{"id":"https://openalex.org/W4415318193","doi":"https://doi.org/10.18653/v1/2026.acl-long.2156","title":"Red-Bandit: Test-Time Adaptation for LLM Red-Teaming via Bandit-Guided LoRA Experts","display_name":"Red-Bandit: Test-Time Adaptation for LLM Red-Teaming via Bandit-Guided LoRA Experts","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W4415318193","doi":"https://doi.org/10.18653/v1/2026.acl-long.2156"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.acl-long.2156","doi":"https://doi.org/10.18653/v1/2026.acl-long.2156","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.2156","pdf_url":"https://aclanthology.org/2026.acl-long.2156.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2026.acl-long.2156.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5054494021","display_name":"Christos Ziakas","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Christos Ziakas","raw_affiliation_strings":["Imperial College London"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Nicholas Loo","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Nicholas Loo","raw_affiliation_strings":["Imperial College London"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Nishita Jain","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Nishita Jain","raw_affiliation_strings":["Imperial College London"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"last","author":{"id":null,"display_name":"Alessandra Russo","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Alessandra Russo","raw_affiliation_strings":["Imperial College London"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Imperial College London","institution_ids":["https://openalex.org/I47508984"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I47508984"],"apc_list":null,"apc_paid":null,"fwci":8.9348,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.93111819,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":93,"max":96},"biblio":{"volume":null,"issue":null,"first_page":"46462","last_page":"46478"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":1.0,"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":1.0,"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":0.0,"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/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.0,"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/test-time-adaptation","display_name":"test-time adaptation","score":0.8309999704360962},{"id":"https://openalex.org/keywords/multi-armed-bandits","display_name":"multi-armed bandits","score":0.8569999933242798},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"reinforcement learning","score":0.6119999885559082},{"id":"https://openalex.org/keywords/adversarial-prompts","display_name":"adversarial prompts","score":0.8619999885559082},{"id":"https://openalex.org/keywords/model-failure-modes","display_name":"model failure modes","score":0.6119999885559082},{"id":"https://openalex.org/keywords/exploration-exploitation-trade-off","display_name":"exploration-exploitation balance","score":0.6520000100135803}],"concepts":[{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.800000011920929},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7369999885559082},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.6377000212669373},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6277999877929688},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5630000233650208},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.5432000160217285},{"id":"https://openalex.org/C199521495","wikidata":"https://www.wikidata.org/wiki/Q181487","display_name":"Audit","level":2,"score":0.4860999882221222},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.4632999897003174},{"id":"https://openalex.org/C153180980","wikidata":"https://www.wikidata.org/wiki/Q19776675","display_name":"Commit","level":2,"score":0.3804999887943268},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.36890000104904175},{"id":"https://openalex.org/C140547941","wikidata":"https://www.wikidata.org/wiki/Q7797194","display_name":"Threat model","level":2,"score":0.328900009393692},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.32589998841285706},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.3107999861240387},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.30730000138282776},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.302700012922287},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3021000027656555},{"id":"https://openalex.org/C106189395","wikidata":"https://www.wikidata.org/wiki/Q176789","display_name":"Markov decision process","level":3,"score":0.2842999994754791},{"id":"https://openalex.org/C111065885","wikidata":"https://www.wikidata.org/wiki/Q1189053","display_name":"Fuzz testing","level":3,"score":0.2770000100135803},{"id":"https://openalex.org/C2776544517","wikidata":"https://www.wikidata.org/wiki/Q189447","display_name":"Unexpected events","level":2,"score":0.2671999931335449},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.2662000060081482},{"id":"https://openalex.org/C77967617","wikidata":"https://www.wikidata.org/wiki/Q4677561","display_name":"Active learning (machine learning)","level":2,"score":0.25519999861717224},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.2542000114917755},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2535000145435333}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.2156","doi":"https://doi.org/10.18653/v1/2026.acl-long.2156","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.2156","pdf_url":"https://aclanthology.org/2026.acl-long.2156.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2510.07239","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2510.07239","pdf_url":"https://arxiv.org/pdf/2510.07239","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2510.07239","doi":"https://doi.org/10.48550/arxiv.2510.07239","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.07239","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,"listed_in":[],"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"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.2156","doi":"https://doi.org/10.18653/v1/2026.acl-long.2156","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.2156","pdf_url":"https://aclanthology.org/2026.acl-long.2156.pdf","source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G6469877690","display_name":"Smart Integration of Process Systems Engineering & Machine Learning for Improved Process Safety in Process Industries","funder_award_id":"101119358","funder_id":"https://openalex.org/F4320338438","funder_display_name":"HORIZON EUROPE Marie Sklodowska-Curie Actions"},{"id":"https://openalex.org/G6770767669","display_name":"Smart Integration of Process Systems Engineering & Machine Learning for Improved Process Safety in Process Industries (PROSAFE)","funder_award_id":"EP/Y037111/1","funder_id":"https://openalex.org/F4320334627","funder_display_name":"Engineering and Physical Sciences Research Council"}],"funders":[{"id":"https://openalex.org/F4320314731","display_name":"UK Research and Innovation","ror":"https://ror.org/001aqnf71"},{"id":"https://openalex.org/F4320334627","display_name":"Engineering and Physical Sciences Research Council","ror":"https://ror.org/0439y7842"},{"id":"https://openalex.org/F4320338438","display_name":"HORIZON EUROPE Marie Sklodowska-Curie Actions","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415318193.pdf","grobid_xml":"https://content.openalex.org/works/W4415318193.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4415313671","https://openalex.org/W7117306638","https://openalex.org/W4416034565","https://openalex.org/W6967492529","https://openalex.org/W4404782310","https://openalex.org/W4388717799","https://openalex.org/W4409348331","https://openalex.org/W4403851788","https://openalex.org/W7157187412","https://openalex.org/W7163689423"],"abstract_inverted_index":{"Warning:":[0],"This":[1],"paper":[2],"contains":[3],"content":[4],"that":[5,46,80],"may":[6],"be":[7],"inappropriate,":[8],"offensive,":[9],"or":[10],"harmful.Automated":[11],"red-teaming":[12,44],"has":[13],"emerged":[14],"as":[15,144],"a":[16,43,64,73,88,93,145],"scalable":[17],"approach":[18],"for":[19,72,148],"auditing":[20],"Large":[21],"Language":[22],"Models":[23],"(LLMs)":[24],"prior":[25],"to":[26,33,36,49],"deployment,":[27],"yet":[28],"existing":[29],"approaches":[30],"lack":[31],"mechanisms":[32],"efficiently":[34],"adapt":[35],"model-specific":[37,150],"vulnerabilities":[38,151],"at":[39],"inference.We":[40],"introduce":[41],"Red-Bandit,":[42],"framework":[45],"adapts":[47],"online":[48],"identify":[50],"and":[51,112,119],"exploit":[52],"model":[53],"failure":[54],"modes":[55],"under":[56,126],"distinct":[57],"attack":[58,75,123,155],"styles":[59,156],"(e.g.,":[60],"manipulation,":[61],"slang).Red-Bandit":[62],"post-trains":[63],"set":[65],"of":[66,84],"parameter-efficient":[67],"LoRA":[68],"experts,":[69],"each":[70],"specialized":[71],"particular":[74],"style,":[76],"using":[77],"reinforcement":[78],"learning":[79],"rewards":[81],"the":[82,105],"generation":[83],"unsafe":[85],"prompts":[86,136],"via":[87],"rule-based":[89],"safety":[90],"model.At":[91],"inference,":[92],"multiarmed":[94],"bandit":[95,141],"policy":[96,142],"dynamically":[97],"selects":[98],"among":[99],"these":[100],"attack-style":[101],"experts":[102],"based":[103],"on":[104,117],"target":[106],"model's":[107],"response":[108],"safety,":[109],"balancing":[110],"exploration":[111,128],"exploitation.Red-Bandit":[113],"outperforms":[114],"stateof-the-art":[115],"methods":[116],"AdvBench":[118],"Harm-Bench,":[120],"achieving":[121],"higher":[122],"success":[124],"rates":[125],"sufficient":[127],"budgets":[129],"(ASR@10),":[130],"while":[131],"generating":[132],"more":[133],"human-readable":[134],"adversarial":[135],"(lower":[137],"perplexity).In":[138],"addition,":[139],"Red-Bandit's":[140],"serves":[143],"diagnostic":[146],"tool":[147],"identifying":[149],"by":[152],"indicating":[153],"which":[154],"most":[157],"effectively":[158],"elicit":[159],"harmful":[160],"behaviors.":[161]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-10-29T09:56:32.521200","created_date":"2025-10-18T00:00:00"}