{"id":"https://openalex.org/W7202201236","doi":"https://doi.org/10.48550/arxiv.2608.08764","title":"Learning from Consensus and Disagreement: Unsupervised On-Policy Self-Distillation with Minority-Trajectory Contrast","display_name":"Learning from Consensus and Disagreement: Unsupervised On-Policy Self-Distillation with Minority-Trajectory Contrast","publication_year":2026,"publication_date":"2026-08-09","ids":{"openalex":"https://openalex.org/W7202201236","doi":"https://doi.org/10.48550/arxiv.2608.08764"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.08764","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.08764","pdf_url":"https://arxiv.org/pdf/2608.08764","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/2608.08764","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5125768049","display_name":"Jiaxin Guo","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guo, Jiaxin","raw_affiliation_strings":["School of Intelligence Science and Technology , Peking University","State Key Laboratory of General Artificial Intelligence , Peking University , Beijing , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Intelligence Science and Technology , Peking University","institution_ids":["https://openalex.org/I20231570"]},{"raw_affiliation_string":"State Key Laboratory of General Artificial Intelligence , Peking University , Beijing , China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037828026","display_name":"Yanwei Yue","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yue, Yanwei","raw_affiliation_strings":["School of Intelligence Science and Technology , Peking University","State Key Laboratory of General Artificial Intelligence , Peking University , Beijing , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Intelligence Science and Technology , Peking University","institution_ids":["https://openalex.org/I20231570"]},{"raw_affiliation_string":"State Key Laboratory of General Artificial Intelligence , Peking University , Beijing , China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101263563","display_name":"Xuanbo Fan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Fan, Xuanbo","raw_affiliation_strings":["School of Intelligence Science and Technology , Peking University","State Key Laboratory of General Artificial Intelligence , Peking University , Beijing , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Intelligence Science and Technology , Peking University","institution_ids":["https://openalex.org/I20231570"]},{"raw_affiliation_string":"State Key Laboratory of General Artificial Intelligence , Peking University , Beijing , China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5140991044","display_name":"Chunyu Yang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Chunyu","raw_affiliation_strings":["Ucap Cloud"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ucap Cloud","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5146527488","display_name":"Yan Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhang, Yan","raw_affiliation_strings":["School of Intelligence Science and Technology , Peking University","State Key Laboratory of General Artificial Intelligence , Peking University , Beijing , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Intelligence Science and Technology , Peking University","institution_ids":["https://openalex.org/I20231570"]},{"raw_affiliation_string":"State Key Laboratory of General Artificial Intelligence , Peking University , Beijing , China","institution_ids":["https://openalex.org/I20231570"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"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/T10028","display_name":"Topic Modeling","score":0.5175999999046326,"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/T10028","display_name":"Topic Modeling","score":0.5175999999046326,"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/T12072","display_name":"Machine Learning and Algorithms","score":0.11869999766349792,"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":0.09109999984502792,"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/on-policy-self-distillation","display_name":"on-policy self-distillation","score":0.9279999732971191},{"id":"https://openalex.org/keywords/language-model-reasoning","display_name":"language model reasoning","score":0.6809999942779541},{"id":"https://openalex.org/keywords/unsupervised-learning","display_name":"unsupervised learning","score":0.6679999828338623},{"id":"https://openalex.org/keywords/consensus-and-disagreement","display_name":"consensus and disagreement","score":0.6549999713897705},{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.49799999594688416}],"concepts":[{"id":"https://openalex.org/C3962253","wikidata":"https://www.wikidata.org/wiki/Q852466","display_name":"Coda","level":2,"score":0.9394000172615051},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.6394000053405762},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6366999745368958},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.6148999929428101},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.5976999998092651},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.5447999835014343},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5424000024795532},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4796000123023987},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.462799996137619},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.46000000834465027},{"id":"https://openalex.org/C66905080","wikidata":"https://www.wikidata.org/wiki/Q17005494","display_name":"Binary classification","level":3,"score":0.3695000112056732},{"id":"https://openalex.org/C38976095","wikidata":"https://www.wikidata.org/wiki/Q752641","display_name":"Asymmetry","level":2,"score":0.3133000135421753},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.29760000109672546},{"id":"https://openalex.org/C166052673","wikidata":"https://www.wikidata.org/wiki/Q83021","display_name":"Empirical evidence","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.27160000801086426},{"id":"https://openalex.org/C8038995","wikidata":"https://www.wikidata.org/wiki/Q1152135","display_name":"Unsupervised learning","level":2,"score":0.2705000042915344},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.2702000141143799},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.2687999904155731},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2662999927997589},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C75949130","wikidata":"https://www.wikidata.org/wiki/Q848010","display_name":"Database transaction","level":2,"score":0.2635999917984009},{"id":"https://openalex.org/C90329073","wikidata":"https://www.wikidata.org/wiki/Q914232","display_name":"Ask price","level":2,"score":0.2583000063896179},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2578999996185303}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.08764","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.08764","pdf_url":"https://arxiv.org/pdf/2608.08764","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.2608.08764","doi":"https://doi.org/10.48550/arxiv.2608.08764","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.08764","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.08764","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.08764","pdf_url":"https://arxiv.org/pdf/2608.08764","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":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7202201236.pdf","grobid_xml":"https://content.openalex.org/works/W7202201236.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7215074916","https://openalex.org/W7201842969","https://openalex.org/W7125861034","https://openalex.org/W7212261691","https://openalex.org/W7161707194","https://openalex.org/W7160995547","https://openalex.org/W7161147084","https://openalex.org/W7161057388","https://openalex.org/W7168989064","https://openalex.org/W7164364162"],"abstract_inverted_index":{"On-policy":[0],"self-distillation":[1],"improves":[2,180],"language-model":[3],"reasoning":[4,87],"by":[5,13,23],"querying":[6],"a":[7,19,52,68,85,91,117,128,145],"teacher":[8,26],"on":[9,34,99,172],"states":[10],"actually":[11],"visited":[12],"the":[14,25,63,79,159,163,166],"student.":[15],"Recent":[16],"methods":[17],"create":[18],"powerful":[20],"information":[21,60],"asymmetry":[22],"exposing":[24],"to":[27,94],"privileged":[28,59],"context,":[29],"yet":[30],"they":[31],"fundamentally":[32],"rely":[33],"external":[35],"supervision---such":[36],"as":[37,138],"gold":[38],"solutions":[39],"or":[40],"verifiers---to":[41],"construct":[42],"this":[43,122],"advantage.":[44],"We":[45],"introduce":[46],"CoDA":[47,73,126,178],"(Consensus":[48],"and":[49,141,185],"Disagreement":[50],"Alignment),":[51],"fully":[53],"unsupervised":[54],"framework":[55],"that":[56,131,162,177],"creates":[57],"reliable":[58],"entirely":[61],"from":[62],"latent":[64],"uncertainty":[65],"structure":[66],"of":[67],"model's":[69],"own":[70],"unlabeled":[71],"rollouts.":[72],"extracts":[74],"two":[75],"complementary":[76],"signals.":[77],"In":[78],"positive":[80],"branch,":[81],"answer-level":[82],"consensus":[83,164],"identifies":[84],"stable":[86],"mode,":[88],"which":[89],"conditions":[90],"frozen":[92],"self-teacher":[93],"provide":[95],"dense":[96],"distributional":[97],"guidance":[98],"fresh":[100],"student":[101],"trajectories.":[102],"However,":[103],"because":[104],"agreement":[105],"does":[106],"not":[107],"guarantee":[108],"correctness,":[109],"positive-only":[110],"distillation":[111],"risks":[112],"amplifying":[113],"correlated":[114],"errors":[115],"into":[116],"false":[118],"consensus.":[119,191],"To":[120],"break":[121],"harmful":[123],"feedback":[124,153],"loop,":[125],"incorporates":[127],"negative":[129],"branch":[130],"exploits":[132],"disagreement:":[133],"minority":[134],"trajectories":[135],"are":[136],"treated":[137],"unstable":[139],"alternatives":[140],"gently":[142],"penalized":[143],"via":[144],"reference-anchored,":[146],"KTO-style":[147],"calibration":[148],"objective.":[149],"This":[150],"unpaired":[151],"binary":[152],"provides":[154],"robust":[155],"regularization":[156],"without":[157],"requiring":[158],"strong":[160],"assumption":[161],"is":[165],"absolute":[167],"ground":[168],"truth.":[169],"Empirical":[170],"evaluations":[171],"competition-level":[173],"mathematical":[174],"benchmarks":[175],"demonstrate":[176],"significantly":[179],"reasoning,":[181],"outperforming":[182],"self-generated":[183],"baselines":[184],"effectively":[186],"stabilizing":[187],"training":[188],"against":[189],"erroneous":[190]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-12T00:00:00"}