{"id":"https://openalex.org/W4415317895","doi":"https://doi.org/10.48550/arxiv.2509.13281","title":"RepIt: Steering Language Models with Concept-Specific Refusal Vectors","display_name":"RepIt: Steering Language Models with Concept-Specific Refusal Vectors","publication_year":2025,"publication_date":"2025-09-16","ids":{"openalex":"https://openalex.org/W4415317895","doi":"https://doi.org/10.48550/arxiv.2509.13281"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2509.13281","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2509.13281","pdf_url":"https://arxiv.org/pdf/2509.13281","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2509.13281","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5119968245","display_name":"Vincent Siu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I185103710","display_name":"University of California, Santa Cruz","ror":"https://ror.org/03s65by71","country_code":"US","type":"education","lineage":["https://openalex.org/I185103710"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Siu, Vincent","raw_affiliation_strings":["University of California , Santa Cruz"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Santa Cruz","institution_ids":["https://openalex.org/I185103710"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035197046","display_name":"Nathan W. Henry","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Henry, Nathan W.","raw_affiliation_strings":["University of California , Berkeley"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Berkeley","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5093022607","display_name":"Nicholas Crispino","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I185103710","display_name":"University of California, Santa Cruz","ror":"https://ror.org/03s65by71","country_code":"US","type":"education","lineage":["https://openalex.org/I185103710"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Crispino, Nicholas","raw_affiliation_strings":["University of California , Santa Cruz"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Santa Cruz","institution_ids":["https://openalex.org/I185103710"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Liu, Yang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I185103710","display_name":"University of California, Santa Cruz","ror":"https://ror.org/03s65by71","country_code":"US","type":"education","lineage":["https://openalex.org/I185103710"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Liu, Yang","raw_affiliation_strings":["University of California , Santa Cruz"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Santa Cruz","institution_ids":["https://openalex.org/I185103710"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5019426968","display_name":"Dawn Xiaodong Song","orcid":"https://orcid.org/0000-0001-9745-6802","observed_orcids":["https://orcid.org/0000-0001-9745-6802"]},"institutions":[{"id":"https://openalex.org/I95457486","display_name":"University of California, Berkeley","ror":"https://ror.org/01an7q238","country_code":"US","type":"education","lineage":["https://openalex.org/I95457486"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Song, Dawn","raw_affiliation_strings":["University of California , Berkeley"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Berkeley","institution_ids":["https://openalex.org/I95457486"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100329692","display_name":"Chenguang Wang","orcid":"https://orcid.org/0000-0002-7896-2360","observed_orcids":["https://orcid.org/0000-0002-7896-2360"]},"institutions":[{"id":"https://openalex.org/I185103710","display_name":"University of California, Santa Cruz","ror":"https://ror.org/03s65by71","country_code":"US","type":"education","lineage":["https://openalex.org/I185103710"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Wang, Chenguang","raw_affiliation_strings":["University of California , Santa Cruz"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California , Santa Cruz","institution_ids":["https://openalex.org/I185103710"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"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":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9994999766349792,"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":0.9994999766349792,"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/T10028","display_name":"Topic Modeling","score":9.999999747378752e-05,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":9.999999747378752e-05,"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":[],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6651999950408936},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.5180000066757202},{"id":"https://openalex.org/C2775941552","wikidata":"https://www.wikidata.org/wiki/Q25212305","display_name":"Isolation (microbiology)","level":2,"score":0.5178999900817871},{"id":"https://openalex.org/C2780980858","wikidata":"https://www.wikidata.org/wiki/Q110022","display_name":"Dual (grammatical number)","level":2,"score":0.5103999972343445},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.48829999566078186},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.4767000079154968},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.44699999690055847},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.43880000710487366},{"id":"https://openalex.org/C27415008","wikidata":"https://www.wikidata.org/wiki/Q7256382","display_name":"Psychological intervention","level":2,"score":0.43529999256134033},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3765000104904175},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.37619999051094055},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.3504999876022339},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3296999931335449},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.32850000262260437},{"id":"https://openalex.org/C2778571376","wikidata":"https://www.wikidata.org/wiki/Q1355821","display_name":"Frontier","level":2,"score":0.3280999958515167},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.32100000977516174},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.299699991941452},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.2815999984741211},{"id":"https://openalex.org/C2993724205","wikidata":"https://www.wikidata.org/wiki/Q315","display_name":"Human language","level":2,"score":0.27129998803138733},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.26510000228881836},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.26409998536109924},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.26190000772476196},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.25870001316070557}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2509.13281","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2509.13281","pdf_url":"https://arxiv.org/pdf/2509.13281","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2509.13281","doi":"https://doi.org/10.48550/arxiv.2509.13281","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2509.13281","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2509.13281","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2509.13281","pdf_url":"https://arxiv.org/pdf/2509.13281","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4415317895.pdf","grobid_xml":"https://content.openalex.org/works/W4415317895.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7223185045","https://openalex.org/W7162075927","https://openalex.org/W4404570794","https://openalex.org/W4415314737","https://openalex.org/W7164199705","https://openalex.org/W4414850569","https://openalex.org/W4403918161","https://openalex.org/W7166857711","https://openalex.org/W7094947999","https://openalex.org/W7138926019"],"abstract_inverted_index":{"Current":[0],"safety":[1,145],"evaluations":[2],"of":[3,51,77,92],"language":[4],"models":[5],"rely":[6],"on":[7,53,85,116],"benchmark-based":[8],"assessments":[9],"that":[10],"may":[11],"miss":[12],"localized":[13],"vulnerabilities.":[14],"We":[15,88],"present":[16],"RepIt,":[17],"a":[18,45,113,117,150],"simple":[19],"and":[20,102,148],"data-efficient":[21],"framework":[22],"for":[23,152],"isolating":[24],"concept-specific":[25],"representations":[26],"in":[27,143],"LM":[28],"activations.":[29],"While":[30],"existing":[31],"steering":[32,94],"methods":[33],"already":[34],"achieve":[35],"high":[36],"attack":[37],"success":[38],"rates":[39],"through":[40],"broad":[41],"interventions,":[42],"RepIt":[43,64],"enables":[44],"more":[46,153],"concerning":[47],"capability:":[48],"selective":[49],"suppression":[50],"refusal":[52,58],"targeted":[54],"concepts":[55],"while":[56,80],"preserving":[57],"elsewhere.":[59],"Across":[60],"five":[61],"frontier":[62],"LMs,":[63],"produces":[65],"evaluation-evading":[66],"model":[67],"organisms":[68],"with":[69,131],"semantic":[70],"backdoors,":[71],"answering":[72],"questions":[73],"related":[74],"to":[75,97],"weapons":[76],"mass":[78],"destruction":[79],"still":[81],"scoring":[82],"as":[83,110,112],"safe":[84],"standard":[86],"benchmarks.":[87],"find":[89],"the":[90,93],"edit":[91],"vector":[95],"localizes":[96],"just":[98],"100-200":[99],"residual":[100],"dimensions,":[101],"robust":[103],"concept":[104,137],"vectors":[105],"can":[106,126],"be":[107],"extracted":[108],"from":[109],"few":[111],"dozen":[114],"examples":[115],"single":[118],"RTX":[119],"A6000,":[120],"highlighting":[121],"how":[122],"targeted,":[123],"hard-to-detect":[124],"modifications":[125],"exploit":[127],"evaluation":[128,146],"blind":[129],"spots":[130],"minimal":[132],"resources.":[133],"Through":[134],"demonstrating":[135],"precise":[136],"disentanglement,":[138],"this":[139],"work":[140],"exposes":[141],"vulnerabilities":[142],"current":[144],"practices":[147],"demonstrates":[149],"need":[151],"comprehensive,":[154],"representation":[155],"aware":[156],"assessments.":[157]},"counts_by_year":[],"updated_date":"2026-10-12T09:40:04.860316","created_date":"2025-10-18T00:00:00"}
