{"id":"https://openalex.org/W7166861970","doi":"https://doi.org/10.18653/v1/2026.acl-long.1482","title":"HiddenGuard: Fine-Grained Safe Generation with Specialized Representation Router","display_name":"HiddenGuard: Fine-Grained Safe Generation with Specialized Representation Router","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166861970","doi":"https://doi.org/10.18653/v1/2026.acl-long.1482"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.acl-long.1482","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1482","pdf_url":"https://aclanthology.org/2026.acl-long.1482.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.acl-long.1482.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5093784797","display_name":"Lingrui Mei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lingrui Mei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100677836","display_name":"Shenghua Liu","orcid":"https://orcid.org/0000-0003-2120-3598"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shenghua Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139804582","display_name":"Yiwei Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yiwei Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5094348722","display_name":"Baolong Bi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Baolong Bi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067150934","display_name":"Ruibin Yuan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ruibin Yuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139812510","display_name":"Xueqi Cheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xueqi Cheng","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":0.81242079,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"32124","last_page":"32156"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10743","display_name":"Software Testing and Debugging Techniques","score":0.12409999966621399,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/T10743","display_name":"Software Testing and Debugging Techniques","score":0.12409999966621399,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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":0.07829999923706055,"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/T11598","display_name":"Internet Traffic Analysis and Secure E-voting","score":0.05790000036358833,"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/representation","display_name":"Representation (politics)","score":0.392300009727478},{"id":"https://openalex.org/keywords/router","display_name":"Router","score":0.29600000381469727},{"id":"https://openalex.org/keywords/intersection","display_name":"Intersection (aeronautics)","score":0.2676999866962433},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.26579999923706055},{"id":"https://openalex.org/keywords/troubleshooting","display_name":"Troubleshooting","score":0.2248000055551529}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6384999752044678},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.392300009727478},{"id":"https://openalex.org/C2775896111","wikidata":"https://www.wikidata.org/wiki/Q642560","display_name":"Router","level":2,"score":0.29600000381469727},{"id":"https://openalex.org/C64543145","wikidata":"https://www.wikidata.org/wiki/Q162942","display_name":"Intersection (aeronautics)","level":2,"score":0.2676999866962433},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.26579999923706055},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.24539999663829803},{"id":"https://openalex.org/C147494362","wikidata":"https://www.wikidata.org/wiki/Q2078905","display_name":"Troubleshooting","level":2,"score":0.2248000055551529},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.22470000386238098},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.22450000047683716},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.2207999974489212}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.1482","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1482","pdf_url":"https://aclanthology.org/2026.acl-long.1482.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.1482","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.1482","pdf_url":"https://aclanthology.org/2026.acl-long.1482.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G1907283378","display_name":null,"funder_award_id":"62441229","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G2706053043","display_name":null,"funder_award_id":"62377043","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G8808570173","display_name":null,"funder_award_id":"U25B2076","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":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166861970.pdf","grobid_xml":"https://content.openalex.org/works/W7166861970.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"As":[0],"Large":[1],"Language":[2],"Models":[3],"(LLMs)":[4],"grow":[5],"increasingly":[6],"powerful,":[7],"ensuring":[8],"their":[9],"safety":[10],"and":[11,63,106,171,180],"alignment":[12,32],"with":[13,65,147],"human":[14],"values":[15],"remains":[16],"a":[17,80,144],"critical":[18],"challenge.Ideally,":[19],"LLMs":[20],"should":[21],"provide":[22],"informative":[23,129],"responses":[24,130],"while":[25,131,175],"avoiding":[26],"harmful":[27,109,153,173],"or":[28,41,67,134],"sensitive":[29,136],"disclosures.However,":[30],"current":[31],"strategies,":[33],"reliant":[34],"on":[35],"binary":[36],"refusal":[37],"(e.g.,":[38],"rejecting":[39],"prompts":[40],"coarse":[42],"filtering),":[43],"lack":[44],"nuance.This":[45],"leads":[46],"to":[47,50,60,101,127],"overcensorship,":[48],"failure":[49],"detect":[51],"subtle":[52],"harm":[53],"(like":[54],"withholding":[55],"public":[56],"medication":[57],"information":[58,154],"due":[59],"misuse":[61],"concerns),":[62],"difficulty":[64],"mixed":[66],"context-dependent":[68],"sensitivities,":[69],"often":[70],"overcensoring":[71],"benign":[72],"content.To":[73],"overcome":[74],"these":[75],"challenges,":[76],"we":[77],"introduce":[78],"HIDDENGUARD,":[79],"novel":[81],"framework":[82],"for":[83,93,119,169],"fine-grained,":[84],"safe":[85],"generation":[86],"in":[87,165],"LLMs.HIDDENGUARD":[88],"incorporates":[89],"PRISM":[90],"(Representation":[91],"Router":[92],"In-Stream":[94],"Moderation),":[95],"which":[96],"operates":[97],"alongside":[98],"the":[99,125,177,183],"LLM":[100],"enable":[102],"real-time,":[103],"token-level":[104,148],"detection":[105],"redaction":[107],"of":[108,151,182],"content":[110,174],"by":[111],"leveraging":[112],"intermediate":[113],"hidden":[114],"states.This":[115],"fine-grained":[116],"approach":[117],"allows":[118],"more":[120],"nuanced,":[121],"contextaware":[122],"moderation,":[123],"enabling":[124],"model":[126],"generate":[128],"selectively":[132],"redacting":[133,172],"replacing":[135],"information,":[137],"rather":[138],"than":[139],"outright":[140],"refusal.We":[141],"also":[142],"contribute":[143],"comprehensive":[145],"dataset":[146],"finegrained":[149],"annotations":[150],"potentially":[152],"across":[155],"diverse":[156],"contexts.Our":[157],"experiments":[158],"demonstrate":[159],"that":[160],"HIDDENGUARD":[161],"achieves":[162],"over":[163],"90%":[164],"F":[166],"1":[167],"score":[168],"detecting":[170],"preserving":[176],"overall":[178],"utility":[179],"informativeness":[181],"model's":[184],"responses.":[185]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}