{"id":"https://openalex.org/W7166879563","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1182","title":"RLShield: Dynamic Jailbreak Detection for LLMs via Reinforced Adaptive Learning","display_name":"RLShield: Dynamic Jailbreak Detection for LLMs via Reinforced Adaptive Learning","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166879563","doi":"https://doi.org/10.18653/v1/2026.findings-acl.1182"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.1182","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1182","pdf_url":"https://aclanthology.org/2026.findings-acl.1182.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.1182.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139836562","display_name":"Tong Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao Tong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039054062","display_name":"Pengfei Yang","orcid":"https://orcid.org/0000-0002-5803-8859"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pengfei Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139773710","display_name":"Yimeng Gu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yimeng Gu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014178940","display_name":"Haichao Shi","orcid":"https://orcid.org/0000-0001-8846-1853"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Haichao Shi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139833999","display_name":"Qiang Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qiang Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049694379","display_name":"Xingcheng Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xingcheng Xu","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/A5139766446","display_name":"Xiao-Yu Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiao-Yu Zhang","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":0.8676189,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"23615","last_page":"23630"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11652","display_name":"Imbalanced Data Classification Techniques","score":0.2070000022649765,"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/T11652","display_name":"Imbalanced Data Classification Techniques","score":0.2070000022649765,"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/T12034","display_name":"Digital and Cyber Forensics","score":0.1446000039577484,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.034699998795986176,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/adaptive-control","display_name":"Adaptive control","score":0.35249999165534973},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.30820000171661377},{"id":"https://openalex.org/keywords/adaptive-learning","display_name":"Adaptive learning","score":0.30480000376701355},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.29100000858306885},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.27410000562667847}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5482000112533569},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.45809999108314514},{"id":"https://openalex.org/C107464732","wikidata":"https://www.wikidata.org/wiki/Q235781","display_name":"Adaptive control","level":3,"score":0.35249999165534973},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.30820000171661377},{"id":"https://openalex.org/C125014702","wikidata":"https://www.wikidata.org/wiki/Q4680749","display_name":"Adaptive learning","level":2,"score":0.30480000376701355},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.30480000376701355},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.28790000081062317},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.27410000562667847},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.27070000767707825},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.2628999948501587},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.2581999897956848},{"id":"https://openalex.org/C52970973","wikidata":"https://www.wikidata.org/wiki/Q2497134","display_name":"Adaptive system","level":2,"score":0.2549000084400177}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.1182","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1182","pdf_url":"https://aclanthology.org/2026.findings-acl.1182.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.1182","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.1182","pdf_url":"https://aclanthology.org/2026.findings-acl.1182.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/G8504642471","display_name":null,"funder_award_id":"62376265","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/W7166879563.pdf","grobid_xml":"https://content.openalex.org/works/W7166879563.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"While":[0],"prompt":[1],"engineering":[2],"enhances":[3],"the":[4,18,40],"capabilities":[5],"of":[6,21,42,157],"Large":[7],"Language":[8],"Models":[9],"(LLMs),":[10],"it":[11,146],"also":[12],"exposes":[13],"critical":[14],"safety":[15,24],"concerns.Due":[16],"to":[17,29,35,59,61,100,110,124,151],"inherent":[19],"brittleness":[20],"their":[22,57],"static":[23,49],"boundaries,":[25],"LLMs":[26],"are":[27,168],"vulnerable":[28],"jailbreak":[30,77],"prompts,":[31],"i.e.adversarial":[32],"inputs":[33],"designed":[34],"bypass":[36],"safeguards":[37],"and":[38,65,96,114],"induce":[39],"generation":[41],"harmful":[43],"content.Existing":[44],"detection":[45,78,128,139],"mechanisms":[46],"rely":[47],"on":[48],"model":[50,67],"components":[51],"or":[52],"fixed":[53],"decision":[54],"thresholds,":[55],"limiting":[56],"ability":[58],"generalize":[60],"evolving":[62],"attack":[63],"patterns":[64],"continual":[66],"updates.To":[68],"bridge":[69],"this":[70,171],"gap,":[71],"we":[72],"propose":[73],"RLShield,":[74],"a":[75,93,106,116],"dynamic":[76,94],"framework":[79],"that":[80,122,132],"employs":[81],"reinforcement":[82],"learning":[83],"for":[84],"adaptive":[85],"threshold":[86],"selection.RLShield":[87],"incorporates":[88],"three":[89],"key":[90],"innovations:":[91],"(i)":[92],"retrieval":[95],"LLM-based":[97],"rewriting":[98],"module":[99],"simulate":[101],"diverse":[102],"adversarial":[103],"contexts;":[104],"(ii)":[105],"crosslayer":[107],"representation":[108],"analysis":[109],"pinpoint":[111],"safetycritical":[112],"parameters;":[113],"(iii)":[115],"Soft":[117],"Actor-Critic":[118],"(SAC)":[119],"based":[120],"agent":[121],"learns":[123],"predict":[125],"optimal,":[126],"sample-specific":[127],"thresholds.Experimental":[129],"results":[130],"demonstrate":[131],"RLShield":[133],"consistently":[134],"outperforms":[135],"state-of-the-art":[136],"baselines":[137],"in":[138,160],"performance":[140],"while":[141,153],"maintaining":[142],"high":[143],"computational":[144],"efficiency.Notably,":[145],"improves":[147],"F1":[148],"by":[149],"up":[150],"7.3%,":[152],"achieving":[154],"an":[155],"average":[156],"3\u00d7":[158],"gain":[159],"inference":[161],"efficiency":[162],"across":[163],"multiple":[164],"LLM":[165],"backbones.Our":[166],"codes":[167],"available":[169],"at":[170],"website.":[172]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}