{"id":"https://openalex.org/W4412888268","doi":"https://doi.org/10.18653/v1/2025.findings-acl.613","title":"DeTAM: Defending LLMs Against Jailbreak Attacks via Targeted Attention Modification","display_name":"DeTAM: Defending LLMs Against Jailbreak Attacks via Targeted Attention Modification","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412888268","doi":"https://doi.org/10.18653/v1/2025.findings-acl.613"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.613","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.613","pdf_url":"https://aclanthology.org/2025.findings-acl.613.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.613.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101983825","display_name":"Li Yu","orcid":"https://orcid.org/0000-0002-2304-3770"},"institutions":[{"id":"https://openalex.org/I116953780","display_name":"Tongji University","ror":"https://ror.org/03rc6as71","country_code":"CN","type":"education","lineage":["https://openalex.org/I116953780"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yu Li","raw_affiliation_strings":["Department of Computer Science and Technology Tongji University Shanghai , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology Tongji University Shanghai , China","institution_ids":["https://openalex.org/I116953780"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5080281559","display_name":"Han Jiang","orcid":"https://orcid.org/0000-0003-3222-4140"},"institutions":[{"id":"https://openalex.org/I116953780","display_name":"Tongji University","ror":"https://ror.org/03rc6as71","country_code":"CN","type":"education","lineage":["https://openalex.org/I116953780"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Han Jiang","raw_affiliation_strings":["Department of Computer Science and Technology Tongji University Shanghai , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology Tongji University Shanghai , China","institution_ids":["https://openalex.org/I116953780"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5004301353","display_name":"Zhihua Wei","orcid":"https://orcid.org/0000-0002-5937-3907"},"institutions":[{"id":"https://openalex.org/I116953780","display_name":"Tongji University","ror":"https://ror.org/03rc6as71","country_code":"CN","type":"education","lineage":["https://openalex.org/I116953780"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhihua Wei","raw_affiliation_strings":["Department of Computer Science and Technology Tongji University Shanghai , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Technology Tongji University Shanghai , China","institution_ids":["https://openalex.org/I116953780"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I116953780"],"apc_list":null,"apc_paid":null,"fwci":1.6508,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.87212743,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"11781","last_page":"11797"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.994700014591217,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.994700014591217,"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":0.9846000075340271,"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/T11241","display_name":"Advanced Malware Detection Techniques","score":0.902999997138977,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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-security","display_name":"Computer security","score":0.49983692169189453},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.4138021469116211},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.3329184055328369},{"id":"https://openalex.org/keywords/internet-privacy","display_name":"Internet privacy","score":0.33182668685913086}],"concepts":[{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.49983692169189453},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4138021469116211},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.3329184055328369},{"id":"https://openalex.org/C108827166","wikidata":"https://www.wikidata.org/wiki/Q175975","display_name":"Internet privacy","level":1,"score":0.33182668685913086}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.613","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.613","pdf_url":"https://aclanthology.org/2025.findings-acl.613.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.613","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.613","pdf_url":"https://aclanthology.org/2025.findings-acl.613.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":[{"id":"https://metadata.un.org/sdg/5","display_name":"Gender equality","score":0.5}],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412888268.pdf","grobid_xml":"https://content.openalex.org/works/W4412888268.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":{"Warning:":[0],"This":[1],"paper":[2],"contains":[3],"examples":[4],"of":[5,11,75],"harmful":[6,32],"outputs.With":[7],"the":[8,69,83,95,148],"widespread":[9],"adoption":[10],"Large":[12],"Language":[13],"Models":[14],"(LLMs),":[15],"jailbreak":[16,73,100,125,142],"attacks":[17,74,133],"have":[18],"become":[19],"an":[20],"increasingly":[21],"pressing":[22],"safety":[23],"concern.While":[24],"safety-aligned":[25],"LLMs":[26,76],"can":[27],"effectively":[28],"defend":[29],"against":[30,72],"normal":[31],"queries,":[33],"they":[34],"remain":[35],"vulnerable":[36],"to":[37,93,99,106],"such":[38],"attacks.Existing":[39],"defense":[40,65,126],"methods":[41],"primarily":[42],"rely":[43],"on":[44,140,150],"finetuning":[45],"or":[46],"input":[47],"modification,":[48],"which":[49],"often":[50],"suffer":[51],"from":[52,113],"limited":[53],"generalization":[54,130],"and":[55,90,127,134,160],"reduced":[56],"utility.To":[57],"address":[58],"this,":[59],"we":[60,81,103,144],"introduce":[61],"DETAM,":[62],"a":[63],"finetuning-free":[64],"approach":[66],"that":[67,119],"improves":[68],"defensive":[70],"capabilities":[71],"via":[77],"targeted":[78],"attention":[79,86,96,105],"modification.Specifically,":[80],"analyze":[82],"differences":[84],"in":[85,124],"scores":[87],"between":[88,158],"successful":[89],"unsuccessful":[91],"defenses":[92],"identify":[94],"heads":[97],"sensitive":[98],"attacks.During":[101],"inference,":[102],"reallocate":[104],"emphasize":[107],"users'":[108],"core":[109],"intentions,":[110],"minimizing":[111],"interference":[112],"attack":[114],"tokens.Our":[115],"experimental":[116],"results":[117],"demonstrate":[118],"DETAM":[120,146],"outperforms":[121],"various":[122],"baselines":[123,149],"exhibits":[128],"robust":[129],"across":[131],"different":[132],"models,":[135],"maintaining":[136],"its":[137,155],"effectiveness":[138],"even":[139],"in-thewild":[141],"data.Furthermore,":[143],"compare":[145],"with":[147],"over-defense":[151],"datasets,":[152],"further":[153],"validating":[154],"superior":[156],"balance":[157],"helpfulness":[159],"harmlessness.":[161]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-06-26T08:34:08.712188","created_date":"2025-10-10T00:00:00"}