{"id":"https://openalex.org/W4416036313","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.654","title":"Advancing Oversight Reasoning across Languages for Audit Sycophantic Behaviour via X-Agent","display_name":"Advancing Oversight Reasoning across Languages for Audit Sycophantic Behaviour via X-Agent","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416036313","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.654"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.654","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.654","pdf_url":"https://aclanthology.org/2025.emnlp-main.654.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.654.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5020089275","display_name":"Giulia Pucci","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Giulia Pucci","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5062023845","display_name":"Leonardo Ranaldi","orcid":"https://orcid.org/0000-0001-8488-4146"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leonardo Ranaldi","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":2,"corresponding_author_ids":["https://openalex.org/A5020089275"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.18376113,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"12960","last_page":"12976"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10456","display_name":"Multi-Agent Systems and Negotiation","score":0.08640000224113464,"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/T10456","display_name":"Multi-Agent Systems and Negotiation","score":0.08640000224113464,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.03709999844431877,"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/T13083","display_name":"Advanced Text Analysis Techniques","score":0.036400001496076584,"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/audit","display_name":"Audit","score":0.5728999972343445},{"id":"https://openalex.org/keywords/government","display_name":"Government (linguistics)","score":0.28040000796318054},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.2770000100135803},{"id":"https://openalex.org/keywords/action","display_name":"Action (physics)","score":0.26930001378059387},{"id":"https://openalex.org/keywords/field","display_name":"Field (mathematics)","score":0.2524000108242035}],"concepts":[{"id":"https://openalex.org/C199521495","wikidata":"https://www.wikidata.org/wiki/Q181487","display_name":"Audit","level":2,"score":0.5728999972343445},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.40869998931884766},{"id":"https://openalex.org/C56739046","wikidata":"https://www.wikidata.org/wiki/Q192060","display_name":"Knowledge management","level":1,"score":0.3587000072002411},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3582000136375427},{"id":"https://openalex.org/C39549134","wikidata":"https://www.wikidata.org/wiki/Q133080","display_name":"Public relations","level":1,"score":0.33489999175071716},{"id":"https://openalex.org/C121955636","wikidata":"https://www.wikidata.org/wiki/Q4116214","display_name":"Accounting","level":1,"score":0.3199000060558319},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.3059999942779541},{"id":"https://openalex.org/C195094911","wikidata":"https://www.wikidata.org/wiki/Q14167904","display_name":"Process management","level":1,"score":0.29829999804496765},{"id":"https://openalex.org/C2778137410","wikidata":"https://www.wikidata.org/wiki/Q2732820","display_name":"Government (linguistics)","level":2,"score":0.28040000796318054},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.2770000100135803},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.26930001378059387},{"id":"https://openalex.org/C55587333","wikidata":"https://www.wikidata.org/wiki/Q1133029","display_name":"Engineering ethics","level":1,"score":0.26080000400543213},{"id":"https://openalex.org/C539667460","wikidata":"https://www.wikidata.org/wiki/Q2414942","display_name":"Management science","level":1,"score":0.25679999589920044},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.2524000108242035},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.25110000371932983}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.654","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.654","pdf_url":"https://aclanthology.org/2025.emnlp-main.654.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"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.654","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.654","pdf_url":"https://aclanthology.org/2025.emnlp-main.654.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":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416036313.pdf","grobid_xml":"https://content.openalex.org/works/W4416036313.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"have":[4],"demonstrated":[5],"capabilities":[6],"that":[7,33,83,111,127,149],"are":[8,42],"highly":[9],"satisfactory":[10],"to":[11,19,30,53,184],"a":[12,28,108,137,162,173,182,189],"wide":[13],"range":[14],"of":[15,51,196],"users":[16],"by":[17,101],"adapting":[18],"their":[20],"culture":[21],"and":[22,67,74,92,115,120,146,157,168],"wisdom.Yet,":[23],"this":[24],"can":[25],"translate":[26],"into":[27],"propensity":[29],"produce":[31],"responses":[32,56],"align":[34,61],"with":[35,62],"users'":[36],"viewpoints,":[37],"even":[38],"when":[39],"the":[40,49,63,94,193],"latter":[41],"wrong.This":[43],"behaviour":[44,114],"is":[45,179],"known":[46],"as":[47,57,59,129,188],"sycophancy,":[48,134,153],"tendency":[50],"LLMs":[52],"generate":[54],"misleading":[55],"long":[58],"they":[60],"user's,":[64],"inducing":[65],"bias":[66],"reducing":[68],"reliability.To":[69],"make":[70],"interactions":[71],"consistent,":[72],"reliable":[73],"safe,":[75],"we":[76],"introduce":[77],"X-Agent,":[78],"an":[79],"Oversight":[80],"Reasoning":[81],"framework":[82],"audits":[84],"human-model":[85],"dialogues,":[86],"reasons":[87],"about":[88],"them,":[89],"captures":[90],"sycophancy":[91],"corrects":[93],"final":[95],"outputs.Concretely,":[96],"X-Agent":[97,142],"extends":[98],"debate-based":[99],"frameworks":[100],"(i)":[102],"auditing":[103],"user-model":[104,167],"conversations,":[105],"(ii)":[106],"applying":[107],"defence":[109],"layer":[110],"steers":[112],"model":[113],"goes":[116],"beyond":[117],"user":[118],"beliefs,":[119],"(iii)":[121],"extracting":[122],"reasoning":[123,178],"traces":[124],"from":[125],"evaluations":[126],"serve":[128],"training":[130],"signals":[131],"for":[132,165,191],"mitigating":[133],"all":[135],"in":[136,176],"completely":[138],"unsupervised":[139],"way.We":[140],"evaluate":[141],"across":[143],"diverse":[144],"scenarios":[145],"languages,":[147],"showing":[148],"it":[150],"consistently":[151],"detects":[152],"reduces":[154],"unwarranted":[155],"agreement,":[156],"improves":[158],"crossturn":[159],"consistency,":[160],"advancing":[161],"reasoning-asoverview":[163],"paradigm":[164,175],"safer":[166],"model-model":[169],"interaction.Our":[170],"approach":[171],"introduces":[172],"novel":[174],"which":[177],"not":[180],"merely":[181],"means":[183],"solve":[185],"problems,":[186],"but":[187],"mechanism":[190],"overseeing":[192],"problem-solving":[194],"processes":[195],"other":[197],"models.":[198]},"counts_by_year":[],"updated_date":"2026-03-13T14:20:09.374765","created_date":"2025-11-08T00:00:00"}