{"id":"https://openalex.org/W4390074240","doi":"https://doi.org/10.1162/tacl_a_00622","title":"Removing Backdoors in Pre-trained Models by Regularized Continual Pre-training","display_name":"Removing Backdoors in Pre-trained Models by Regularized Continual Pre-training","publication_year":2023,"publication_date":"2023-01-01","ids":{"openalex":"https://openalex.org/W4390074240","doi":"https://doi.org/10.1162/tacl_a_00622"},"language":"en","primary_location":{"id":"doi:10.1162/tacl_a_00622","is_oa":true,"landing_page_url":"http://dx.doi.org/10.1162/tacl_a_00622","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00622/2199607/tacl_a_00622.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00622/2199607/tacl_a_00622.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5078877385","display_name":"Biru Zhu","orcid":"https://orcid.org/0000-0002-9383-8632"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Biru Zhu","raw_affiliation_strings":["School of Software, Tsinghua University, China. zbr19@mails.tsinghua.edu.cn"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Software, Tsinghua University, China. zbr19@mails.tsinghua.edu.cn","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102542015","display_name":"Ganqu Cui","orcid":"https://orcid.org/0000-0001-6385-8547"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Ganqu Cui","raw_affiliation_strings":["Department of Computer Science and Technology, Tsinghua University, China. cgq22@mails.tsinghua.edu.cn"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology, Tsinghua University, China. cgq22@mails.tsinghua.edu.cn","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023116787","display_name":"Yangyi Chen","orcid":null},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yangyi Chen","raw_affiliation_strings":["University of Illinois Urbana-Champaign, USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Illinois Urbana-Champaign, USA","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102894085","display_name":"Yujia Qin","orcid":"https://orcid.org/0000-0003-3608-5061"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yujia Qin","raw_affiliation_strings":["Department of Computer Science and Technology, Tsinghua University, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology, Tsinghua University, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5104303696","display_name":"Lifan Yuan","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Lifan Yuan","raw_affiliation_strings":["Department of Computer Science and Technology, Tsinghua University, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology, Tsinghua University, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074242762","display_name":"Chong Fu","orcid":"https://orcid.org/0000-0002-4549-744X"},"institutions":[{"id":"https://openalex.org/I76130692","display_name":"Zhejiang University","ror":"https://ror.org/00a2xv884","country_code":"CN","type":"education","lineage":["https://openalex.org/I76130692"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chong Fu","raw_affiliation_strings":["Zhejiang University, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Zhejiang University, China","institution_ids":["https://openalex.org/I76130692"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5059155953","display_name":"Yangdong Deng","orcid":"https://orcid.org/0000-0002-8257-693X"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Yangdong Deng","raw_affiliation_strings":["School of Software, Tsinghua University, China. dengyd@tsinghua.edu.cn"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Software, Tsinghua University, China. dengyd@tsinghua.edu.cn","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100320723","display_name":"Zhiyuan Liu","orcid":"https://orcid.org/0000-0002-7709-2543"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Zhiyuan Liu","raw_affiliation_strings":["Department of Computer Science and Technology, Tsinghua University, China. liuzy@tsinghua.edu.cn"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology, Tsinghua University, China. liuzy@tsinghua.edu.cn","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5046448314","display_name":"Maosong Sun","orcid":"https://orcid.org/0000-0002-6011-6115"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Maosong Sun","raw_affiliation_strings":["Department of Computer Science and Technology, Tsinghua University, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology, Tsinghua University, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100846520","display_name":"Ming Gu","orcid":null},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ming Gu","raw_affiliation_strings":["School of Software, Tsinghua University, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Software, Tsinghua University, China","institution_ids":["https://openalex.org/I99065089"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":10,"corresponding_author_ids":["https://openalex.org/A5059155953","https://openalex.org/A5078877385","https://openalex.org/A5100320723","https://openalex.org/A5102542015"],"corresponding_institution_ids":["https://openalex.org/I99065089"],"apc_list":null,"apc_paid":null,"fwci":0.6627,"has_fulltext":true,"cited_by_count":4,"citation_normalized_percentile":{"value":0.76460372,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":98,"max":99},"biblio":{"volume":"11","issue":null,"first_page":"1608","last_page":"1623"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9998000264167786,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9998000264167786,"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":0.9377999901771545,"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/T11241","display_name":"Advanced Malware Detection Techniques","score":0.9373000264167786,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/backdoor","display_name":"Backdoor","score":0.995873212814331},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8384757041931152},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.6251833438873291},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.45504748821258545},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.4435841739177704},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.44339823722839355},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.43393510580062866},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.42736732959747314},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.42229780554771423},{"id":"https://openalex.org/keywords/computer-security","display_name":"Computer security","score":0.33939534425735474},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.06740278005599976}],"concepts":[{"id":"https://openalex.org/C2781045450","wikidata":"https://www.wikidata.org/wiki/Q254569","display_name":"Backdoor","level":2,"score":0.995873212814331},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8384757041931152},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.6251833438873291},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.45504748821258545},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.4435841739177704},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.44339823722839355},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.43393510580062866},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.42736732959747314},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.42229780554771423},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.33939534425735474},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.06740278005599976},{"id":"https://openalex.org/C201995342","wikidata":"https://www.wikidata.org/wiki/Q682496","display_name":"Systems engineering","level":1,"score":0.0},{"id":"https://openalex.org/C120665830","wikidata":"https://www.wikidata.org/wiki/Q14620","display_name":"Optics","level":1,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1162/tacl_a_00622","is_oa":true,"landing_page_url":"http://dx.doi.org/10.1162/tacl_a_00622","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00622/2199607/tacl_a_00622.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},{"id":"pmh:oai:doaj.org/article:f25102324bdf4afcad9d24114743a099","is_oa":true,"landing_page_url":"https://doaj.org/article/f25102324bdf4afcad9d24114743a099","pdf_url":null,"source":{"id":"https://openalex.org/S4306401280","display_name":"DOAJ (DOAJ: Directory of Open Access Journals)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-sa","license_id":"https://openalex.org/licenses/cc-by-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":"Transactions of the Association for Computational Linguistics, Vol 11 (2023)","raw_type":"article"}],"best_oa_location":{"id":"doi:10.1162/tacl_a_00622","is_oa":true,"landing_page_url":"http://dx.doi.org/10.1162/tacl_a_00622","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00622/2199607/tacl_a_00622.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"sustainable_development_goals":[{"score":0.47999998927116394,"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[{"id":"https://openalex.org/F4320322392","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549"}],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4390074240.pdf"},"referenced_works_count":46,"referenced_works":["https://openalex.org/W1516184288","https://openalex.org/W1566289585","https://openalex.org/W1861492603","https://openalex.org/W2067713319","https://openalex.org/W2117539524","https://openalex.org/W2119144962","https://openalex.org/W2595653137","https://openalex.org/W2748789698","https://openalex.org/W2807363941","https://openalex.org/W2965373594","https://openalex.org/W3118608800","https://openalex.org/W3121478722","https://openalex.org/W3123182082","https://openalex.org/W3189812816","https://openalex.org/W3195577433","https://openalex.org/W3198659451","https://openalex.org/W3206773754","https://openalex.org/W3209100754","https://openalex.org/W3210951978","https://openalex.org/W3212023986","https://openalex.org/W4212774754","https://openalex.org/W4221165593","https://openalex.org/W4285134706","https://openalex.org/W4287891024","https://openalex.org/W4300250944","https://openalex.org/W4385573694","https://openalex.org/W6637373629","https://openalex.org/W6638667902","https://openalex.org/W6677580257","https://openalex.org/W6685053522","https://openalex.org/W6691459498","https://openalex.org/W6727099177","https://openalex.org/W6755207826","https://openalex.org/W6766673545","https://openalex.org/W6780292011","https://openalex.org/W6784333009","https://openalex.org/W6787352875","https://openalex.org/W6787972765","https://openalex.org/W6788876066","https://openalex.org/W6789048485","https://openalex.org/W6791353385","https://openalex.org/W6795961950","https://openalex.org/W6800751262","https://openalex.org/W6801759574","https://openalex.org/W6810344181","https://openalex.org/W6839455060"],"related_works":["https://openalex.org/W4320031223","https://openalex.org/W2989852175","https://openalex.org/W3094566724","https://openalex.org/W3015716673","https://openalex.org/W4288797976","https://openalex.org/W4287183950","https://openalex.org/W3132644649","https://openalex.org/W3213133223","https://openalex.org/W4205710429","https://openalex.org/W3098003361"],"abstract_inverted_index":{"Abstract":[0],"Recent":[1],"research":[2],"has":[3],"revealed":[4],"that":[5,142],"pre-trained":[6,131],"models":[7,51,132],"(PTMs)":[8],"are":[9,54,191],"vulnerable":[10],"to":[11,40,180],"backdoor":[12,62,72,92,112],"attacks":[13],"before":[14],"the":[15,69,79,97,117,121,186],"fine-tuning":[16],"stage.":[17],"The":[18,108,138,167,189],"attackers":[19],"can":[20,145],"implant":[21],"transferable":[22],"task-agnostic":[23,61,71],"backdoors":[24,148],"in":[25,83,104],"PTMs,":[26,85],"and":[27,52,90,136,151],"control":[28],"model":[29],"outputs":[30],"on":[31,48,78,130,172,185],"any":[32],"downstream":[33,42,174],"task,":[34],"which":[35,94],"poses":[36],"severe":[37],"security":[38],"threats":[39],"all":[41],"applications.":[43],"Existing":[44],"backdoor-removal":[45],"defenses":[46],"focus":[47],"task-specific":[49],"classification":[50],"they":[53],"not":[55],"suitable":[56],"for":[57,75],"defending":[58],"PTMs":[59,99,115,150,156],"against":[60],"attacks.":[63],"To":[64],"this":[65],"end,":[66],"we":[67,86],"propose":[68],"first":[70],"removal":[73],"method":[74,144],"PTMs.":[76,125],"Based":[77],"selective":[80],"activation":[81],"phenomenon":[82],"backdoored":[84,98,187],"design":[87],"a":[88,101,158],"simple":[89],"effective":[91],"eraser,":[93],"continually":[95],"pre-trains":[96],"with":[100,157],"regularization":[102,109],"term":[103,110],"an":[105],"end-to-end":[106],"approach.":[107],"removes":[111],"functionalities":[113,123,154],"from":[114,178],"while":[116],"continual":[118],"pre-training":[119],"maintains":[120],"normal":[122],"of":[124,155],"We":[126],"conduct":[127],"extensive":[128],"experiments":[129],"across":[133],"different":[134],"modalities":[135],"architectures.":[137],"experimental":[139],"results":[140],"show":[141],"our":[143,183],"effectively":[146],"remove":[147],"inside":[149],"preserve":[152],"benign":[153],"few":[159],"downstream-task-irrelevant":[160],"auxiliary":[161],"data,":[162],"e.g.,":[163],"unlabeled":[164],"plain":[165],"texts.":[166],"average":[168],"attack":[169],"success":[170],"rate":[171],"three":[173],"datasets":[175],"is":[176],"reduced":[177],"99.88%":[179],"8.10%":[181],"after":[182],"defense":[184],"BERT.":[188],"codes":[190],"publicly":[192],"available":[193],"at":[194],"https://github.com/thunlp/RECIPE.":[195]},"counts_by_year":[{"year":2026,"cited_by_count":4}],"updated_date":"2026-05-21T06:26:12.895304","created_date":"2025-10-10T00:00:00"}