{"id":"https://openalex.org/W7203880136","doi":"https://doi.org/10.48550/arxiv.2608.18271","title":"Rethinking Privileged Information in On-Policy Self-Distillation","display_name":"Rethinking Privileged Information in On-Policy Self-Distillation","publication_year":2026,"publication_date":"2026-08-18","ids":{"openalex":"https://openalex.org/W7203880136","doi":"https://doi.org/10.48550/arxiv.2608.18271"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.18271","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.18271","pdf_url":"https://arxiv.org/pdf/2608.18271","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.18271","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5132576977","display_name":"Samyak Shrestha","orcid":"https://orcid.org/0009-0008-9351-3644","observed_orcids":["https://orcid.org/0009-0008-9351-3644"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shrestha, Samyak","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5082190233","display_name":"Alexander Tessier","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tessier, Alexander","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":12,"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":0.3977000117301941,"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":0.3977000117301941,"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":0.3905999958515167,"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":0.09120000153779984,"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":0.9959999918937683},{"id":"https://openalex.org/keywords/privileged-information","display_name":"privileged information","score":0.8420000076293945},{"id":"https://openalex.org/keywords/teacher-student-learning","display_name":"teacher-student learning","score":0.5910000205039978},{"id":"https://openalex.org/keywords/token-level-supervision","display_name":"token-level supervision","score":0.7099999785423279},{"id":"https://openalex.org/keywords/distributional-alignment","display_name":"distributional alignment","score":0.5929999947547913},{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.6629999876022339},{"id":"https://openalex.org/keywords/scientific-reasoning","display_name":"scientific reasoning","score":0.8519999980926514},{"id":"https://openalex.org/keywords/qwen3","display_name":"Qwen3","score":0.8420000076293945},{"id":"https://openalex.org/keywords/self-supervised-learning","display_name":"self-supervised learning","score":0.4729999899864197}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6021000146865845},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.5745999813079834},{"id":"https://openalex.org/C150189527","wikidata":"https://www.wikidata.org/wiki/Q356674","display_name":"Reference model","level":2,"score":0.4171999990940094},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3865000009536743},{"id":"https://openalex.org/C4554734","wikidata":"https://www.wikidata.org/wiki/Q593744","display_name":"Knowledge base","level":2,"score":0.34369999170303345},{"id":"https://openalex.org/C59656382","wikidata":"https://www.wikidata.org/wiki/Q191536","display_name":"Conjunction (astronomy)","level":2,"score":0.32429999113082886},{"id":"https://openalex.org/C190839683","wikidata":"https://www.wikidata.org/wiki/Q2448197","display_name":"Train","level":2,"score":0.31290000677108765},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.30869999527931213},{"id":"https://openalex.org/C115051666","wikidata":"https://www.wikidata.org/wiki/Q6522493","display_name":"Ranging","level":2,"score":0.3012999892234802},{"id":"https://openalex.org/C145420912","wikidata":"https://www.wikidata.org/wiki/Q853077","display_name":"Mathematics education","level":1,"score":0.29980000853538513},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.27480000257492065},{"id":"https://openalex.org/C60478076","wikidata":"https://www.wikidata.org/wiki/Q3036835","display_name":"Reference data","level":2,"score":0.2743000090122223},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.27090001106262207},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.2565999925136566}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.18271","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.18271","pdf_url":"https://arxiv.org/pdf/2608.18271","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.2608.18271","doi":"https://doi.org/10.48550/arxiv.2608.18271","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.18271","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:2608.18271","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.18271","pdf_url":"https://arxiv.org/pdf/2608.18271","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":0.8278999924659729,"display_name":"Quality education"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7203880136.pdf","grobid_xml":"https://content.openalex.org/works/W7203880136.grobid-xml"},"referenced_works_count":25,"referenced_works":["https://openalex.org/W1821462560","https://openalex.org/W2963736842","https://openalex.org/W3101005014","https://openalex.org/W4388926373","https://openalex.org/W4414939355","https://openalex.org/W4417440793","https://openalex.org/W7125861034","https://openalex.org/W7125915072","https://openalex.org/W7126034091","https://openalex.org/W7128026050","https://openalex.org/W7128814765","https://openalex.org/W7130402983","https://openalex.org/W7131375838","https://openalex.org/W7139913273","https://openalex.org/W7140517510","https://openalex.org/W7154299763","https://openalex.org/W7161046900","https://openalex.org/W7161054869","https://openalex.org/W7161147084","https://openalex.org/W7162670085","https://openalex.org/W7167214743","https://openalex.org/W7167642481","https://openalex.org/W7171552453","https://openalex.org/W7202181542","https://openalex.org/W7202204426"],"related_works":["https://openalex.org/W7213581439","https://openalex.org/W7202181542","https://openalex.org/W7172413156","https://openalex.org/W7223337476","https://openalex.org/W7204449334","https://openalex.org/W7223227413","https://openalex.org/W7213073827","https://openalex.org/W7202213474","https://openalex.org/W7220699741","https://openalex.org/W7214731163"],"abstract_inverted_index":{"On-policy":[0],"self-distillation":[1],"(OPSD)":[2],"trains":[3],"a":[4,106,127,186],"student":[5,32,208],"on":[6,18,56,137],"its":[7],"own":[8],"responses":[9],"using":[10,61],"token-level":[11],"supervision":[12,74,81,157],"from":[13,27,65,79,129,165,190],"the":[14,31,34,37,49,73,77,80,84,87,97,123,134,149,156,160,178,191],"same":[15],"model":[16,114],"conditioned":[17],"privileged":[19,203],"reference":[20,38,78,88,102,180,204],"information.":[21],"We":[22,52],"investigate":[23],"whether":[24],"performance":[25,108,188],"gains":[26,194],"OPSD":[28,54],"show":[29],"that":[30],"learned":[33],"information":[35,205],"in":[36,48,96,210],"or":[39],"instead":[40],"reflect":[41],"recovery":[42],"of":[43,170],"reasoning":[44,140],"behavior":[45,153],"already":[46],"present":[47],"base":[50,150],"model.":[51],"perform":[53],"experiments":[55],"science":[57],"and":[58,89,116,126,195],"mathematics":[59],"datasets":[60],"Qwen3":[62],"models":[63],"ranging":[64],"1.7B":[66],"to":[67,177,207],"8B.":[68],"Our":[69],"analysis":[70],"framework":[71],"separates":[72],"induced":[75,158],"by":[76,83,159],"provided":[82],"teacher":[85,111],"without":[86,122],"measures":[90],"how":[91,202],"each":[92],"aligns":[93],"with":[94,148,155,185],"changes":[95],"student's":[98,143],"predictions.":[99],"The":[100,142],"correct":[101,124,135,179],"does":[103,181],"not":[104,182],"provide":[105],"consistent":[107],"benefit":[109,189],"across":[110],"generation":[112],"modes,":[113],"sizes,":[115],"training":[117],"datasets.":[118],"Students":[119],"can":[120,132],"improve":[121],"reference,":[125,161],"solution":[128,136],"another":[130],"problem":[131],"outperform":[133],"several":[138],"mathematical":[139],"benchmarks.":[141],"predictions":[144],"align":[145],"more":[146],"strongly":[147],"model's":[151],"thinking":[152],"than":[154],"but":[162],"controls":[163],"constructed":[164],"other":[166],"problems":[167],"reproduce":[168],"much":[169],"both":[171],"alignments.":[172],"Moreover,":[173],"stronger":[174],"alignment":[175,197],"attributable":[176],"reliably":[183],"coincide":[184],"greater":[187],"reference.":[192],"Performance":[193],"distributional":[196],"alone":[198],"therefore":[199],"cannot":[200],"determine":[201],"contributes":[206],"learning":[209],"OPSD.":[211]},"counts_by_year":[{"year":2026,"cited_by_count":12}],"updated_date":"2026-11-09T13:22:03.289127","created_date":"2026-08-22T00:00:00"}