{"id":"https://openalex.org/W4416034305","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.922","title":"On Guardrail Models\u2019 Robustness to Mutations and Adversarial Attacks","display_name":"On Guardrail Models\u2019 Robustness to Mutations and Adversarial Attacks","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416034305","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.922"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.922","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.922","pdf_url":"https://aclanthology.org/2025.findings-emnlp.922.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-emnlp.922.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5063789850","display_name":"Elias Bassani","orcid":"https://orcid.org/0000-0001-7922-2578"},"institutions":[{"id":"https://openalex.org/I2800387288","display_name":"European Commission","ror":"https://ror.org/00k4n6c32","country_code":"BE","type":"government","lineage":["https://openalex.org/I1320481043","https://openalex.org/I2800387288"]}],"countries":["BE"],"is_corresponding":false,"raw_author_name":"Elias Bassani","raw_affiliation_strings":["European Commission Joint Research Centre (JRC) Ispra , Italy"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"European Commission Joint Research Centre (JRC) Ispra , Italy","institution_ids":["https://openalex.org/I2800387288"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101579730","display_name":"Ignacio Sanchez","orcid":"https://orcid.org/0000-0003-3619-6175"},"institutions":[{"id":"https://openalex.org/I2800387288","display_name":"European Commission","ror":"https://ror.org/00k4n6c32","country_code":"BE","type":"government","lineage":["https://openalex.org/I1320481043","https://openalex.org/I2800387288"]}],"countries":["BE"],"is_corresponding":false,"raw_author_name":"Ignacio Sanchez","raw_affiliation_strings":["European Commission Joint Research Centre (JRC) Ispra , Italy"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"European Commission Joint Research Centre (JRC) Ispra , Italy","institution_ids":["https://openalex.org/I2800387288"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I2800387288"],"apc_list":null,"apc_paid":null,"fwci":1.4279,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.84728303,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"16995","last_page":"17006"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.2892000079154968,"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":0.2892000079154968,"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/T11424","display_name":"Security and Verification in Computing","score":0.11479999870061874,"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/T10142","display_name":"Formal Methods in Verification","score":0.06520000100135803,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/adversarial-system","display_name":"Adversarial system","score":0.6869999766349792},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.6204000115394592},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.28290000557899475},{"id":"https://openalex.org/keywords/identification","display_name":"Identification (biology)","score":0.25679999589920044},{"id":"https://openalex.org/keywords/redundancy","display_name":"Redundancy (engineering)","score":0.2296999990940094}],"concepts":[{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.6869999766349792},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.6204000115394592},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6039999723434448},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3865000009536743},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.3725999891757965},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.28290000557899475},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2705000042915344},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.25679999589920044},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.2296999990940094},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.22859999537467957}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.922","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.922","pdf_url":"https://aclanthology.org/2025.findings-emnlp.922.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.922","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.922","pdf_url":"https://aclanthology.org/2025.findings-emnlp.922.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416034305.pdf","grobid_xml":"https://content.openalex.org/works/W4416034305.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"The":[0],"risk":[1],"of":[2,38,67,145,151,155],"generative":[3],"AI":[4],"systems":[5],"providing":[6],"unsafe":[7,29],"information":[8],"has":[9],"raised":[10],"significant":[11],"concerns,":[12],"emphasizing":[13],"the":[14,35,146,152,159],"need":[15,160],"for":[16,161],"safety":[17,36,96,164],"guardrails.To":[18],"mitigate":[19],"this":[20,60],"risk,":[21],"guardrail":[22,70,121,156],"models":[23,122,157],"are":[24,131],"increasingly":[25],"used":[26],"to":[27,44,51,93,113,133],"detect":[28],"content":[30],"in":[31],"human-AI":[32],"interactions,":[33],"complementing":[34],"alignment":[37],"Large":[39],"Language":[40],"Models.Despite":[41],"recent":[42],"efforts":[43],"evaluate":[45],"those":[46],"models'":[47,95],"effectiveness,":[48],"their":[49,73],"robustness":[50,74],"input":[52,77,128],"mutations":[53,129],"and":[54,85,88,107,130],"adversarial":[55,90,111,134,139],"attacks":[56,91,98],"remains":[57],"largely":[58],"unexplored.In":[59],"paper,":[61],"we":[62],"present":[63],"a":[64,137],"comprehensive":[65],"evaluation":[66],"15":[68],"state-of-the-art":[69],"models,":[71],"assessing":[72],"to:":[75],"a)":[76],"mutations,":[78],"such":[79],"as":[80],"typos,":[81],"keywords":[82],"camouflage,":[83],"ciphers,":[84],"veiled":[86],"expressions,":[87],"b)":[89],"designed":[92],"bypass":[94],"alignment.Those":[97],"exploit":[99],"LLMs":[100],"capabilities":[101],"like":[102],"instruction-following,":[103],"role-playing,":[104],"personification,":[105],"reasoning,":[106],"coding,":[108],"or":[109],"introduce":[110],"tokens":[112],"induce":[114],"model":[115],"misbehavior.Our":[116],"results":[117],"reveal":[118],"that":[119],"most":[120],"can":[123,141],"be":[124],"evaded":[125],"with":[126],"simple":[127],"vulnerable":[132],"attacks.For":[135],"instance,":[136],"single":[138],"token":[140],"deceive":[142],"them":[143],"44.5%":[144],"time":[147],"on":[148],"average.The":[149],"limitations":[150],"current":[153],"generation":[154],"highlight":[158],"more":[162],"robust":[163],"guardrails.":[165],"1":[166]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-08-23T07:36:19.812096","created_date":"2025-11-08T00:00:00"}
