[33],"budget":[35],"select":[37],"final":[39,124],"afterwards":[41],"by":[42],"evaluating":[43],"several":[44],"saved":[45],"checkpoints":[46],"on":[47,131],"downstream":[48,96],"validation":[49],"data.":[50],"Two":[51],"sources":[52],"of":[53,138,161,165,179],"unnecessary":[54],"computation":[55],"limit":[56],"scalability:":[57],"beyond":[59],"desired":[61],"forget-retain":[62,156],"trade-off,":[63],"checkpoint":[65],"selection":[66],"that":[67,92,171],"requires":[68],"extra":[69],"storage":[70],"repeated":[72],"evaluations.":[73],"To":[74],"address":[75],"these":[76],"limitations,":[77],"we":[78,173],"introduce":[79],"MArgin":[80],"Self-Correction":[81],"(MASC),":[82],"an":[83,88],"efficient":[84],"method":[86],"online":[89],"stopping":[90],"rule":[91],"does":[93],"require":[95],"evaluation.":[97],"Given":[98],"text":[100],"sequence":[101],"be":[103],"forgotten,":[104],"MASC":[105,152,186],"actively":[106],"reduces":[107],"logit":[109],"gap":[110,128],"between":[111],"original":[113],"next":[114],"token":[115,139],"most":[118],"likely":[119],"alternatives.":[120],"It":[121],"outputs":[122],"once":[126],"this":[127],"is":[129],"small":[130],"average":[132],"over":[133],"sufficiently":[135],"large":[136],"proportion":[137],"positions":[140],"across":[141],"all":[142],"forget":[143,191],"sequences.":[144],"On":[145],"TOFU,":[146],"MUSE":[147,150],"News,":[148],"Books,":[151],"achieves":[153],"competitive":[155],"trade-off":[157],"at":[158],"fraction":[160],"computational":[163],"cost":[164],"existing":[166],"baselines.":[167],"We":[168],"further":[169],"observe":[170],"increase":[174],"size":[176],"(a.k.a.":[177],"number":[178],"parameters),":[180],"trade-offs":[182],"improve":[183],"for":[184],"both":[185],"SimNPO":[188],"--":[189],"metrics":[192],"remain":[193],"comparable":[194],"retain":[196],"increases.":[198]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-06-04T00:00:00"}