{"id":"https://openalex.org/W7160664940","doi":"https://doi.org/10.48550/arxiv.2605.05940","title":"Near-Policy: Accelerating On-Policy Distillation via Asynchronous Generation and Selective Packing","display_name":"Near-Policy: Accelerating On-Policy Distillation via Asynchronous Generation and Selective Packing","publication_year":2026,"publication_date":"2026-05-07","ids":{"openalex":"https://openalex.org/W7160664940","doi":"https://doi.org/10.48550/arxiv.2605.05940"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.05940","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05940","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":"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.05940","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135653179","display_name":"Miao Rang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rang, Miao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5033574239","display_name":"Zhenni Bi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bi, Zhenni","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135698189","display_name":"Hang Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Hang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135655292","display_name":"Kai Han","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Kai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135716747","display_name":"Xuechun Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Xuechun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017898444","display_name":"An Xiao","orcid":"https://orcid.org/0000-0003-0909-3234"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiao, An","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135710582","display_name":"Xinghao Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Xinghao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135713501","display_name":"Yunhe Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yunhe","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135656086","display_name":"Hanting Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Hanting","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":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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.3605000078678131,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.3605000078678131,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.12549999356269836,"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":0.10350000113248825,"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/asynchronous-communication","display_name":"Asynchronous communication","score":0.79339998960495},{"id":"https://openalex.org/keywords/distillation","display_name":"Distillation","score":0.6969000101089478},{"id":"https://openalex.org/keywords/heuristic","display_name":"Heuristic","score":0.5764999985694885},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.5662000179290771},{"id":"https://openalex.org/keywords/sample","display_name":"Sample (material)","score":0.5547000169754028},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.5365999937057495},{"id":"https://openalex.org/keywords/sequence","display_name":"Sequence (biology)","score":0.45089998841285706},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.44350001215934753},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.4309999942779541}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8036999702453613},{"id":"https://openalex.org/C151319957","wikidata":"https://www.wikidata.org/wiki/Q752739","display_name":"Asynchronous communication","level":2,"score":0.79339998960495},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.6969000101089478},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.5764999985694885},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5662000179290771},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.5547000169754028},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.5365999937057495},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.45089998841285706},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.44350001215934753},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.4309999942779541},{"id":"https://openalex.org/C154030694","wikidata":"https://www.wikidata.org/wiki/Q1436074","display_name":"Fractionating column","level":3,"score":0.41440001130104065},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.38429999351501465},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3774000108242035},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.3714999854564667},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.37049999833106995},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.36500000953674316},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.35589998960494995},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.3278999924659729},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.32739999890327454},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.31459999084472656},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.3012999892234802},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.3003000020980835},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.29829999804496765},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.2703000009059906},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2556999921798706},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2547999918460846}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.05940","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05940","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.05940","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05940","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":"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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Standard":[0],"knowledge":[1],"distillation":[2],"for":[3,147],"autoregressive":[4],"models":[5],"often":[6],"suffers":[7],"from":[8,44,72,111],"distribution":[9],"mismatch.":[10],"While":[11],"on-policy":[12,133],"methods":[13],"mitigate":[14],"this":[15,78],"by":[16,138,141],"leveraging":[17],"student-generated":[18],"outputs,":[19],"they":[20],"rely":[21],"on":[22],"computationally":[23],"expensive":[24],"Reinforcement":[25],"Learning":[26],"(RL)":[27],"frameworks.":[28],"To":[29,76],"improve":[30],"efficiency,":[31,81],"we":[32],"propose":[33],"Near-Policy":[34],"Distillation":[35],"(NPD),":[36],"an":[37],"asynchronous":[38,56],"approach":[39],"that":[40,97],"decouples":[41],"student":[42,85],"generation":[43],"training.":[45],"This":[46],"reformulation":[47],"enables":[48,152],"Supervised":[49],"Fine-Tuning":[50],"(SFT)":[51],"with":[52],"sequence":[53],"packing.":[54],"However,":[55],"updates":[57,86,116],"inevitably":[58],"introduce":[59],"policy":[60],"lag":[61],"and":[62,87,135],"sample":[63,94],"noise,":[64],"which":[65],"can":[66],"cause":[67],"the":[68,88,100,113,125,144,162],"behavior":[69],"to":[70,154],"drift":[71],"near-policy":[73],"toward":[74],"off-policy.":[75],"counteract":[77],"without":[79],"sacrificing":[80],"NPD":[82,126],"integrates":[83],"sparse":[84],"$\u0394$-IFD":[89,108],"filtering":[90,104],"mechanism,":[91],"a":[92,119,129,156],"heuristic":[93],"selection":[95],"mechanism":[96],"empirically":[98],"stabilizes":[99],"optimization":[101],"trajectory.":[102],"By":[103],"extreme":[105],"out-of-distribution":[106],"samples,":[107],"prevents":[109],"noise":[110],"dominating":[112],"gradients,":[114],"ensuring":[115],"remain":[117],"within":[118],"safe":[120],"proximal":[121],"learning":[122],"zone.":[123],"Empirically,":[124],"framework":[127],"achieves":[128],"8.1x":[130],"speedup":[131],"over":[132],"baselines":[134],"outperforms":[136],"SFT":[137],"8.09%.":[139],"Crucially,":[140],"effectively":[142],"narrowing":[143],"exploration":[145],"space":[146],"subsequent":[148],"RL,":[149],"our":[150],"method":[151],"openPangu-Embedded-1B":[153],"reach":[155],"state-of-the-art":[157],"score":[158],"of":[159],"68.73%,":[160],"outperforming":[161],"substantially":[163],"larger":[164],"Qwen3-1.7B.":[165],"Codes":[166],"will":[167],"be":[168],"released":[169],"soon.":[170]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-09T00:00:00"}