{"id":"https://openalex.org/W7202369261","doi":"https://doi.org/10.48550/arxiv.2608.11829","title":"Towards Understanding On-Policy Distillation through the Lens of Test-Time Scaling","display_name":"Towards Understanding On-Policy Distillation through the Lens of Test-Time Scaling","publication_year":2026,"publication_date":"2026-08-12","ids":{"openalex":"https://openalex.org/W7202369261","doi":"https://doi.org/10.48550/arxiv.2608.11829"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.11829","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.11829","pdf_url":"https://arxiv.org/pdf/2608.11829","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/2608.11829","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137096081","display_name":"Xinmu Ge","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ge, Xinmu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035619686","display_name":"Zizhuo Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I141568987","display_name":"Hong Kong Baptist University","ror":"https://ror.org/0145fw131","country_code":"HK","type":"education","lineage":["https://openalex.org/I141568987"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Zhang, Zizhuo","raw_affiliation_strings":["Hong Kong Baptist University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hong Kong Baptist University","institution_ids":["https://openalex.org/I141568987"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5055069752","display_name":"Yu Huang","orcid":"https://orcid.org/0000-0003-2730-5077","observed_orcids":["https://orcid.org/0000-0003-2730-5077"]},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]},{"id":"https://openalex.org/I4405270328","display_name":"Ant Group (China)","ror":"https://ror.org/00g5w7n47","country_code":"CN","type":"company","lineage":["https://openalex.org/I4405270328"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huang, Yu","raw_affiliation_strings":["Ant Group","Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ant Group","institution_ids":["https://openalex.org/I4405270328"]},{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147136485","display_name":"Jianing Zhu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I86519309","display_name":"The University of Texas at Austin","ror":"https://ror.org/00hj54h04","country_code":"US","type":"education","lineage":["https://openalex.org/I86519309"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhu, Jianing","raw_affiliation_strings":["University of Texas at Austin"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Texas at Austin","institution_ids":["https://openalex.org/I86519309"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147131317","display_name":"Lin Yuan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405270328","display_name":"Ant Group (China)","ror":"https://ror.org/00g5w7n47","country_code":"CN","type":"company","lineage":["https://openalex.org/I4405270328"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yuan, Lin","raw_affiliation_strings":["Ant Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ant Group","institution_ids":["https://openalex.org/I4405270328"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5064526570","display_name":"Wanli Gu","orcid":"https://orcid.org/0009-0002-9982-823X","observed_orcids":["https://orcid.org/0009-0002-9982-823X"]},"institutions":[{"id":"https://openalex.org/I4405270328","display_name":"Ant Group (China)","ror":"https://ror.org/00g5w7n47","country_code":"CN","type":"company","lineage":["https://openalex.org/I4405270328"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Gu, Wanli","raw_affiliation_strings":["Ant Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ant Group","institution_ids":["https://openalex.org/I4405270328"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147106392","display_name":"Weichang Wu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405270328","display_name":"Ant Group (China)","ror":"https://ror.org/00g5w7n47","country_code":"CN","type":"company","lineage":["https://openalex.org/I4405270328"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wu, Weichang","raw_affiliation_strings":["Ant Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ant Group","institution_ids":["https://openalex.org/I4405270328"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125945718","display_name":"Weiran Huang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]},{"id":"https://openalex.org/I4405279436","display_name":"Shanghai Innovation Institute","ror":"https://ror.org/01fatbw66","country_code":"CN","type":"education","lineage":["https://openalex.org/I4405279436"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huang, Weiran","raw_affiliation_strings":["Shanghai Innovation Institute","Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Innovation Institute","institution_ids":["https://openalex.org/I4405279436"]},{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124308314","display_name":"Xiaolu Zhang","orcid":"https://orcid.org/0009-0003-8332-5539","observed_orcids":["https://orcid.org/0009-0003-8332-5539"]},"institutions":[{"id":"https://openalex.org/I141568987","display_name":"Hong Kong Baptist University","ror":"https://ror.org/0145fw131","country_code":"HK","type":"education","lineage":["https://openalex.org/I141568987"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Han, Bo","raw_affiliation_strings":["Hong Kong Baptist University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Hong Kong Baptist University","institution_ids":["https://openalex.org/I141568987"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5147111663","display_name":"Bo Han","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405270328","display_name":"Ant Group (China)","ror":"https://ror.org/00g5w7n47","country_code":"CN","type":"company","lineage":["https://openalex.org/I4405270328"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhang, Xiaolu","raw_affiliation_strings":["Ant Group"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ant Group","institution_ids":["https://openalex.org/I4405270328"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5153501114","display_name":"Jiangchao Yao","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yao, Jiangchao","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":5,"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":0.6474999785423279,"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":0.6474999785423279,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.13600000739097595,"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":0.04309999942779541,"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-distillation","display_name":"on-policy distillation","score":1.0},{"id":"https://openalex.org/keywords/test-time-scaling","display_name":"test-time scaling","score":0.8709999918937683},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.7250000238418579},{"id":"https://openalex.org/keywords/llm-reasoning","display_name":"LLM reasoning","score":0.7089999914169312},{"id":"https://openalex.org/keywords/sampling-efficiency","display_name":"sampling efficiency","score":0.6209999918937683},{"id":"https://openalex.org/keywords/capability-boundaries","display_name":"capability boundaries","score":0.5149999856948853}],"concepts":[{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.753600001335144},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.6919999718666077},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.6826000213623047},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5223000049591064},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.4641000032424927},{"id":"https://openalex.org/C15336307","wikidata":"https://www.wikidata.org/wiki/Q1766051","display_name":"Lens (geology)","level":2,"score":0.4564000070095062},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4456000030040741},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.4185999929904938},{"id":"https://openalex.org/C4554734","wikidata":"https://www.wikidata.org/wiki/Q593744","display_name":"Knowledge base","level":2,"score":0.412200003862381},{"id":"https://openalex.org/C43091099","wikidata":"https://www.wikidata.org/wiki/Q1067788","display_name":"Through-the-lens metering","level":3,"score":0.37940001487731934},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.36419999599456787},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.31540000438690186},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.31040000915527344},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.3098999857902527},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3073999881744385},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.2994000017642975},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.2671000063419342},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2587999999523163},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.25609999895095825},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2549999952316284},{"id":"https://openalex.org/C13736549","wikidata":"https://www.wikidata.org/wiki/Q4489420","display_name":"Industrial engineering","level":1,"score":0.25110000371932983}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.11829","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.11829","pdf_url":"https://arxiv.org/pdf/2608.11829","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.2608.11829","doi":"https://doi.org/10.48550/arxiv.2608.11829","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.11829","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:2608.11829","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.11829","pdf_url":"https://arxiv.org/pdf/2608.11829","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":[{"id":"https://metadata.un.org/sdg/4","score":0.6164000034332275,"display_name":"Quality education"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7202369261.pdf","grobid_xml":"https://content.openalex.org/works/W7202369261.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7141388625","https://openalex.org/W7154505894","https://openalex.org/W7221072085","https://openalex.org/W7204908525","https://openalex.org/W7161054869","https://openalex.org/W7168964786","https://openalex.org/W7214769125","https://openalex.org/W7214953522","https://openalex.org/W7214772452","https://openalex.org/W7214800163"],"abstract_inverted_index":{"On-policy":[0],"distillation":[1],"(OPD)":[2],"has":[3,142],"emerged":[4],"as":[5],"a":[6,127,155,182,214],"promising":[7],"post-training":[8],"technique":[9],"for":[10,217],"enhancing":[11],"LLM":[12],"reasoning.":[13],"Under":[14],"the":[15,19,25,31,34,38,50,59,74,81,140,161,193,200,205,221],"reverse":[16,132],"KL":[17,133],"objective,":[18],"idealized":[20,131],"optimum":[21],"of":[22,30,62,76,204,211],"OPD":[23,44,98,173,207],"aligns":[24],"student":[26],"distribution":[27],"with":[28,88,199],"that":[29,43,124],"teacher.":[32],"When":[33],"teacher":[35,141,178,218],"consistently":[36],"outperforms":[37],"student,":[39],"this":[40,67,71],"naturally":[41],"suggests":[42],"should":[45],"yield":[46],"broad":[47],"improvements":[48,56,203],"over":[49],"pre-OPD":[51],"student.":[52],"However,":[53],"do":[54],"such":[55,126],"extend":[57],"across":[58],"entire":[60],"range":[61],"test-time":[63,77],"sampling":[64,82,107,157],"budgets?":[65],"In":[66],"work,":[68],"we":[69,93,159],"revisit":[70],"expectation":[72],"through":[73],"lens":[75],"scaling":[78],"by":[79,196],"varying":[80],"budget":[83],"$K$":[84],"and":[85,105,129,190],"evaluating":[86],"performance":[87,115],"pass@$K$.":[89,119,186,223],"Across":[90,187],"multiple":[91],"settings,":[92],"observe":[94],"two":[95,151],"distinct":[96],"patterns:":[97],"can":[99,111,169],"improve":[100,113],"pass@$K$":[101,202],"at":[102,154,165,220],"both":[103],"small":[104],"large":[106],"budgets,":[108],"but":[109],"it":[110,136],"also":[112],"small-budget":[114],"while":[116],"reducing":[117],"large-budget":[118],"We":[120],"show":[121],"one":[122],"condition":[123],"guarantees":[125],"reversal":[128],"an":[130],"counterexample":[134],"where":[135],"occurs":[137],"even":[138],"when":[139],"higher":[143],"accuracy":[144],"on":[145],"every":[146],"problem.":[147],"To":[148],"choose":[149],"between":[150],"candidate":[152],"teachers":[153],"target":[156,222],"budget,":[158],"propose":[160],"\\textit{Teacher":[162],"Advantage":[163],"Score":[164],"$K$}":[166],"(TAS@$K$),":[167],"which":[168,177],"be":[170],"computed":[171],"before":[172],"training":[174],"to":[175,181],"predict":[176],"will":[179],"lead":[180],"larger":[183],"improvement":[184],"in":[185,209],"three":[188],"domains":[189],"thirteen":[191],"benchmarks,":[192],"ordering":[194],"predicted":[195],"TAS@$K$":[197],"agrees":[198],"observed":[201],"resulting":[206],"models":[208],"83.6\\%":[210],"experiments,":[212],"providing":[213],"useful":[215],"signal":[216],"selection":[219]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-14T00:00:00"}