{"id":"https://openalex.org/W7151254787","doi":"https://doi.org/10.48550/arxiv.2604.03993","title":"Can LLMs Learn to Reason Robustly under Noisy Supervision?","display_name":"Can LLMs Learn to Reason Robustly under Noisy Supervision?","publication_year":2026,"publication_date":"2026-04-05","ids":{"openalex":"https://openalex.org/W7151254787","doi":"https://doi.org/10.48550/arxiv.2604.03993"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.03993","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.03993","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.03993","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133142537","display_name":"Shenzhi Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Yang, Shenzhi","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5000700814","display_name":"G. Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Guangcheng","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133096445","display_name":"Bowen Song","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Song, Bowen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133119483","display_name":"Sharon Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Sharon","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133102555","display_name":"Haobo Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Haobo","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114804130","display_name":"Xing Zheng","orcid":"https://orcid.org/0000-0003-3171-6506"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Xing","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133098217","display_name":"Yingfan Ma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Yingfan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102157705","display_name":"Zhongqi Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Zhongqi","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133131710","display_name":"Weiqiang Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Weiqiang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133095449","display_name":"Gang Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Gang","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":10,"corresponding_author_ids":["https://openalex.org/A5133142537"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.942799985408783,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.942799985408783,"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":0.01080000028014183,"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":0.008899999782443047,"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/robustness","display_name":"Robustness (evolution)","score":0.6279000043869019},{"id":"https://openalex.org/keywords/correctness","display_name":"Correctness","score":0.6276000142097473},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.47679999470710754},{"id":"https://openalex.org/keywords/property","display_name":"Property (philosophy)","score":0.46149998903274536},{"id":"https://openalex.org/keywords/verifiable-secret-sharing","display_name":"Verifiable secret sharing","score":0.45159998536109924},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.450300008058548},{"id":"https://openalex.org/keywords/coherence","display_name":"Coherence (philosophical gambling strategy)","score":0.3481000065803528},{"id":"https://openalex.org/keywords/noisy-data","display_name":"Noisy data","score":0.34299999475479126}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6365000009536743},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.6279000043869019},{"id":"https://openalex.org/C55439883","wikidata":"https://www.wikidata.org/wiki/Q360812","display_name":"Correctness","level":2,"score":0.6276000142097473},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6114000082015991},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5350000262260437},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.47679999470710754},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.46149998903274536},{"id":"https://openalex.org/C85847156","wikidata":"https://www.wikidata.org/wiki/Q59015987","display_name":"Verifiable secret sharing","level":3,"score":0.45159998536109924},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.450300008058548},{"id":"https://openalex.org/C2781181686","wikidata":"https://www.wikidata.org/wiki/Q4226068","display_name":"Coherence (philosophical gambling strategy)","level":2,"score":0.3481000065803528},{"id":"https://openalex.org/C2781170535","wikidata":"https://www.wikidata.org/wiki/Q30587856","display_name":"Noisy data","level":2,"score":0.34299999475479126},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.3361000120639801},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.30239999294281006},{"id":"https://openalex.org/C75778745","wikidata":"https://www.wikidata.org/wiki/Q342626","display_name":"Lag","level":2,"score":0.28119999170303345},{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.27959999442100525},{"id":"https://openalex.org/C29265498","wikidata":"https://www.wikidata.org/wiki/Q7047719","display_name":"Noise measurement","level":3,"score":0.275299996137619},{"id":"https://openalex.org/C75553542","wikidata":"https://www.wikidata.org/wiki/Q178161","display_name":"A priori and a posteriori","level":2,"score":0.2590000033378601},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.2581999897956848},{"id":"https://openalex.org/C109747225","wikidata":"https://www.wikidata.org/wiki/Q815758","display_name":"Scarcity","level":2,"score":0.25619998574256897},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.25459998846054077}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.03993","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.03993","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.03993","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.03993","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reinforcement":[0],"Learning":[1],"with":[2,122,170],"Verifiable":[3],"Rewards":[4],"(RLVR)":[5],"effectively":[6],"trains":[7],"reasoning":[8,207],"models":[9],"that":[10,75,80],"rely":[11],"on":[12,63,67,87,120,143,203,248,255],"abundant":[13],"perfect":[14],"labels,":[15,98,106],"but":[16],"its":[17],"vulnerability":[18],"to":[19,24,50,83,136,228,246,253],"unavoidable":[20],"noisy":[21,43,84,97,105,123,133,147,168],"labels":[22,169],"due":[23],"expert":[25],"scarcity":[26],"remains":[27],"critically":[28],"underexplored.":[29],"In":[30,48],"this":[31,88,156],"work,":[32],"we":[33,90,125,158],"take":[34],"the":[35,69,113,181,197],"first":[36],"step":[37],"toward":[38,115],"a":[39,57,60,78,177],"systematic":[40],"analysis":[41],"of":[42,94,244],"label":[44],"mechanisms":[45],"in":[46,139,151,180],"RLVR.":[47],"contrast":[49],"supervised":[51],"classification,":[52],"most":[53],"RLVR":[54],"algorithms":[55],"incorporate":[56],"rollout-based":[58],"condition:":[59],"label's":[61],"influence":[62],"training":[64,121],"is":[65],"contingent":[66],"whether":[68],"current":[70],"policy":[71,198],"can":[72],"generate":[73],"rollouts":[74],"realize":[76],"it,":[77],"property":[79],"naturally":[81],"extends":[82],"labels.":[85],"Based":[86],"observation,":[89],"distinguish":[91],"two":[92,174],"types":[93],"noise:":[95],"inactive":[96,236],"which":[99,107,164],"reduce":[100],"data":[101],"efficiency,":[102],"and":[103,110,146,187,213,215,221,237,251],"active":[104,238],"are":[108],"reinforced":[109],"risk":[111],"skewing":[112],"model":[114],"incorrect":[116],"distributions.":[117],"From":[118],"experiments":[119],"samples,":[124],"identify":[126],"an":[127],"Early":[128],"Correctness":[129],"Coherence":[130],"phenomenon:":[131],"although":[132],"samples":[134,148],"begin":[135],"lag":[137],"behind":[138],"later":[140],"stages,":[141],"accuracy":[142],"both":[144,235],"clean":[145],"increases":[149],"similarly":[150],"early":[152],"training.":[153],"Motivated":[154],"by":[155],"dynamic,":[157],"propose":[159],"Online":[160],"Label":[161],"Refinement":[162],"(OLR),":[163],"progressively":[165],"corrects":[166],"potentially":[167],"majority-voted":[171],"answers":[172],"when":[173],"conditions":[175],"hold:":[176],"positive":[178],"slope":[179],"majority":[182],"answer's":[183],"rollout":[184],"pass":[185],"rate":[186],"stable":[188],"historical":[189],"consistency":[190],"across":[191],"updates,":[192],"enabling":[193],"gradual":[194],"self-correction":[195],"as":[196],"improves.":[199],"We":[200],"evaluate":[201],"OLR":[202,230],"six":[204],"in-distribution":[205,249],"mathematical":[206],"benchmarks":[208,250],"(AIME24/25,":[209],"AMC,":[210],"MATH-500,":[211],"Minerva,":[212],"Olympiad)":[214],"three":[216],"out-of-distribution":[217,256],"tasks":[218],"(ARC-c,":[219],"GPQA-diamond,":[220],"MMLU-pro).":[222],"Across":[223],"noise":[224],"ratios":[225],"from":[226],"0.1":[227],"0.9,":[229],"consistently":[231],"improves":[232],"robustness":[233],"under":[234],"noisy-label":[239],"settings,":[240],"achieving":[241],"average":[242],"gains":[243],"3.6%":[245],"3.9%":[247],"3.3%":[252],"4.6%":[254],"evaluations.":[257]},"counts_by_year":[],"updated_date":"2026-04-08T06:07:18.267832","created_date":"2026-04-08T00:00:00"}