{"id":"https://openalex.org/W4412888121","doi":"https://doi.org/10.18653/v1/2025.findings-acl.885","title":"Beyond the Tip of Efficiency: Uncovering the Submerged Threats of Jailbreak Attacks in Small Language Models","display_name":"Beyond the Tip of Efficiency: Uncovering the Submerged Threats of Jailbreak Attacks in Small Language Models","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888121","doi":"https://doi.org/10.18653/v1/2025.findings-acl.885"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.885","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.885","pdf_url":"https://aclanthology.org/2025.findings-acl.885.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 2025","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.885.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5112529653","display_name":"Sibo Yi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sibo Yi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5002440210","display_name":"Tianshuo Cong","orcid":"https://orcid.org/0000-0003-3189-8223"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tianshuo Cong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101911145","display_name":"Xinlei He","orcid":"https://orcid.org/0000-0001-5725-2848"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinlei He","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035868158","display_name":"Qi Li","orcid":"https://orcid.org/0000-0002-7280-7378"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qi Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5056116846","display_name":"Jiaxing Song","orcid":"https://orcid.org/0009-0008-5366-4675"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiaxing Song","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.7588,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.87754215,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"17221","last_page":"17234"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.9142000079154968,"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/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.9142000079154968,"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/computer-science","display_name":"Computer science","score":0.5702410936355591},{"id":"https://openalex.org/keywords/computer-security","display_name":"Computer security","score":0.4438212215900421}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5702410936355591},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.4438212215900421}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.885","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.885","pdf_url":"https://aclanthology.org/2025.findings-acl.885.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 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.885","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.885","pdf_url":"https://aclanthology.org/2025.findings-acl.885.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 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16","score":0.4099999964237213}],"awards":[],"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/W4412888121.pdf","grobid_xml":"https://content.openalex.org/works/W4412888121.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":{"Small":[0],"language":[1,52],"models":[2,53],"(SLMs)":[3],"have":[4,44],"become":[5],"increasingly":[6],"prominent":[7],"in":[8,114,158],"the":[9,27,39,66,101,116,122,146],"deployment":[10],"on":[11],"edge":[12,168],"devices":[13],"due":[14],"to":[15,25,50,64,85,96,155],"their":[16,112],"high":[17],"efficiency":[18],"and":[19,35,110,137,151,162],"low":[20],"computational":[21],"cost.While":[22],"researchers":[23],"continue":[24],"advance":[26],"capabilities":[28],"of":[29,42,69,91,118,149],"SLMs":[30,43,72,81,150],"through":[31],"innovative":[32],"training":[33],"strategies":[34],"model":[36],"compression":[37],"techniques,":[38],"security":[40,67,117,124,147],"risks":[41],"received":[45],"considerably":[46],"less":[47],"attention":[48],"compared":[49],"large":[51],"(LLMs).To":[54],"fill":[55],"this":[56],"gap,":[57],"we":[58,104],"provide":[59,152],"a":[60],"comprehensive":[61],"empirical":[62],"study":[63],"evaluate":[65,105],"performance":[68],"13":[70],"state-of-the-art":[71],"under":[73],"various":[74],"jailbreak":[75,87],"attacks.Our":[76],"experiments":[77],"demonstrate":[78,111],"that":[79,141],"most":[80],"are":[82,93],"quite":[83],"susceptible":[84],"existing":[86],"attacks,":[88],"while":[89],"some":[90],"them":[92],"even":[94],"vulnerable":[95],"direct":[97],"harmful":[98],"prompts.To":[99],"address":[100],"safety":[102],"concerns,":[103],"several":[106],"representative":[107],"defense":[108],"methods":[109],"effectiveness":[113],"enhancing":[115],"SLMs.We":[119],"further":[120],"analyze":[121],"potential":[123],"degradation":[125],"caused":[126],"by":[127],"different":[128],"SLM":[129],"techniques":[130],"including":[131],"architecture":[132],"compression,":[133],"quantization,":[134],"knowledge":[135],"distillation,":[136],"so":[138],"on.We":[139],"expect":[140],"our":[142],"research":[143],"can":[144],"highlight":[145],"challenges":[148],"valuable":[153],"insights":[154],"future":[156],"work":[157],"developing":[159],"more":[160],"robust":[161],"secure":[163],"SLMs.Llama3.2-1BLlama3.2-3BLlama":[164],"Family":[165,176],"MobiLlama-1B":[166],"MobileLlama-1.4BTinyLlama-1.1BKnowl":[167],"Distilla":[169],"tion":[170],"Param":[171],"eter":[172],"Sharin":[173],"g":[174],"Phi":[175],"Llama3.2-1BPhi-3-3":[177]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}