{"id":"https://openalex.org/W6891998366","doi":"https://doi.org/10.48550/arxiv.2508.04281","title":"Reasoning Enhances Robustness to Prompt Injection in LLM-Based Consensus","display_name":"Reasoning Enhances Robustness to Prompt Injection in LLM-Based Consensus","publication_year":2025,"publication_date":"2025-08-06","ids":{"openalex":"https://openalex.org/W6891998366","doi":"https://doi.org/10.48550/arxiv.2508.04281"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2508.04281","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2508.04281","pdf_url":"https://arxiv.org/pdf/2508.04281","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-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","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/2508.04281","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100049953","display_name":"Jairo Gudi\u00f1o","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gudi\u00f1o-Rosero, Jairo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Zambrano, Juan Ignacio","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zambrano, Juan Ignacio","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037466565","display_name":"Umberto Grandi","orcid":"https://orcid.org/0000-0002-1908-5142","observed_orcids":["https://orcid.org/0000-0002-1908-5142"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Grandi, Umberto","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5064297795","display_name":"C\u00e9sar A. Hidalgo","orcid":"https://orcid.org/0000-0002-6977-9492","observed_orcids":["https://orcid.org/0000-0002-6977-9492","https://orcid.org/0000-0002-6031-5982"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hidalgo, C\u00e9sar A.","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":false,"cited_by_count":1,"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":true,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9868000149726868,"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":0.9868000149726868,"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":0.00430000014603138,"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"}},{"id":"https://openalex.org/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.002199999988079071,"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/prompt-injection","display_name":"prompt injection","score":0.9760000109672546},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.9229999780654907},{"id":"https://openalex.org/keywords/digital-democracy","display_name":"digital democracy","score":0.9860000014305115},{"id":"https://openalex.org/keywords/opinion-aggregation","display_name":"opinion aggregation","score":0.6389999985694885},{"id":"https://openalex.org/keywords/adversarial-prompts","display_name":"adversarial prompts","score":0.7360000014305115},{"id":"https://openalex.org/keywords/llm-robustness","display_name":"LLM robustness","score":0.746999979019165},{"id":"https://openalex.org/keywords/political-manifestos","display_name":"political manifestos","score":0.4129999876022339}],"concepts":[{"id":"https://openalex.org/C192562157","wikidata":"https://www.wikidata.org/wiki/Q316694","display_name":"Rhetorical question","level":2,"score":0.660099983215332},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.6010000109672546},{"id":"https://openalex.org/C555826173","wikidata":"https://www.wikidata.org/wiki/Q7174","display_name":"Democracy","level":3,"score":0.5978999733924866},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.5045999884605408},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.4975999891757965},{"id":"https://openalex.org/C190253527","wikidata":"https://www.wikidata.org/wiki/Q295354","display_name":"Law and economics","level":1,"score":0.4519999921321869},{"id":"https://openalex.org/C17744445","wikidata":"https://www.wikidata.org/wiki/Q36442","display_name":"Political science","level":0,"score":0.44369998574256897},{"id":"https://openalex.org/C134698397","wikidata":"https://www.wikidata.org/wiki/Q17946","display_name":"Public opinion","level":3,"score":0.40070000290870667},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.3896999955177307},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.37290000915527344},{"id":"https://openalex.org/C2776552730","wikidata":"https://www.wikidata.org/wiki/Q189656","display_name":"Disinformation","level":3,"score":0.36800000071525574},{"id":"https://openalex.org/C2777363581","wikidata":"https://www.wikidata.org/wiki/Q15098235","display_name":"Harm","level":2,"score":0.36329999566078186},{"id":"https://openalex.org/C2779039943","wikidata":"https://www.wikidata.org/wiki/Q316748","display_name":"Scientific consensus","level":4,"score":0.36239999532699585},{"id":"https://openalex.org/C39549134","wikidata":"https://www.wikidata.org/wiki/Q133080","display_name":"Public relations","level":1,"score":0.3499999940395355},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3400000035762787},{"id":"https://openalex.org/C2780586970","wikidata":"https://www.wikidata.org/wiki/Q1357284","display_name":"Popularity","level":2,"score":0.3346000015735626},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.33219999074935913},{"id":"https://openalex.org/C2776325391","wikidata":"https://www.wikidata.org/wiki/Q6917865","display_name":"Motivated reasoning","level":3,"score":0.296999990940094},{"id":"https://openalex.org/C108827166","wikidata":"https://www.wikidata.org/wiki/Q175975","display_name":"Internet privacy","level":1,"score":0.28600001335144043},{"id":"https://openalex.org/C138921699","wikidata":"https://www.wikidata.org/wiki/Q47555","display_name":"Political economy","level":1,"score":0.2842999994754791},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.28139999508857727},{"id":"https://openalex.org/C144024400","wikidata":"https://www.wikidata.org/wiki/Q21201","display_name":"Sociology","level":0,"score":0.2791000008583069},{"id":"https://openalex.org/C110875604","wikidata":"https://www.wikidata.org/wiki/Q75","display_name":"The Internet","level":2,"score":0.2612999975681305},{"id":"https://openalex.org/C3017605391","wikidata":"https://www.wikidata.org/wiki/Q1021509","display_name":"Consensus conference","level":2,"score":0.2565999925136566}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2508.04281","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2508.04281","pdf_url":"https://arxiv.org/pdf/2508.04281","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-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2508.04281","doi":"https://doi.org/10.48550/arxiv.2508.04281","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2508.04281","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-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2508.04281","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2508.04281","pdf_url":"https://arxiv.org/pdf/2508.04281","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-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[{"score":0.7297000288963318,"id":"https://metadata.un.org/sdg/16","display_name":"Peace, justice, and strong institutions"}],"sustainable_development_goals_aurora":[{"score":0.6296433210372925,"id":"https://metadata.un.org/sdg/16","display_name":"Peace, justice, and strong institutions"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"are":[4,127],"gaining":[5],"traction":[6],"as":[7],"a":[8,70,79,102],"method":[9],"to":[10,41,51,151],"generate":[11],"consensus":[12,60,99,142],"statements":[13],"and":[14,35,77,84,93,98,106,125,136,165,194],"aggregate":[15],"preferences":[16],"in":[17,27,74,201],"digital":[18,202],"democracy":[19,203],"experiments.":[20],"Yet,":[21],"participants":[22],"can":[23],"introduce":[24],"critical":[25],"vulnerabilities":[26,193],"LLM-based":[28],"systems.":[29],"Here,":[30],"we":[31,81,113],"examine":[32],"the":[33,75,192,195],"vulnerability":[34],"robustness":[36,156],"of":[37,46,87,190,198],"off-the-shelf":[38],"consensus-generating":[39,199],"LLMs":[40,117,200],"prompt-injection":[42],"attacks,":[43,184],"which":[44],"consist":[45],"introducing":[47],"strategically":[48],"designed":[49],"texts":[50],"amplify":[52],"particular":[53],"viewpoints,":[54],"erase":[55],"certain":[56],"opinions,":[57],"or":[58,63],"divert":[59],"toward":[61,143],"unrelated":[62],"irrelevant":[64],"topics.":[65],"Using":[66],"data":[67],"collected":[68],"from":[69],"2023":[71],"experiment":[72],"conducted":[73],"UK":[76],"predefining":[78],"majority-rule,":[80],"construct":[82],"attack-free":[83],"adversarial":[85],"variants":[86],"prompts":[88],"containing":[89],"public":[90],"policy":[91],"questions":[92],"opinion":[94,97,163],"texts,":[95],"classify":[96],"valences":[100],"with":[101,146],"fine-tuned":[103],"BERT":[104],"model,":[105],"estimate":[107],"LLM--human":[108],"majority":[109],"agreement":[110,126],"rates.":[111],"Overall,":[112],"find":[114],"that":[115,140],"default":[116],"exhibit":[118],"widespread":[119],"vulnerability,":[120],"especially":[121],"when:":[122],"(i)":[123],"disagreement":[124],"finely":[128],"balanced,":[129],"(ii)":[130],"under":[131],"rational,":[132],"instruction-like":[133],"rhetorical":[134],"strategies,":[135],"(iii)":[137],"for":[138],"attacks":[139],"shift":[141],"positions":[144],"aligned":[145],"GB-unionist":[147],"conservative":[148],"manifestos":[149],"relative":[150],"pro-independence":[152],"left":[153],"manifestos.":[154],"A":[155],"pipeline":[157],"combining":[158],"GPT-OSS-SafeGuard":[159],"injection":[160],"detection,":[161],"structured":[162],"representations,":[164],"GSPO-based":[166],"reinforcement":[167],"learning":[168],"substantially":[169],"reduces":[170],"directional":[171],"failures,":[172],"outperforming":[173],"state-of-the-art":[174],"alternatives.":[175],"While":[176],"more":[177,182],"effective":[178],"detectors":[179],"may":[180],"enable":[181],"sophisticated":[183],"these":[185],"findings":[186],"advance":[187],"our":[188],"understanding":[189],"both":[191],"potential":[196],"defenses":[197],"applications.":[204]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-10-13T10:08:12.033832","created_date":"2025-10-10T00:00:00"}