{"id":"https://openalex.org/W7154048428","doi":"https://doi.org/10.48550/arxiv.2604.09405","title":"EGLOCE: Training-Free Energy-Guided Latent Optimization for Concept Erasure","display_name":"EGLOCE: Training-Free Energy-Guided Latent Optimization for Concept Erasure","publication_year":2026,"publication_date":"2026-04-10","ids":{"openalex":"https://openalex.org/W7154048428","doi":"https://doi.org/10.48550/arxiv.2604.09405"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.09405","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09405","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.09405","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5110035590","display_name":"Junyeong Ahn","orcid":"https://orcid.org/0009-0008-1456-7020"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Ahn, Junyeong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133503171","display_name":"Seojin Yoon","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yoon, Seojin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5048206537","display_name":"Sungyong Baik","orcid":"https://orcid.org/0000-0001-5702-4618"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Baik, Sungyong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5110035590"],"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.46230000257492065,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.46230000257492065,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.2467000037431717,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.10119999945163727,"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/erasure","display_name":"Erasure","score":0.7720999717712402},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5579000115394592},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.42489999532699585},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.3799999952316284},{"id":"https://openalex.org/keywords/component","display_name":"Component (thermodynamics)","score":0.3571999967098236},{"id":"https://openalex.org/keywords/energy","display_name":"Energy (signal processing)","score":0.3402999937534332},{"id":"https://openalex.org/keywords/construct","display_name":"Construct (python library)","score":0.33379998803138733},{"id":"https://openalex.org/keywords/optimization-problem","display_name":"Optimization problem","score":0.3174999952316284}],"concepts":[{"id":"https://openalex.org/C2778790127","wikidata":"https://www.wikidata.org/wiki/Q484885","display_name":"Erasure","level":2,"score":0.7720999717712402},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6671000123023987},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5579000115394592},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4609000086784363},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.42489999532699585},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.3799999952316284},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.3571999967098236},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.3402999937534332},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.33379998803138733},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3319000005722046},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.328900009393692},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.3174999952316284},{"id":"https://openalex.org/C205372313","wikidata":"https://www.wikidata.org/wiki/Q355645","display_name":"Latent image","level":3,"score":0.30160000920295715},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.29339998960494995},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.2833000123500824},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.28189998865127563},{"id":"https://openalex.org/C2780980858","wikidata":"https://www.wikidata.org/wiki/Q110022","display_name":"Dual (grammatical number)","level":2,"score":0.2815000116825104},{"id":"https://openalex.org/C51167844","wikidata":"https://www.wikidata.org/wiki/Q4422623","display_name":"Latent variable","level":2,"score":0.2775000035762787},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2761000096797943},{"id":"https://openalex.org/C206688291","wikidata":"https://www.wikidata.org/wiki/Q7617819","display_name":"Stochastic gradient descent","level":3,"score":0.267300009727478},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.2632000148296356},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.2572999894618988},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2500999867916107}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.09405","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09405","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.09405","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.09405","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":"article"},"sustainable_development_goals":[{"score":0.8384003639221191,"display_name":"Affordable and clean energy","id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"As":[0],"text-to-image":[1],"diffusion":[2,71],"models":[3],"grow":[4],"increasingly":[5],"prevalent,":[6],"the":[7,37,53,60,68,126,179,187],"ability":[8],"to":[9,125,189],"remove":[10],"specific":[11],"concepts-mostly":[12],"explicit":[13],"content":[14],"and":[15,25,117,150,169,196],"many":[16],"copyrighted":[17],"characters":[18],"or":[19,45,140],"styles-has":[20],"become":[21],"essential":[22],"for":[23,65,78,194],"safety":[24],"compliance.":[26],"Existing":[27],"unlearning":[28],"approaches":[29,132],"often":[30],"require":[31,135],"costly":[32],"re-training,":[33],"modify":[34],"parameters":[35],"at":[36,148],"cost":[38],"of":[39,41,55,62,67,70,181],"degradation":[40],"unrelated":[42],"concept":[43,56,163],"fidelity,":[44],"depend":[46],"on":[47],"indirect":[48],"inference-time":[49,143],"adjustment":[50],"that":[51,85,104,121,133,160],"compromise":[52],"effectiveness":[54],"erasure.":[57],"Inspired":[58],"by":[59,89],"success":[61],"energy-guided":[63],"sampling":[64],"preservation":[66],"condition":[69],"models,":[72],"we":[73],"introduce":[74],"Energy-Guided":[75],"Latent":[76],"Optimization":[77],"Concept":[79],"Erasure":[80],"(EGLOCE),":[81],"a":[82,98,101,118,191],"training-free":[83],"approach":[84],"removes":[86],"unwanted":[87],"concepts":[88,110],"re-directing":[90],"noisy":[91],"latent":[92,115],"during":[93,204],"inference.":[94],"Our":[95],"method":[96],"employs":[97],"dual-objective":[99],"framework:":[100],"repulsion":[102],"energy":[103,120],"steers":[105],"generation":[106,199],"away":[107],"from":[108],"target":[109],"via":[111],"gradient":[112],"descent":[113],"in":[114],"space,":[116],"retention":[119],"preserves":[122],"semantic":[123],"alignment":[124,171],"original":[127],"prompt.":[128],"Combined":[129],"with":[130,175],"previous":[131],"either":[134],"erroneous":[136],"modified":[137],"model":[138],"weights":[139],"provide":[141],"weak":[142],"guidance,":[144],"EGLOCE":[145,161],"operates":[146],"entirely":[147],"inference":[149],"enhances":[151],"erasure":[152],"performance,":[153],"enabling":[154],"plug-and-play":[155],"integration.":[156],"Extensive":[157],"experiments":[158],"demonstrate":[159],"improves":[162],"removal":[164],"while":[165],"maintaining":[166],"image":[167,198],"quality":[168],"prompt":[170],"across":[172],"baselines,":[173],"even":[174],"adversarial":[176],"attacks.":[177],"To":[178],"best":[180],"our":[182,184],"knowledge,":[183],"work":[185],"is":[186],"first":[188],"establish":[190],"new":[192],"paradigm":[193],"safe":[195],"controllable":[197],"through":[200],"dual":[201],"energy-based":[202],"guidance":[203],"sampling.":[205]},"counts_by_year":[],"updated_date":"2026-05-05T08:41:31.759640","created_date":"2026-04-14T00:00:00"}