{"id":"https://openalex.org/W7225467883","doi":"https://doi.org/10.5281/zenodo.24045650","title":"Told the Answer Is Not Taught to Reason: What On-Policy Self-Distillation From Privileged Context Transfers to a Language Model, Why Its Gains and Losses Divide Along Whether the Student Can Ever Hold What the Teacher Was Told, and the Controls That Would Separate Three Signals Under One Loss","display_name":"Told the Answer Is Not Taught to Reason: What On-Policy Self-Distillation From Privileged Context Transfers to a Language Model, Why Its Gains and Losses Divide Along Whether the Student Can Ever Hold What the Teacher Was Told, and the Controls That Would Separate Three Signals Under One Loss","publication_year":2026,"publication_date":"2026-11-08","ids":{"openalex":"https://openalex.org/W7225467883","doi":"https://doi.org/10.5281/zenodo.24045650"},"language":"en","primary_location":{"id":"doi:10.5281/zenodo.24045650","doi":"https://doi.org/10.5281/zenodo.24045650","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24045650","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.5281/zenodo.24045650","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5147720734","display_name":"Pranay M. Mahendrakar","orcid":"https://orcid.org/0009-0003-7224-029X","observed_orcids":["https://orcid.org/0009-0003-7224-029X"]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Mahendrakar, Pranay","raw_affiliation_strings":["SONYTECH"],"raw_orcid":"https://orcid.org/0009-0003-7224-029X","affiliations":[{"raw_affiliation_string":"SONYTECH","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5147720734"],"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.30959999561309814,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.30959999561309814,"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":0.2087000012397766,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.15629999339580536,"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":0.9900000095367432},{"id":"https://openalex.org/keywords/language-models","display_name":"language models","score":0.6809999942779541},{"id":"https://openalex.org/keywords/reasoning-models","display_name":"reasoning models","score":0.3529999852180481},{"id":"https://openalex.org/keywords/information-signals","display_name":"information signals","score":0.9810000061988831},{"id":"https://openalex.org/keywords/mathematical-problem-solving","display_name":"mathematical problem solving","score":0.47099998593330383}],"concepts":[{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.5996000170707703},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.46970000863075256},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.462799996137619},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4514999985694885},{"id":"https://openalex.org/C149629883","wikidata":"https://www.wikidata.org/wiki/Q660926","display_name":"Fraction (chemistry)","level":2,"score":0.4383000135421753},{"id":"https://openalex.org/C2777363581","wikidata":"https://www.wikidata.org/wiki/Q15098235","display_name":"Harm","level":2,"score":0.4277999997138977},{"id":"https://openalex.org/C77805123","wikidata":"https://www.wikidata.org/wiki/Q161272","display_name":"Social psychology","level":1,"score":0.42419999837875366},{"id":"https://openalex.org/C145420912","wikidata":"https://www.wikidata.org/wiki/Q853077","display_name":"Mathematics education","level":1,"score":0.39100000262260437},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.375},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.3711000084877014},{"id":"https://openalex.org/C71889745","wikidata":"https://www.wikidata.org/wiki/Q1783264","display_name":"Counterfactual conditional","level":3,"score":0.3650999963283539},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.3634999990463257},{"id":"https://openalex.org/C48677424","wikidata":"https://www.wikidata.org/wiki/Q6888088","display_name":"Mode (computer interface)","level":2,"score":0.35920000076293945},{"id":"https://openalex.org/C143299363","wikidata":"https://www.wikidata.org/wiki/Q900584","display_name":"Attribution","level":2,"score":0.3564000129699707},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.35370001196861267},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3434999883174896},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.31709998846054077},{"id":"https://openalex.org/C2776175482","wikidata":"https://www.wikidata.org/wiki/Q1195816","display_name":"Transfer (computing)","level":2,"score":0.2969000041484833},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.29100000858306885},{"id":"https://openalex.org/C2777267654","wikidata":"https://www.wikidata.org/wiki/Q3519023","display_name":"Test (biology)","level":2,"score":0.28850001096725464},{"id":"https://openalex.org/C150899416","wikidata":"https://www.wikidata.org/wiki/Q1820378","display_name":"Transfer of learning","level":2,"score":0.28839999437332153},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.2851000130176544},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2825999855995178},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.2797999978065491},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.2574999928474426},{"id":"https://openalex.org/C47177190","wikidata":"https://www.wikidata.org/wiki/Q207137","display_name":"Curriculum","level":2,"score":0.25450000166893005},{"id":"https://openalex.org/C2776050585","wikidata":"https://www.wikidata.org/wiki/Q7439360","display_name":"Scrutiny","level":2,"score":0.2515000104904175},{"id":"https://openalex.org/C2779178101","wikidata":"https://www.wikidata.org/wiki/Q6987274","display_name":"Negative transfer","level":3,"score":0.2502000033855438}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.5281/zenodo.24045650","doi":"https://doi.org/10.5281/zenodo.24045650","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24045650","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},{"id":"doi:10.5281/zenodo.24045651","doi":"https://doi.org/10.5281/zenodo.24045651","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24045651","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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":"doi:10.5281/zenodo.24045650","doi":"https://doi.org/10.5281/zenodo.24045650","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.24045650","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","score":0.9235000014305115,"display_name":"Quality education"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7213581439","https://openalex.org/W7167642481","https://openalex.org/W7215074916","https://openalex.org/W7225382284","https://openalex.org/W7204470223","https://openalex.org/W7214987755","https://openalex.org/W7202181542","https://openalex.org/W7223227413","https://openalex.org/W7203880136","https://openalex.org/W7160995547"],"abstract_inverted_index":{"(c)":[0],"2026":[1,48,65],"Pranay":[2],"Mahendrakar.":[3],"Licensed":[4],"under":[5],"CC":[6],"BY":[7],"4.0.":[8],"Life":[9],"of":[10,26,60,155,169,227],"Research":[11],"On-policy":[12],"self-distillation":[13],"trains":[14],"a":[15,36,39,45,58,145,301,304,307],"language":[16],"model":[17,29],"on":[18,70,81],"its":[19,267,277],"own":[20,210,278],"samples":[21],"against":[22],"the":[23,27,32,51,61,68,106,116,121,138,150,159,167,171,188,193,204,208,215,228,232,239,248,257,263,270,281,315],"token":[24],"distribution":[25],"same":[28],"shown":[30],"something":[31],"student":[33,258],"never":[34],"sees:":[35],"reference":[37],"solution,":[38],"final":[40],"answer,":[41],"execution":[42],"feedback":[43],"or":[44,186,275],"demonstration.":[46],"Early":[47],"reports":[49],"credited":[50],"recipe":[52],"with":[53,177],"matching":[54],"reinforcement":[55],"learning":[56],"at":[57,131],"fraction":[59],"sampling":[62],"cost.":[63],"Later":[64],"studies":[66],"report":[67],"opposite":[69],"thinking":[71],"models:":[72],"accuracy":[73],"falls":[74],"by":[75,84,314],"up":[76,85],"to":[77,86,113,253,261,288],"17":[78,311],"percent":[79,88],"relative":[80],"competition":[82],"mathematics,":[83],"40":[87],"in":[89,180,207,266,276,300],"another":[90],"study,":[91],"and":[92,109,142,153,202,220,244,303,319],"agents":[93],"trained":[94,179],"this":[95,104],"way":[96],"can":[97,259],"end":[98],"below":[99],"their":[100],"untrained":[101],"base.":[102],"In":[103],"analysis":[105],"author":[107],"reviews":[108],"synthesises":[110],"published":[111,298,312],"results":[112,241],"show":[114,191,224],"that":[115,127,157,184,192,200,203,225,247,283,323],"disagreement":[117],"is":[118,213,231,251],"not":[119,197],"about":[120,124],"method":[122],"but":[123],"three":[125],"signals":[126],"one":[128],"loss":[129,222],"carries":[130],"once.":[132],"A":[133],"mode":[134,212,216],"signal":[135,147,164,250,316],"comes":[136,148,165],"from":[137,149,166,199,242],"teacher's":[139],"prompt":[140],"format":[141],"reasoning":[143],"mode;":[144],"behaviour":[146,233,249],"confidence,":[151],"brevity":[152],"absence":[154],"checking":[156],"knowing":[158],"answer":[160],"licenses;":[161],"an":[162],"information":[163,264,271],"content":[168,190],"what":[170,326],"teacher":[172],"was":[173],"told.":[174],"On":[175],"mathematics":[176],"students":[178],"non-thinking":[181,211],"mode,":[182],"ablations":[183],"shuffle":[185],"delete":[187],"privileged":[189],"reported":[194,229],"gain":[195,205],"does":[196],"come":[198,260],"content,":[201],"measured":[206],"student's":[209],"largely":[214],"signal.":[217,234],"Marker":[218],"counts":[219],"per-token":[221],"attributions":[223],"much":[226],"harm":[230],"The":[235,294],"paper":[236,295],"argues,":[237],"using":[238],"imitation-gap":[240],"robot":[243],"control":[245],"learning,":[246],"safe":[252],"transfer":[254],"only":[255],"when":[256],"hold":[262],"-":[265],"weights,":[268],"because":[269],"recurs":[272],"across":[273],"instances,":[274],"context":[279],"before":[280],"step":[282],"needs":[284],"it.":[285],"Per-problem":[286],"answers":[287],"unseen":[289],"problems":[290],"meet":[291],"neither":[292],"condition.":[293],"consolidates":[296],"34":[297],"values":[299],"table":[302],"figure,":[305],"states":[306],"reading":[308],"procedure,":[309],"sorts":[310],"repairs":[313],"each":[317],"removes,":[318],"names":[320],"six":[321],"measurements":[322],"would":[324],"settle":[325],"remains":[327],"open.":[328]},"counts_by_year":[],"updated_date":"2026-11-09T13:22:03.289127","created_date":"2026-11-08T00:00:00"}