{"id":"https://openalex.org/W4412945329","doi":"https://doi.org/10.18653/v1/2025.acl-long.738","title":"Why Safeguarded Ships Run Aground? Aligned Large Language Models\u2019 Safety Mechanisms Tend to Be Anchored in The Template Region","display_name":"Why Safeguarded Ships Run Aground? Aligned Large Language Models\u2019 Safety Mechanisms Tend to Be Anchored in The Template Region","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945329","doi":"https://doi.org/10.18653/v1/2025.acl-long.738"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.738","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.738","pdf_url":"https://aclanthology.org/2025.acl-long.738.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 63rd 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/2025.acl-long.738.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5009277283","display_name":"Chak Tou Leong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chak Tou Leong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100610166","display_name":"Qingyu Yin","orcid":"https://orcid.org/0009-0006-1129-4704"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qingyu Yin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100370357","display_name":"Jian Wang","orcid":"https://orcid.org/0000-0001-7430-1218"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jian Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5100408983","display_name":"Wenjie Li","orcid":"https://orcid.org/0000-0002-7360-8864"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wenjie Li","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":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"15212","last_page":"15229"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11622","display_name":"Maritime Navigation and Safety","score":0.9606000185012817,"subfield":{"id":"https://openalex.org/subfields/2212","display_name":"Ocean Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11622","display_name":"Maritime Navigation and Safety","score":0.9606000185012817,"subfield":{"id":"https://openalex.org/subfields/2212","display_name":"Ocean Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11357","display_name":"Risk and Safety Analysis","score":0.9279000163078308,"subfield":{"id":"https://openalex.org/subfields/1804","display_name":"Statistics, Probability and Uncertainty"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.595942497253418},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.33108362555503845}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.595942497253418},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.33108362555503845}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.738","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.738","pdf_url":"https://aclanthology.org/2025.acl-long.738.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.738","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.738","pdf_url":"https://aclanthology.org/2025.acl-long.738.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412945329.pdf","grobid_xml":"https://content.openalex.org/works/W4412945329.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"The":[0],"safety":[1,72,80,92,118,139],"alignment":[2,93,140],"of":[3],"large":[4],"language":[5],"models":[6],"(LLMs)":[7],"remains":[8],"vulnerable,":[9],"as":[10,78],"their":[11,52],"initial":[12,32],"behavior":[13],"can":[14],"be":[15],"easily":[16],"jailbroken":[17],"by":[18],"even":[19],"relatively":[20],"simple":[21],"attacks.Since":[22],"infilling":[23],"a":[24,36,48],"fixed":[25],"template":[26,46,65,122,147],"between":[27],"the":[28,60,64,121,146],"input":[29],"instruction":[30],"and":[31,88],"model":[33],"output":[34],"is":[35,47,94,124],"common":[37],"practice":[38],"for":[39],"existing":[40],"LLMs,":[41],"we":[42,84,114],"hypothesize":[43],"that":[44,90,116,142],"this":[45,76,82],"key":[49],"factor":[50],"behind":[51],"vulnerabilities:":[53],"LLMs'":[54],"safety-related":[55],"decision-making":[56],"overly":[57],"relies":[58],"on":[59,145],"aggregated":[61],"information":[62],"from":[63,120],"region,":[66],"which":[67],"largely":[68],"influences":[69],"these":[70],"models'":[71,107],"behavior.We":[73],"refer":[74],"to":[75,106,129,135],"issue":[77],"template-anchored":[79,91],"alignment.In":[81],"paper,":[83],"conduct":[85],"extensive":[86],"experiments":[87],"verify":[89],"widespread":[95],"across":[96],"various":[97],"aligned":[98],"LLMs.Our":[99],"mechanistic":[100],"analyses":[101],"demonstrate":[102],"how":[103],"it":[104],"leads":[105],"susceptibility":[108],"when":[109],"encountering":[110],"inference-time":[111],"jailbreak":[112,130],"attacks.Furthermore,":[113],"show":[115],"detaching":[117],"mechanisms":[119],"region":[123],"promising":[125],"in":[126],"mitigating":[127],"vulnerabilities":[128],"attacks.We":[131],"encourage":[132],"future":[133],"research":[134],"develop":[136],"more":[137],"robust":[138],"techniques":[141],"reduce":[143],"reliance":[144],"region.":[148]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}