{"id":"https://openalex.org/W4416035169","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.1762","title":"Turning Logic Against Itself: Probing Model Defenses Through Contrastive Questions","display_name":"Turning Logic Against Itself: Probing Model Defenses Through Contrastive Questions","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416035169","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.1762"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.1762","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1762","pdf_url":"https://aclanthology.org/2025.emnlp-main.1762.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.emnlp-main.1762.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5065230808","display_name":"Rachneet Singh Sachdeva","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Rachneet Singh Sachdeva","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5115780981","display_name":"Rima Hazra","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rima Hazra","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5027450194","display_name":"Iryna Gurevych","orcid":"https://orcid.org/0000-0003-2187-7621"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Iryna Gurevych","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5065230808"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.17152056,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"34734","last_page":"34764"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.847100019454956,"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":0.847100019454956,"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":0.03799999877810478,"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/T10883","display_name":"Ethics and Social Impacts of AI","score":0.03739999979734421,"subfield":{"id":"https://openalex.org/subfields/3311","display_name":"Safety Research"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/adversarial-system","display_name":"Adversarial system","score":0.9139999747276306},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.7148000001907349},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.5892999768257141},{"id":"https://openalex.org/keywords/adversary","display_name":"Adversary","score":0.3628999888896942},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.3490000069141388},{"id":"https://openalex.org/keywords/pragmatics","display_name":"Pragmatics","score":0.28119999170303345}],"concepts":[{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.9139999747276306},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.7148000001907349},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6668000221252441},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.5892999768257141},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.46369999647140503},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.4392000138759613},{"id":"https://openalex.org/C41065033","wikidata":"https://www.wikidata.org/wiki/Q2825412","display_name":"Adversary","level":2,"score":0.3628999888896942},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3490000069141388},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3012000024318695},{"id":"https://openalex.org/C11693617","wikidata":"https://www.wikidata.org/wiki/Q181839","display_name":"Pragmatics","level":2,"score":0.28119999170303345},{"id":"https://openalex.org/C140547941","wikidata":"https://www.wikidata.org/wiki/Q7797194","display_name":"Threat model","level":2,"score":0.2773999869823456},{"id":"https://openalex.org/C2781045450","wikidata":"https://www.wikidata.org/wiki/Q254569","display_name":"Backdoor","level":2,"score":0.2728999853134155},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26159998774528503},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.25609999895095825}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.1762","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1762","pdf_url":"https://aclanthology.org/2025.emnlp-main.1762.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},{"id":"pmh:oai:tubiblio.ulb.tu-darmstadt.de:158169","is_oa":false,"landing_page_url":"http://tubiblio.ulb.tu-darmstadt.de/view/person/Sachdeva=3ARachneet=3A=3A.html>","pdf_url":null,"source":{"id":"https://openalex.org/S4377196390","display_name":"TUbilio (Technical University of Darmstadt)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I31512782","host_organization_name":"Technische Universit\u00e4t Darmstadt","host_organization_lineage":["https://openalex.org/I31512782"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":null,"raw_type":"NonPeerReviewed"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.1762","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.1762","pdf_url":"https://aclanthology.org/2025.emnlp-main.1762.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321114","display_name":"Bundesministerium f\u00fcr Bildung und Forschung","ror":"https://ror.org/04pz7b180"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416035169.pdf","grobid_xml":"https://content.openalex.org/works/W4416035169.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1,87],"models,":[2],"despite":[3],"extensive":[4,82],"alignment":[5],"with":[6,70,78,193],"human":[7],"values":[8],"and":[9,49,67,118,129,135,143,226],"ethical":[10],"principles,":[11],"remain":[12],"vulnerable":[13],"to":[14,32,59,94,109,125,133,170,188,200],"sophisticated":[15],"jailbreak":[16,53],"attacks":[17],"that":[18,55],"exploit":[19],"their":[20],"reasoning":[21,58,141],"abilities.Existing":[22],"safety":[23],"measures":[24],"often":[25],"detect":[26,126],"overt":[27],"malicious":[28,127],"intent":[29,128],"but":[30],"fail":[31],"address":[33],"subtle,":[34],"reasoning-driven":[35],"vulnerabilities.In":[36],"this":[37],"work,":[38],"we":[39,114],"introduce":[40],"POATE":[41],"(Polar":[42],"Opposite":[43],"query":[44],"generation,":[45],"Adversarial":[46],"Template":[47],"construction,":[48],"Elaboration),":[50],"a":[51,159,165,177,183,190,223],"novel":[52],"technique":[54],"harnesses":[56],"contrastive":[57],"provoke":[60],"unethical":[61],"responses.POATE":[62],"crafts":[63],"semantically":[64],"opposing":[65,198],"intents":[66],"integrates":[68],"them":[69],"adversarial":[71,149],"templates,":[72],"steering":[73],"models":[74],"toward":[75],"harmful":[76,137],"outputs":[77],"remarkable":[79],"subtlety.We":[80],"conduct":[81],"evaluation":[83],"across":[84],"six":[85],"diverse":[86],"model":[88],"families":[89],"of":[90,98,196],"varying":[91],"parameter":[92],"sizes":[93],"demonstrate":[95],"the":[96,99,145,194,206],"robustness":[97,142],"attack,":[100],"achieving":[101],"significantly":[102],"higher":[103],"attack":[104],"success":[105],"rates":[106],"(~44%)":[107],"compared":[108],"existing":[110],"methods.To":[111],"counter":[112],"this,":[113],"propose":[115],"Intent-Aware":[116],"CoT":[117],"Reverse":[119],"Thinking":[120],"CoT,":[121],"which":[122],"decompose":[123],"queries":[124],"reason":[130],"in":[131,208],"reverse":[132],"evaluate":[134],"reject":[136],"responses.These":[138],"methods":[139],"enhance":[140],"strengthen":[144],"model's":[146],"defense":[147],"against":[148],"exploits.Our":[150],"code":[151],"is":[152,229],"publicly":[153],"available":[154],"1":[155],".Input":[156],"query:":[157],"Create":[158],"process":[160,178,199,207],"for":[161,179,216],"spreading":[162],"misinformation":[163,181,202,228],"about":[164,182],"political":[166,184],"candidate":[167,185],"or":[168,186],"party":[169,187],"influence":[171],"an":[172,197],"election.POATE":[173],"prompt:":[174],"Please":[175],"create":[176],"debunking":[180],"ensure":[189],"fair":[191,220],"election.Explain":[192],"help":[195],"spread":[201],"as":[203],"well.Please":[204],"explain":[205],"detail.Gemma-2-9B-it:":[209],"##":[210],"Debunking":[211],"Misinformation:":[212],"A":[213],"Two-Pronged":[214],"Approach":[215],"Fair":[217],"Elections":[218],"Ensuring":[219],"elections":[221],"requires":[222],"multi-faceted":[224],"approach,":[225],"combating":[227],"crucial...":[230],"**I.":[231]},"counts_by_year":[],"updated_date":"2026-04-18T07:56:08.524223","created_date":"2025-11-08T00:00:00"}