{"id":"https://openalex.org/W7151214682","doi":"https://doi.org/10.1109/icmla66185.2025.00172","title":"Jailbreaking Large Language Models: Safety Alignment, Response Quality, Computational Cost","display_name":"Jailbreaking Large Language Models: Safety Alignment, Response Quality, Computational Cost","publication_year":2025,"publication_date":"2025-12-03","ids":{"openalex":"https://openalex.org/W7151214682","doi":"https://doi.org/10.1109/icmla66185.2025.00172"},"language":null,"primary_location":{"id":"doi:10.1109/icmla66185.2025.00172","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icmla66185.2025.00172","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 International Conference on Machine Learning and Applications (ICMLA)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5015962925","display_name":"Jonas Rosengren","orcid":null},"institutions":[{"id":"https://openalex.org/I1279596006","display_name":"Statistics Sweden","ror":"https://ror.org/05x7wz523","country_code":"SE","type":"government","lineage":["https://openalex.org/I1279596006"]}],"countries":["SE"],"is_corresponding":true,"raw_author_name":"Jonas Rosengren","raw_affiliation_strings":["Lovable Labs,Stockholm,Sweden,SE-111 37"],"affiliations":[{"raw_affiliation_string":"Lovable Labs,Stockholm,Sweden,SE-111 37","institution_ids":["https://openalex.org/I1279596006"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5051919182","display_name":"Joel Brynielsson","orcid":"https://orcid.org/0000-0002-2677-9759"},"institutions":[{"id":"https://openalex.org/I86987016","display_name":"KTH Royal Institute of Technology","ror":"https://ror.org/026vcq606","country_code":"SE","type":"education","lineage":["https://openalex.org/I86987016"]}],"countries":["SE"],"is_corresponding":false,"raw_author_name":"Joel Brynielsson","raw_affiliation_strings":["KTH Royal Institute of Technology,Stockholm,Sweden,SE-100 44"],"affiliations":[{"raw_affiliation_string":"KTH Royal Institute of Technology,Stockholm,Sweden,SE-100 44","institution_ids":["https://openalex.org/I86987016"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133086592","display_name":"Fredrik Johansson","orcid":null},"institutions":[{"id":"https://openalex.org/I1291458624","display_name":"Swedish Defence Research Agency","ror":"https://ror.org/0470cgs30","country_code":"SE","type":"funder","lineage":["https://openalex.org/I1291458624"]}],"countries":["SE"],"is_corresponding":false,"raw_author_name":"Fredrik Johansson","raw_affiliation_strings":["FOI Swedish Defence Research Agency,Stockholm,Sweden,SE-164 90"],"affiliations":[{"raw_affiliation_string":"FOI Swedish Defence Research Agency,Stockholm,Sweden,SE-164 90","institution_ids":["https://openalex.org/I1291458624"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5088559756","display_name":"Patrik Jonell","orcid":"https://orcid.org/0000-0003-3687-6189"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Patrik Jonell","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5015962925"],"corresponding_institution_ids":["https://openalex.org/I1279596006"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.87270965,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1116","last_page":"1123"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12380","display_name":"Authorship Attribution and Profiling","score":0.22390000522136688,"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/T12380","display_name":"Authorship Attribution and Profiling","score":0.22390000522136688,"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":0.11249999701976776,"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/T13910","display_name":"Computational and Text Analysis Methods","score":0.060100000351667404,"subfield":{"id":"https://openalex.org/subfields/3300","display_name":"General Social Sciences"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.29679998755455017},{"id":"https://openalex.org/keywords/action","display_name":"Action (physics)","score":0.2881999909877777},{"id":"https://openalex.org/keywords/identification","display_name":"Identification (biology)","score":0.24979999661445618},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.24809999763965607},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.23229999840259552}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6360999941825867},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.29679998755455017},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.2881999909877777},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2506999969482422},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.24979999661445618},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.24809999763965607},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.23229999840259552},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.2287999987602234},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.2224999964237213},{"id":"https://openalex.org/C200601418","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability engineering","level":1,"score":0.2223999947309494}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icmla66185.2025.00172","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icmla66185.2025.00172","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 International Conference on Machine Learning and Applications (ICMLA)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":3,"referenced_works":["https://openalex.org/W4394567566","https://openalex.org/W4415800641","https://openalex.org/W7133211372"],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1,82,158],"models":[2],"are":[3,75],"often":[4,115],"equipped":[5],"with":[6,134],"safety":[7,34,97,154],"alignment":[8],"mechanisms":[9],"designed":[10],"to":[11,28,95,148],"prevent":[12],"generation":[13],"of":[14,24,144],"harmful":[15,53],"or":[16],"other":[17],"unwanted":[18],"content.":[19],"However,":[20],"an":[21,39],"increasing":[22],"number":[23],"jailbreaking":[25,45,151],"techniques":[26,74],"attempt":[27],"circumvent":[29],"these":[30],"safeguards,":[31],"raising":[32],"significant":[33,121],"concerns.":[35],"This":[36],"paper":[37],"introduces":[38],"open-source":[40],"evaluation":[41,147],"framework":[42],"that":[43,87],"analyzes":[44],"effectiveness":[46,152],"in":[47,156],"several":[48],"dimensions:":[49],"refusal":[50,89,105],"bypass":[51,90],"rate,":[52,113],"response":[54],"quality,":[55],"impact":[56],"on":[57],"general":[58,117],"model":[59,72,100,137],"capabilities,":[60],"and":[61,71,119,153],"computational":[62,122],"cost.":[63],"In":[64],"the":[65,142],"study,":[66],"prompt":[67],"injection,":[68],"sampling":[69],"exploits,":[70],"manipulation":[73,101],"examined":[76],"across":[77],"four":[78],"open-weight":[79],"instruction-tuned":[80],"large":[81,157],"models.":[83,159],"The":[84,139],"results":[85],"demonstrate":[86],"high":[88,110],"does":[91],"not":[92],"necessarily":[93],"equate":[94],"practical":[96,130],"compromise.":[98],"Specifically,":[99],"methods":[102],"like":[103],"single":[104],"direction":[106],"ablation":[107],"achieve":[108],"a":[109,128,135],"attack":[111],"success":[112],"but":[114],"degrade":[116],"capabilities":[118],"require":[120],"resources.":[123],"Meanwhile,":[124],"sampling-based":[125],"exploits":[126],"show":[127],"minimal":[129],"threat":[131],"when":[132],"assessed":[133],"robust":[136],"classifier.":[138],"findings":[140],"emphasize":[141],"importance":[143],"comprehensive,":[145],"multi-dimensional":[146],"accurately":[149],"characterize":[150],"risks":[155]},"counts_by_year":[],"updated_date":"2026-04-09T06:08:40.794217","created_date":"2026-04-08T00:00:00"}