{"id":"https://openalex.org/W7125861034","doi":"https://doi.org/10.48550/arxiv.2601.18734","title":"Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models","display_name":"Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models","publication_year":2026,"publication_date":"2026-01-26","ids":{"openalex":"https://openalex.org/W7125861034","doi":"https://doi.org/10.48550/arxiv.2601.18734"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2601.18734","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2601.18734","pdf_url":"https://arxiv.org/pdf/2601.18734","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":"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/2601.18734","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101257955","display_name":"Siyan Zhao","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Siyan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123886423","display_name":"Zhihui Xie","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Zhihui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123890591","display_name":"Mengchen Liu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Mengchen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124014336","display_name":"Jing Huang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Jing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124047361","display_name":"Guan Pang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pang, Guan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100643583","display_name":"Feiyu Chen","orcid":"https://orcid.org/0000-0002-2259-698X","observed_orcids":["https://orcid.org/0000-0002-2259-698X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Feiyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5014409944","display_name":"Aditya Grover","orcid":"https://orcid.org/0000-0002-4217-6624","observed_orcids":["https://orcid.org/0000-0002-4217-6624"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Grover, Aditya","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"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":0.6255000233650208,"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":0.6255000233650208,"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":0.07479999959468842,"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":0.051600001752376556,"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":0.7929999828338623},{"id":"https://openalex.org/keywords/knowledge-distillation","display_name":"knowledge distillation","score":0.9399999976158142},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.8619999885559082},{"id":"https://openalex.org/keywords/llm-reasoning","display_name":"LLM reasoning","score":0.8100000023841858},{"id":"https://openalex.org/keywords/privileged-information","display_name":"privileged information","score":0.6600000262260437},{"id":"https://openalex.org/keywords/reasoning-traces","display_name":"reasoning traces","score":0.7950000166893005},{"id":"https://openalex.org/keywords/distribution-mismatch","display_name":"distribution mismatch","score":0.6420000195503235},{"id":"https://openalex.org/keywords/token-efficiency","display_name":"token efficiency","score":0.6759999990463257},{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.8119999766349792},{"id":"https://openalex.org/keywords/token-level-supervision","display_name":"token-level supervision","score":0.37400001287460327}],"concepts":[{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.6915000081062317},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6815999746322632},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.613099992275238},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.578499972820282},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5249999761581421},{"id":"https://openalex.org/C132010649","wikidata":"https://www.wikidata.org/wiki/Q189222","display_name":"Intuition","level":2,"score":0.5012000203132629},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4009000062942505},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3912000060081482},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.38839998841285706},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.3172999918460846},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.3034000098705292},{"id":"https://openalex.org/C207390915","wikidata":"https://www.wikidata.org/wiki/Q1230525","display_name":"Divergence (linguistics)","level":2,"score":0.3021000027656555},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.25440001487731934}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2601.18734","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2601.18734","pdf_url":"https://arxiv.org/pdf/2601.18734","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2601.18734","doi":"https://doi.org/10.48550/arxiv.2601.18734","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2601.18734","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:2601.18734","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2601.18734","pdf_url":"https://arxiv.org/pdf/2601.18734","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7125861034.pdf","grobid_xml":"https://content.openalex.org/works/W7125861034.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7162670085","https://openalex.org/W7215074916","https://openalex.org/W7158506380","https://openalex.org/W7220990175","https://openalex.org/W7161057388","https://openalex.org/W7162137784","https://openalex.org/W7164364162","https://openalex.org/W7161264723","https://openalex.org/W7214987755","https://openalex.org/W7171930018"],"abstract_inverted_index":{"Knowledge":[0],"distillation":[1,21,51,55,174],"improves":[2],"large":[3],"language":[4],"model":[5],"(LLM)":[6],"reasoning":[7,73,88,125,158],"by":[8,25,76],"compressing":[9],"the":[10,27,42,77,128,133,137,144,150],"knowledge":[11],"of":[12,152],"a":[13,34,58,80,100,104],"teacher":[14,35,62,110,117],"LLM":[15,36,63,83,106],"to":[16,165],"train":[17],"smaller":[18],"LLMs.":[19],"On-policy":[20],"advances":[22],"this":[23],"approach":[24],"having":[26],"student":[28,112,129],"sample":[29],"its":[30,92],"own":[31,146],"trajectories":[32],"while":[33,127],"provides":[37],"dense":[38],"token-level":[39],"supervision,":[40],"addressing":[41],"distribution":[43],"mismatch":[44],"between":[45,140],"training":[46,135],"and":[47,64,90,111,169],"inference":[48],"in":[49,72],"off-policy":[50,173],"methods.":[52,175],"However,":[53],"on-policy":[54],"typically":[56],"requires":[57],"separate,":[59],"often":[60],"larger,":[61],"does":[65],"not":[66],"explicitly":[67],"leverage":[68],"ground-truth":[69],"solutions":[70],"available":[71],"datasets.":[74],"Inspired":[75],"intuition":[78],"that":[79],"sufficiently":[81],"capable":[82],"can":[84],"rationalize":[85],"external":[86],"privileged":[87,121],"traces":[89],"teach":[91],"weaker":[93],"self,":[94],"we":[95],"introduce":[96],"On-Policy":[97],"Self-Distillation":[98],"(OPSD),":[99],"learning":[101,167],"algorithm":[102],"where":[103],"single":[105],"acts":[107],"as":[108],"both":[109],"with":[113],"different":[114],"contexts.":[115],"The":[116],"policy":[118,130],"conditions":[119],"on":[120,155],"information":[122],"(e.g.,":[123],"verified":[124],"traces)":[126],"sees":[131],"only":[132],"question;":[134],"minimizes":[136],"per-token":[138],"divergence":[139],"these":[141],"distributions":[142],"over":[143,172],"student's":[145],"rollouts.":[147],"We":[148],"demonstrate":[149],"efficacy":[151],"our":[153],"method":[154],"multiple":[156],"mathematical":[157],"benchmarks,":[159],"achieving":[160],"superior":[161],"token":[162],"efficiency":[163],"compared":[164],"reinforcement":[166],"methods":[168],"better":[170],"performance":[171],"Code":[176],"repo:":[177],"https://github.com/siyan-zhao/OPSD.":[178]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-01-28T00:00:00"}