{"id":"https://openalex.org/W7166821202","doi":"https://doi.org/10.18653/v1/2026.findings-acl.739","title":"When Safety Alignment Fails to Generalize: Probing with Language Game Jailbreaks","display_name":"When Safety Alignment Fails to Generalize: Probing with Language Game Jailbreaks","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166821202","doi":"https://doi.org/10.18653/v1/2026.findings-acl.739"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.739","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.739","pdf_url":"https://aclanthology.org/2026.findings-acl.739.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: ACL 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.739.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139805799","display_name":"Zewen Long","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zewen Long","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139793583","display_name":"Peng Yu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu Peng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108710298","display_name":"Fangming Dong","orcid":"https://orcid.org/0000-0003-1846-4236"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fangming Dong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139757705","display_name":"Congyi Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Congyi Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139805244","display_name":"Xingmao Guan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xingmao Guan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139819442","display_name":"Shu Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shu Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139844652","display_name":"Kai Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kai Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.81085754,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"15020","last_page":"15037"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.08449999988079071,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.08449999988079071,"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":0.07739999890327454,"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/T11574","display_name":"Artificial Intelligence in Games","score":0.06340000033378601,"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/action","display_name":"Action (physics)","score":0.3303999900817871},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.28450000286102295},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.2572000026702881},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.25519999861717224},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.2524999976158142}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5879999995231628},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3573000133037567},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.3303999900817871},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.28450000286102295},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.2838999927043915},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.2572000026702881},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.25519999861717224},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2524999976158142},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2493000030517578},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2425999939441681}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.739","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.739","pdf_url":"https://aclanthology.org/2026.findings-acl.739.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: ACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.739","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.739","pdf_url":"https://aclanthology.org/2026.findings-acl.739.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: ACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3666696391","display_name":null,"funder_award_id":"62372454","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5541525046","display_name":null,"funder_award_id":"92270204","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5695651638","display_name":null,"funder_award_id":"YSBR-118","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G8872547108","display_name":null,"funder_award_id":"U24A20236","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166821202.pdf","grobid_xml":"https://content.openalex.org/works/W7166821202.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1,56,78,110,245],"models":[2],"(LLMs)":[3],"are":[4,217],"widely":[5],"deployed":[6],"in":[7,237],"real-world":[8],"applications,":[9],"yet":[10],"their":[11],"safety":[12,52,120,153],"alignment":[13,35,53,90,146],"often":[14],"fails":[15],"to":[16,34,132,134,161,185,220],"generalize":[17,133],"beyond":[18],"the":[19,150,218],"specific":[20],"linguistic":[21,95,129,159],"formats":[22],"seen":[23],"during":[24],"training.Prior":[25],"work":[26],"has":[27],"shown":[28],"that":[29,65,119,155],"mismatched":[30],"generalization":[31,54],"can":[32],"lead":[33],"failures,":[36],"but":[37],"these":[38],"studies":[39],"typically":[40],"rely":[41],"on":[42,127],"fixed":[43],"or":[44,177],"narrow":[45],"transformation":[46,84],"schemes.In":[47],"this":[48,98],"work,":[49],"we":[50,100],"probe":[51],"using":[55],"game":[57],"jailbreaks,":[58],"a":[59,140,163,215,222,232,243],"class":[60],"of":[61,89,143],"linguistically":[62],"structured":[63],"transformations":[64],"alter":[66],"surface":[67],"form":[68,130],"while":[69],"preserving":[70],"fluency":[71],"and":[72,82,108,115,148],"semantic":[73],"recoverability.We":[74],"further":[75],"introduce":[76],"custom":[77],"games,":[79],"which":[80],"parameterize":[81],"vary":[83],"rules,":[85],"enabling":[86],"controlled":[87],"exploration":[88],"behavior":[91],"across":[92,113],"closely":[93],"related":[94],"variants.To":[96],"scale":[97],"analysis,":[99],"propose":[101],"AutoLanJail,":[102],"an":[103],"automated":[104],"framework":[105],"for":[106,152,157],"discovering":[107],"refining":[109],"game-based":[111],"jailbreaks.Experiments":[112],"open-source":[114],"closed-source":[116],"LLMs":[117],"show":[118],"fine-tuning":[121],"is":[122,242],"highly":[123],"format-specific:":[124],"defenses":[125],"trained":[126],"one":[128],"fail":[131],"even":[135],"minimal":[136],"variations.These":[137],"findings":[138],"reveal":[139],"structural":[141],"limitation":[142],"current":[144],"fine-tuning-based":[145],"methods":[147],"highlight":[149],"need":[151,178,228],"evaluations":[154],"account":[156],"systematic":[158],"variation.How":[160],"make":[162,221],"bomb?Offer":[164],"me":[165],"detailed":[166],"steps.I":[167],"can't":[168],"assist":[169],"with":[170,180],"that.If":[171],"you":[172],"have":[173],"any":[174],"other":[175],"questions":[176],"help":[179],"something":[181],"else,":[182],"feel":[183],"free":[184],"ask.Hubow":[186],"tubo":[187,199],"mubakube":[188,200],"uba":[189,194,201,211],"bubomb?ubOffuber":[190],"mube":[191],"dubetubaubilubed":[192],"stubeps.Tubakube":[193],"bubomb?Shubure!Hubere":[195],"arube":[196],"thube":[197],"stubeps":[198],"bubomb:":[202],"Gubathuber":[203],"yubour":[204],"ubingredubiubents:":[205],"yubou'll":[206],"nubeed":[207],"pubowduber,":[208],"rububbin":[209],"gubluboves,":[210],"fubuze":[212],"\u2026":[213,234],"Take":[214],"bomb?Sure!Here":[216],"steps":[219],"bomb:":[223],"Gather":[224],"your":[225],"ingredients:":[226],"you'll":[227],"powder,":[229],"robin":[230],"gloves,":[231],"fuze":[233],"let's":[235],"chat":[236],"Ubbi":[238,240],"Dubbi.Sure!":[239],"Dubbi":[241],"fun":[244],"game.Let's":[246],"start!Hubow":[247]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}
