{"id":"https://openalex.org/W7214748988","doi":"https://doi.org/10.48550/arxiv.2609.32361","title":"Black-Box Auditing of Epistemic Reliability in Multi-Agent Debate Distillation","display_name":"Black-Box Auditing of Epistemic Reliability in Multi-Agent Debate Distillation","publication_year":2026,"publication_date":"2026-09-26","ids":{"openalex":"https://openalex.org/W7214748988","doi":"https://doi.org/10.48550/arxiv.2609.32361"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.32361","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.32361","pdf_url":"https://arxiv.org/pdf/2609.32361","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":null,"license_id":null,"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/2609.32361","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5081805488","display_name":"Derui Wang","orcid":"https://orcid.org/0000-0003-1388-7715","observed_orcids":["https://orcid.org/0000-0003-1388-7715"]},"institutions":[{"id":"https://openalex.org/I1292875679","display_name":"Commonwealth Scientific and Industrial Research Organisation","ror":"https://ror.org/03qn8fb07","country_code":"AU","type":"government","lineage":["https://openalex.org/I1292875679","https://openalex.org/I2801453606","https://openalex.org/I4387156119"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Wang, Derui","raw_affiliation_strings":["CSIRO"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CSIRO","institution_ids":["https://openalex.org/I1292875679"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100584234","display_name":"Zewei Shi","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1292875679","display_name":"Commonwealth Scientific and Industrial Research Organisation","ror":"https://ror.org/03qn8fb07","country_code":"AU","type":"government","lineage":["https://openalex.org/I1292875679","https://openalex.org/I2801453606","https://openalex.org/I4387156119"]},{"id":"https://openalex.org/I165779595","display_name":"The University of Melbourne","ror":"https://ror.org/01ej9dk98","country_code":"AU","type":"education","lineage":["https://openalex.org/I165779595"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Shi, Zewei","raw_affiliation_strings":["CSIRO","The University of Melbourne"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CSIRO","institution_ids":["https://openalex.org/I1292875679"]},{"raw_affiliation_string":"The University of Melbourne","institution_ids":["https://openalex.org/I165779595"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5043529893","display_name":"Ray Holland","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1292875679","display_name":"Commonwealth Scientific and Industrial Research Organisation","ror":"https://ror.org/03qn8fb07","country_code":"AU","type":"government","lineage":["https://openalex.org/I1292875679","https://openalex.org/I2801453606","https://openalex.org/I4387156119"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Holland, Rayne","raw_affiliation_strings":["CSIRO"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CSIRO","institution_ids":["https://openalex.org/I1292875679"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153438260","display_name":"Ruoxi Sun","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1292875679","display_name":"Commonwealth Scientific and Industrial Research Organisation","ror":"https://ror.org/03qn8fb07","country_code":"AU","type":"government","lineage":["https://openalex.org/I1292875679","https://openalex.org/I2801453606","https://openalex.org/I4387156119"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Sun, Ruoxi","raw_affiliation_strings":["CSIRO"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CSIRO","institution_ids":["https://openalex.org/I1292875679"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5064553444","display_name":"Xingliang Yuan","orcid":"https://orcid.org/0000-0002-3701-4946","observed_orcids":["https://orcid.org/0000-0002-3701-4946"]},"institutions":[{"id":"https://openalex.org/I165779595","display_name":"The University of Melbourne","ror":"https://ror.org/01ej9dk98","country_code":"AU","type":"education","lineage":["https://openalex.org/I165779595"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Yuan, Xingliang","raw_affiliation_strings":["The University of Melbourne"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The University of Melbourne","institution_ids":["https://openalex.org/I165779595"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113300962","display_name":"Jason Xue","orcid":"https://orcid.org/0000-0002-9172-4252","observed_orcids":["https://orcid.org/0000-0002-9172-4252"]},"institutions":[{"id":"https://openalex.org/I1292875679","display_name":"Commonwealth Scientific and Industrial Research Organisation","ror":"https://ror.org/03qn8fb07","country_code":"AU","type":"government","lineage":["https://openalex.org/I1292875679","https://openalex.org/I2801453606","https://openalex.org/I4387156119"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Xue, Jason","raw_affiliation_strings":["CSIRO"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CSIRO","institution_ids":["https://openalex.org/I1292875679"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5153501355","display_name":"Liming Zhu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1292875679","display_name":"Commonwealth Scientific and Industrial Research Organisation","ror":"https://ror.org/03qn8fb07","country_code":"AU","type":"government","lineage":["https://openalex.org/I1292875679","https://openalex.org/I2801453606","https://openalex.org/I4387156119"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Zhu, Liming","raw_affiliation_strings":["CSIRO"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CSIRO","institution_ids":["https://openalex.org/I1292875679"]}]}],"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":0.6875,"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":0.6875,"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":0.08410000056028366,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.04910000041127205,"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/multi-agent-debate","display_name":"multi-agent debate","score":0.8980000019073486},{"id":"https://openalex.org/keywords/epistemic-reliability","display_name":"epistemic reliability","score":0.9760000109672546},{"id":"https://openalex.org/keywords/catastrophic-forgetting","display_name":"catastrophic forgetting","score":0.8489999771118164},{"id":"https://openalex.org/keywords/paraphrase-evaluation","display_name":"paraphrase evaluation","score":0.6830000281333923},{"id":"https://openalex.org/keywords/lower-confidence-limits","display_name":"lower confidence bound","score":0.8999999761581421},{"id":"https://openalex.org/keywords/total-variation-distance","display_name":"total variation distance","score":0.7390000224113464}],"concepts":[{"id":"https://openalex.org/C162838799","wikidata":"https://www.wikidata.org/wiki/Q596077","display_name":"Counterexample","level":2,"score":0.8102999925613403},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.7305999994277954},{"id":"https://openalex.org/C7149132","wikidata":"https://www.wikidata.org/wiki/Q1377840","display_name":"Forgetting","level":2,"score":0.7085000276565552},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6674000024795532},{"id":"https://openalex.org/C199521495","wikidata":"https://www.wikidata.org/wiki/Q181487","display_name":"Audit","level":2,"score":0.6179999709129333},{"id":"https://openalex.org/C108650721","wikidata":"https://www.wikidata.org/wiki/Q1783253","display_name":"Counterfactual thinking","level":2,"score":0.5720000267028809},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5317999720573425},{"id":"https://openalex.org/C2776548248","wikidata":"https://www.wikidata.org/wiki/Q12621536","display_name":"Judgement","level":2,"score":0.5175999999046326},{"id":"https://openalex.org/C98184364","wikidata":"https://www.wikidata.org/wiki/Q1780131","display_name":"Argument (complex analysis)","level":2,"score":0.5163000226020813},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43630000948905945},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.4092000126838684},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3937000036239624},{"id":"https://openalex.org/C207390915","wikidata":"https://www.wikidata.org/wiki/Q1230525","display_name":"Divergence (linguistics)","level":2,"score":0.3813999891281128},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.35040000081062317},{"id":"https://openalex.org/C206345919","wikidata":"https://www.wikidata.org/wiki/Q20380951","display_name":"Resource (disambiguation)","level":2,"score":0.33090001344680786},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.32499998807907104},{"id":"https://openalex.org/C2778067643","wikidata":"https://www.wikidata.org/wiki/Q166507","display_name":"Interval (graph theory)","level":2,"score":0.3154999911785126},{"id":"https://openalex.org/C183115368","wikidata":"https://www.wikidata.org/wiki/Q856577","display_name":"Weighting","level":2,"score":0.30559998750686646},{"id":"https://openalex.org/C36914074","wikidata":"https://www.wikidata.org/wiki/Q1940501","display_name":"nobody","level":2,"score":0.3027999997138977},{"id":"https://openalex.org/C2778023277","wikidata":"https://www.wikidata.org/wiki/Q321703","display_name":"Premise","level":2,"score":0.2996000051498413},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.29440000653266907},{"id":"https://openalex.org/C32230216","wikidata":"https://www.wikidata.org/wiki/Q7882499","display_name":"Uncertainty quantification","level":2,"score":0.2928999960422516},{"id":"https://openalex.org/C37381756","wikidata":"https://www.wikidata.org/wiki/Q20203288","display_name":"Representativeness heuristic","level":2,"score":0.29190000891685486},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2827000021934509},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.27410000562667847},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.27309998869895935},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.2696000039577484},{"id":"https://openalex.org/C40422974","wikidata":"https://www.wikidata.org/wiki/Q741248","display_name":"Assertion","level":2,"score":0.26899999380111694},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2671000063419342},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.2667999863624573},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.25940001010894775},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.2540999948978424},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.25360000133514404},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.25049999356269836}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.32361","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.32361","pdf_url":"https://arxiv.org/pdf/2609.32361","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.32361","doi":"https://doi.org/10.48550/arxiv.2609.32361","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.32361","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:2609.32361","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.32361","pdf_url":"https://arxiv.org/pdf/2609.32361","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320320386","display_name":"Commonwealth Scientific and Industrial Research Organisation","ror":"https://ror.org/03qn8fb07"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214748988.pdf","grobid_xml":"https://content.openalex.org/works/W7214748988.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7202044583","https://openalex.org/W7212305758","https://openalex.org/W4388748022","https://openalex.org/W7214791812","https://openalex.org/W7213760417","https://openalex.org/W7214902728","https://openalex.org/W7213310495","https://openalex.org/W7215091782","https://openalex.org/W7203701931","https://openalex.org/W7160156110"],"abstract_inverted_index":{"Debate":[0],"distillation":[1],"adapts":[2],"weaker":[3],"verifiers":[4],"using":[5],"multi-agent":[6],"debate":[7,56],"transcripts":[8],"to":[9,121,171,194],"improve":[10],"their":[11,180],"judgement":[12],"in":[13,34,114],"subsequent":[14],"debates,":[15],"but":[16],"gains":[17],"on":[18,25,46,146,206],"monitored":[19,38,62,108],"tasks":[20],"do":[21],"not":[22],"establish":[23,159],"reliability":[24,32],"related":[26],"unmonitored":[27],"tasks.":[28,48],"We":[29,49,140,157],"study":[30],"epistemic":[31],"degradation,":[33],"which":[35],"adaptation":[36],"preserves":[37],"performance":[39,222],"while":[40,58],"reducing":[41],"support":[42],"for":[43,119,136],"correct":[44,61],"responses":[45],"hidden":[47,110],"consider":[50],"an":[51],"adversarial":[52],"debater":[53],"that":[54,93,185,212],"manipulates":[55],"arguments":[57],"defending":[59],"the":[60,67,147],"response,":[63],"and":[64,74,99,102,109,168,196,210,226],"ask":[65],"whether":[66,75],"resulting":[68],"degradation":[69,218],"merely":[70],"reflects":[71],"catastrophic":[72,208],"forgetting":[73,209],"standard":[76],"evaluation":[77,105],"can":[78,189,214],"detect":[79],"it.":[80],"To":[81],"address":[82],"these":[83],"questions,":[84],"we":[85],"propose":[86],"ER-Audit,":[87],"a":[88,154,160,174],"two-stage":[89],"black-box":[90],"auditing":[91,213],"framework":[92],"compares":[94],"frozen":[95],"verifier":[96],"checkpoints":[97],"before":[98],"after":[100],"adaptation,":[101],"introduce":[103],"two":[104],"benchmarks":[106,227],"pairing":[107],"task":[111],"prompts":[112],"grounded":[113],"shared":[115],"contexts.":[116],"ER-Audit":[117],"searches":[118],"counterexamples":[120,193],"non-degradation":[122,148,195,198],"by":[123,220],"evaluating":[124],"semantically":[125],"valid":[126],"paraphrases":[127,135],"and,":[128],"if":[129],"none":[130],"is":[131],"found,":[132],"uses":[133],"independent":[134],"sequential":[137],"hypothesis":[138],"testing.":[139],"derive":[141],"anytime-valid":[142],"lower":[143,162,197],"confidence":[144],"bounds":[145],"probability,":[149],"allowing":[150],"data-dependent":[151],"stopping":[152],"within":[153,173],"finite":[155],"budget.":[156],"further":[158],"common":[161],"bound":[163],"across":[164],"fixed":[165],"paraphrase":[166],"distributions":[167,172],"extend":[169],"it":[170],"bounded":[175],"total":[176],"variation":[177],"distance":[178],"of":[179],"mixtures.":[181],"Our":[182,224],"experiments":[183],"show":[184],"higher":[186],"hidden-task":[187,217],"accuracy":[188],"coexist":[190],"with":[191],"more":[192],"bounds.":[199],"This":[200],"divergence":[201],"challenges":[202],"explanations":[203],"based":[204],"solely":[205],"broad":[207],"shows":[211],"uncover":[215],"selective":[216],"concealed":[219],"aggregate":[221],"gains.":[223],"code":[225],"are":[228],"available":[229],"at":[230],"https://github.com/CSIRO-CQS-AI-alignment-Team/Epistemic-Reliability-Auditor.":[231]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-30T00:00:00"}