{"id":"https://openalex.org/W7169587485","doi":"https://doi.org/10.1145/3803437.3806718","title":"LLMSafeGuard: A Training-Free Framework for Safeguarding LLM Decoding via Context-Wise Similarity Validation","display_name":"LLMSafeGuard: A Training-Free Framework for Safeguarding LLM Decoding via Context-Wise Similarity Validation","publication_year":2026,"publication_date":"2026-07-05","ids":{"openalex":"https://openalex.org/W7169587485","doi":"https://doi.org/10.1145/3803437.3806718"},"language":null,"primary_location":{"id":"doi:10.1145/3803437.3806718","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3803437.3806718","pdf_url":null,"source":null,"license":"cc-by-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 34th ACM International Conference on the Foundations of Software Engineering","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3803437.3806718","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5113198167","display_name":"Ximing Dong","orcid":null},"institutions":[{"id":"https://openalex.org/I4210115038","display_name":"Huawei Technologies (Canada)","ror":"https://ror.org/026venb53","country_code":"CA","type":"company","lineage":["https://openalex.org/I2250955327","https://openalex.org/I4210115038"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Ximing Dong","raw_affiliation_strings":["Centre for Software Excellence, Huawei Canada, Toronto, Ontario, Canada"],"raw_orcid":"https://orcid.org/0009-0006-7569-0650","affiliations":[{"raw_affiliation_string":"Centre for Software Excellence, Huawei Canada, Toronto, Ontario, Canada","institution_ids":["https://openalex.org/I4210115038"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100664833","display_name":"Shaowei Wang","orcid":"https://orcid.org/0000-0003-3823-1771"},"institutions":[{"id":"https://openalex.org/I46247651","display_name":"University of Manitoba","ror":"https://ror.org/02gfys938","country_code":"CA","type":"education","lineage":["https://openalex.org/I46247651"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Shaowei Wang","raw_affiliation_strings":["University of Manitoba, Winnipeg, Manitoba, Canada"],"raw_orcid":"https://orcid.org/0000-0003-3823-1771","affiliations":[{"raw_affiliation_string":"University of Manitoba, Winnipeg, Manitoba, Canada","institution_ids":["https://openalex.org/I46247651"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079401267","display_name":"Dayi Lin","orcid":"https://orcid.org/0000-0002-4034-6650"},"institutions":[{"id":"https://openalex.org/I4210115038","display_name":"Huawei Technologies (Canada)","ror":"https://ror.org/026venb53","country_code":"CA","type":"company","lineage":["https://openalex.org/I2250955327","https://openalex.org/I4210115038"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Dayi Lin","raw_affiliation_strings":["Centre for Software Excellence, Huawei Canada, Toronto, Ontario, Canada"],"raw_orcid":"https://orcid.org/0000-0002-4034-6650","affiliations":[{"raw_affiliation_string":"Centre for Software Excellence, Huawei Canada, Toronto, Ontario, Canada","institution_ids":["https://openalex.org/I4210115038"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5091586373","display_name":"Ahmed E. Hassan","orcid":"https://orcid.org/0000-0001-7749-5513"},"institutions":[{"id":"https://openalex.org/I204722609","display_name":"Queen's University","ror":"https://ror.org/02y72wh86","country_code":"CA","type":"education","lineage":["https://openalex.org/I204722609"]}],"countries":["CA"],"is_corresponding":false,"raw_author_name":"Ahmed E. Hassan","raw_affiliation_strings":["Queen's University, Kingston, Ontario, Canada"],"raw_orcid":"https://orcid.org/0000-0001-7749-5513","affiliations":[{"raw_affiliation_string":"Queen's University, Kingston, Ontario, Canada","institution_ids":["https://openalex.org/I204722609"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1713","last_page":"1720"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":null,"topics":[],"keywords":[{"id":"https://openalex.org/keywords/safeguarding","display_name":"Safeguarding","score":0.5971999764442444},{"id":"https://openalex.org/keywords/validator","display_name":"Validator","score":0.5425999760627747},{"id":"https://openalex.org/keywords/selection","display_name":"Selection (genetic algorithm)","score":0.5358999967575073},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5231000185012817},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.5127999782562256},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.44589999318122864},{"id":"https://openalex.org/keywords/constraint","display_name":"Constraint (computer-aided design)","score":0.4041000008583069},{"id":"https://openalex.org/keywords/similarity","display_name":"Similarity (geometry)","score":0.4016999900341034},{"id":"https://openalex.org/keywords/control","display_name":"Control (management)","score":0.3571999967098236}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7208999991416931},{"id":"https://openalex.org/C2776743756","wikidata":"https://www.wikidata.org/wiki/Q5097921","display_name":"Safeguarding","level":2,"score":0.5971999764442444},{"id":"https://openalex.org/C35292069","wikidata":"https://www.wikidata.org/wiki/Q1575458","display_name":"Validator","level":2,"score":0.5425999760627747},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.5358999967575073},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5231000185012817},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.5127999782562256},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.44589999318122864},{"id":"https://openalex.org/C2776036281","wikidata":"https://www.wikidata.org/wiki/Q48769818","display_name":"Constraint (computer-aided design)","level":2,"score":0.4041000008583069},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.4016999900341034},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37369999289512634},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3736000061035156},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.3571999967098236},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.35589998960494995},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.35359999537467957},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.33410000801086426},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.32269999384880066},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.31150001287460327},{"id":"https://openalex.org/C40423286","wikidata":"https://www.wikidata.org/wiki/Q284172","display_name":"Selection bias","level":2,"score":0.3111000061035156},{"id":"https://openalex.org/C2780665704","wikidata":"https://www.wikidata.org/wiki/Q959298","display_name":"Intervention (counseling)","level":2,"score":0.30570000410079956},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.29809999465942383},{"id":"https://openalex.org/C93959086","wikidata":"https://www.wikidata.org/wiki/Q6888345","display_name":"Model selection","level":2,"score":0.29350000619888306},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.28029999136924744},{"id":"https://openalex.org/C17923572","wikidata":"https://www.wikidata.org/wiki/Q7250160","display_name":"Propensity score matching","level":2,"score":0.274399995803833},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.27320000529289246},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.27079999446868896},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2630999982357025},{"id":"https://openalex.org/C12174686","wikidata":"https://www.wikidata.org/wiki/Q1058438","display_name":"Risk assessment","level":2,"score":0.26190000772476196},{"id":"https://openalex.org/C2779182362","wikidata":"https://www.wikidata.org/wiki/Q17126187","display_name":"Session (web analytics)","level":2,"score":0.2599000036716461},{"id":"https://openalex.org/C2779662365","wikidata":"https://www.wikidata.org/wiki/Q5416694","display_name":"Event (particle physics)","level":2,"score":0.2549999952316284},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2533999979496002},{"id":"https://openalex.org/C2775973920","wikidata":"https://www.wikidata.org/wiki/Q3252726","display_name":"Selection algorithm","level":3,"score":0.25049999356269836}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3803437.3806718","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3803437.3806718","pdf_url":null,"source":null,"license":"cc-by-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 34th ACM International Conference on the Foundations of Software Engineering","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3803437.3806718","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3803437.3806718","pdf_url":null,"source":null,"license":"cc-by-nc-nd","license_id":"https://openalex.org/licenses/cc-by-nc-nd","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 34th ACM International Conference on the Foundations of Software Engineering","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":10,"referenced_works":["https://openalex.org/W3158172841","https://openalex.org/W3176618728","https://openalex.org/W3202712981","https://openalex.org/W4221144473","https://openalex.org/W4382246105","https://openalex.org/W4386290290","https://openalex.org/W4389520705","https://openalex.org/W4389617257","https://openalex.org/W4402667010","https://openalex.org/W4402671039"],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4,29],"significantly":[5],"advanced":[6],"natural":[7],"language":[8],"processing":[9],"(NLP)":[10],"tasks,":[11],"but":[12],"also":[13],"pose":[14],"ethical":[15],"and":[16,39,50,89,125,132],"societal":[17],"risks":[18],"due":[19],"to":[20,23,47,109,181],"their":[21],"propensity":[22],"generate":[24],"harmful":[25,167],"content.":[26],"Existing":[27],"methods":[28],"limitations,":[30,57],"including":[31],"the":[32,91,133,166,170,182],"need":[33,92],"for":[34,93],"training":[35],"specific":[36],"control":[37,95],"models":[38],"proactive":[40],"intervention":[41],"during":[42],"text":[43],"generation,":[44],"which":[45,106],"leads":[46],"quality":[48],"degradation":[49],"increased":[51],"computational":[52],"overhead.":[53],"To":[54],"mitigate":[55],"those":[56],"we":[58],"propose":[59],"LLMSafeGuard,":[60],"a":[61,81,101,191],"lightweight":[62],"real-time":[63],"framework":[64],"that":[65],"integrates":[66],"an":[67],"external":[68,94],"validator":[69],"into":[70],"decoding,":[71],"rejecting":[72],"unsafe":[73],"outputs":[74],"while":[75,156],"allowing":[76],"valid":[77],"ones.":[78],"We":[79,116],"introduce":[80],"similarity-based":[82],"validation":[83,88],"approach,":[84],"simplifying":[85],"safety":[86],"constraint":[87],"eliminating":[90],"model":[96],"training.":[97],"Additionally,":[98,185],"LLMSafeGuard":[99,108,118,148,164],"employs":[100],"context-wise":[102,187],"timing":[103,188],"selection":[104,189],"strategy,":[105],"allows":[107],"intervene":[110],"in":[111,146],"LLMs":[112,172],"only":[113],"when":[114],"necessary.":[115],"evaluate":[117],"on":[119,141],"three":[120],"tasks:":[121],"detoxification,":[122,147],"copyright":[123],"safeguarding,":[124],"jailbreak":[126,174],"attack":[127,162,175],"defending":[128],"across":[129],"six":[130],"models,":[131],"results":[134],"demonstrate":[135],"its":[136,186],"superiority":[137],"over":[138,195],"SOTA":[139],"baselines":[140],"all":[142],"tasks.":[143],"For":[144,160],"instance,":[145],"reduces":[149,165],"toxic":[150],"output":[151],"by":[152,178],"at":[153,176],"least":[154,177],"38.6%":[155],"preserving":[157],"linguistic":[158],"quality.":[159],"Jailbreak":[161],"defending,":[163],"score":[168],"of":[169,193],"response":[171],"under":[173],"11.4%":[179],"compared":[180],"best":[183],"baseline.":[184],"achieves":[190],"speedup":[192],"1.7X":[194],"per-step":[196],"validation,":[197],"without":[198],"compromising":[199],"effectiveness.":[200]},"counts_by_year":[],"updated_date":"2026-07-18T05:58:53.570196","created_date":"2026-07-18T00:00:00"}