{"id":"https://openalex.org/W7133324420","doi":"https://doi.org/10.48550/arxiv.2603.01792","title":"ALTER: Asymmetric LoRA for Token-Entropy-Guided Unlearning of LLMs","display_name":"ALTER: Asymmetric LoRA for Token-Entropy-Guided Unlearning of LLMs","publication_year":2026,"publication_date":"2026-03-02","ids":{"openalex":"https://openalex.org/W7133324420","doi":"https://doi.org/10.48550/arxiv.2603.01792"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.01792","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01792","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":null,"license_id":null,"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.2603.01792","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5128011186","display_name":"Xunlei Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Xunlei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128004295","display_name":"Jinyu Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Jinyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5032504077","display_name":"Yuang Li","orcid":"https://orcid.org/0000-0002-2244-0184"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Yuang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127987736","display_name":"Zhaokun Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zhaokun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127929115","display_name":"Yi Gong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gong, Yi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128016832","display_name":"Jie Zou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zou, Jie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127892489","display_name":"Jiwei Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Jiwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5127932981","display_name":"Wenhong Tian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tian, Wenhong","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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.2538999915122986,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.2538999915122986,"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/T10028","display_name":"Topic Modeling","score":0.179299995303154,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.049800001084804535,"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/isolation","display_name":"Isolation (microbiology)","score":0.48069998621940613},{"id":"https://openalex.org/keywords/forgetting","display_name":"Forgetting","score":0.4309000074863434},{"id":"https://openalex.org/keywords/decoupling","display_name":"Decoupling (probability)","score":0.3370000123977661},{"id":"https://openalex.org/keywords/fuzzy-logic","display_name":"Fuzzy logic","score":0.31769999861717224},{"id":"https://openalex.org/keywords/entropy","display_name":"Entropy (arrow of time)","score":0.31290000677108765}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.555899977684021},{"id":"https://openalex.org/C2775941552","wikidata":"https://www.wikidata.org/wiki/Q25212305","display_name":"Isolation (microbiology)","level":2,"score":0.48069998621940613},{"id":"https://openalex.org/C7149132","wikidata":"https://www.wikidata.org/wiki/Q1377840","display_name":"Forgetting","level":2,"score":0.4309000074863434},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.4077000021934509},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.37130001187324524},{"id":"https://openalex.org/C205606062","wikidata":"https://www.wikidata.org/wiki/Q5249645","display_name":"Decoupling (probability)","level":2,"score":0.3370000123977661},{"id":"https://openalex.org/C58166","wikidata":"https://www.wikidata.org/wiki/Q224821","display_name":"Fuzzy logic","level":2,"score":0.31769999861717224},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.31290000677108765},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.30399999022483826},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.29739999771118164},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.29589998722076416},{"id":"https://openalex.org/C56739046","wikidata":"https://www.wikidata.org/wiki/Q192060","display_name":"Knowledge management","level":1,"score":0.2782999873161316},{"id":"https://openalex.org/C190253527","wikidata":"https://www.wikidata.org/wiki/Q295354","display_name":"Law and economics","level":1,"score":0.26109999418258667},{"id":"https://openalex.org/C2779536738","wikidata":"https://www.wikidata.org/wiki/Q658179","display_name":"Oryx","level":2,"score":0.25060001015663147}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.01792","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01792","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.01792","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01792","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":null,"license_id":null,"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":{"Large":[0],"language":[1],"models":[2,78],"(LLMs)":[3],"have":[4],"advanced":[5],"to":[6,38,74,99],"encompass":[7],"extensive":[8],"knowledge":[9,43,105],"across":[10],"diverse":[11],"domains.":[12],"Yet":[13],"controlling":[14],"what":[15],"a":[16,93,139,155],"LLMs":[17,34,98],"should":[18],"not":[19],"know":[20],"is":[21,35,49],"important":[22],"for":[23,97,159],"ensuring":[24],"alignment":[25],"and":[26,45,107,121,146,176,184],"thus":[27],"safe":[28],"use.":[29],"However,":[30],"effective":[31],"unlearning":[32,67,95,108,147,161,195],"in":[33,61,128,165],"difficult":[36],"due":[37],"the":[39,70,102,124,150,166],"fuzzy":[40],"boundary":[41],"between":[42],"retention":[44],"forgetting.":[46],"This":[47],"challenge":[48],"exacerbated":[50],"by":[51,131,143],"entangled":[52],"parameter":[53,144],"spaces":[54],"from":[55,196],"continuous":[56],"multi-domain":[57],"training,":[58],"often":[59],"resulting":[60],"collateral":[62],"damage,":[63],"especially":[64],"under":[65],"aggressive":[66],"strategies.":[68],"Furthermore,":[69],"computational":[71],"overhead":[72],"required":[73],"optimize":[75],"State-of-the-Art":[76],"(SOTA)":[77],"with":[79,179],"billions":[80],"of":[81,104,209,216],"parameters":[82],"poses":[83],"an":[84,133],"additional":[85],"barrier.":[86],"In":[87],"this":[88,200],"work,":[89],"we":[90],"present":[91],"ALTER,":[92],"lightweight":[94],"framework":[96,201],"address":[100],"both":[101],"challenges":[103],"entanglement":[106],"efficiency.":[109],"ALTER":[110,169],"operates":[111],"through":[112,189],"two":[113],"phases:":[114],"(I)":[115],"high":[116],"entropy":[117],"tokens":[118,148],"are":[119],"captured":[120],"learned":[122],"via":[123,162],"shared":[125],"A":[126],"matrix":[127],"LoRA,":[129],"followed":[130],"(II)":[132],"asymmetric":[134,167],"LoRA":[135],"architecture":[136],"that":[137],"achieves":[138,170],"specified":[140],"forgetting":[141],"objective":[142],"isolation":[145,164],"within":[149],"target":[151],"subdomains.":[152],"Serving":[153],"as":[154],"new":[156],"research":[157],"direction":[158],"achieving":[160],"token-level":[163],"framework.":[168],"SOTA":[171],"performance":[172],"on":[173],"TOFU,":[174],"WMDP,":[175],"MUSE":[177],"benchmarks":[178],"over":[180,207],"95%":[181],"forget":[182],"quality":[183],"shows":[185],"minimal":[186],"side":[187],"effects":[188],"preserving":[190,206],"foundational":[191],"tokens.":[192],"By":[193],"decoupling":[194],"LLMs'":[197],"billion-scale":[198],"parameters,":[199],"delivers":[202],"excellent":[203],"efficiency":[204],"while":[205],"90%":[208],"model":[210],"utility,":[211],"exceeding":[212],"baseline":[213],"preservation":[214],"rates":[215],"47.8-83.6%.":[217]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-03-04T00:00:00"}