{"id":"https://openalex.org/W7163208236","doi":"https://doi.org/10.48550/arxiv.2606.01476","title":"OmniOPD: Logit-Free On-Policy Distillation via Speculative Verification","display_name":"OmniOPD: Logit-Free On-Policy Distillation via Speculative Verification","publication_year":2026,"publication_date":"2026-05-31","ids":{"openalex":"https://openalex.org/W7163208236","doi":"https://doi.org/10.48550/arxiv.2606.01476"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.01476","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01476","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.2606.01476","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137637457","display_name":"Yuhang Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Yuhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137686069","display_name":"Lizhu Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Lizhu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137633102","display_name":"Yifan Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Yifan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137625201","display_name":"Mingyi Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Mingyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137687896","display_name":"Peng Bo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Peng, Bo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137645649","display_name":"Jiayi Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Jiayi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137698755","display_name":"Xiangjun Fan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fan, Xiangjun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137634271","display_name":"Zhuokai Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Zhuokai","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":0.2061000019311905,"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":0.2061000019311905,"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":0.09179999679327011,"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/T11902","display_name":"Intelligent Tutoring Systems and Adaptive Learning","score":0.06700000166893005,"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/logit","display_name":"Logit","score":0.5669999718666077},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.46560001373291016},{"id":"https://openalex.org/keywords/bayesian-probability","display_name":"Bayesian probability","score":0.4625999927520752},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.4616999924182892},{"id":"https://openalex.org/keywords/variance","display_name":"Variance (accounting)","score":0.4284000098705292},{"id":"https://openalex.org/keywords/categorical-variable","display_name":"Categorical variable","score":0.4009999930858612},{"id":"https://openalex.org/keywords/offset","display_name":"Offset (computer science)","score":0.3635999858379364},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.34630000591278076},{"id":"https://openalex.org/keywords/bayesian-inference","display_name":"Bayesian inference","score":0.34610000252723694}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6615999937057495},{"id":"https://openalex.org/C140331021","wikidata":"https://www.wikidata.org/wiki/Q1868104","display_name":"Logit","level":2,"score":0.5669999718666077},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.46560001373291016},{"id":"https://openalex.org/C107673813","wikidata":"https://www.wikidata.org/wiki/Q812534","display_name":"Bayesian probability","level":2,"score":0.4625999927520752},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.4616999924182892},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.4284000098705292},{"id":"https://openalex.org/C5274069","wikidata":"https://www.wikidata.org/wiki/Q2285707","display_name":"Categorical variable","level":2,"score":0.4009999930858612},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.3635999858379364},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.34630000591278076},{"id":"https://openalex.org/C160234255","wikidata":"https://www.wikidata.org/wiki/Q812535","display_name":"Bayesian inference","level":3,"score":0.34610000252723694},{"id":"https://openalex.org/C207201462","wikidata":"https://www.wikidata.org/wiki/Q182505","display_name":"Bayes' theorem","level":3,"score":0.3382999897003174},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.33309999108314514},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.33230000734329224},{"id":"https://openalex.org/C19499675","wikidata":"https://www.wikidata.org/wiki/Q232207","display_name":"Monte Carlo method","level":2,"score":0.3271999955177307},{"id":"https://openalex.org/C176248197","wikidata":"https://www.wikidata.org/wiki/Q458526","display_name":"Probably approximately correct learning","level":4,"score":0.314300000667572},{"id":"https://openalex.org/C57830394","wikidata":"https://www.wikidata.org/wiki/Q278079","display_name":"Posterior probability","level":3,"score":0.31349998712539673},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.3133000135421753},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.30630001425743103},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.30239999294281006},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.295199990272522},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.2856000065803528},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.2777000069618225},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.27399998903274536},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.27250000834465027},{"id":"https://openalex.org/C2779458634","wikidata":"https://www.wikidata.org/wiki/Q24963715","display_name":"Debiasing","level":2,"score":0.2718999981880188},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.2694000005722046},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26660001277923584},{"id":"https://openalex.org/C63584917","wikidata":"https://www.wikidata.org/wiki/Q333286","display_name":"Bounding overwatch","level":2,"score":0.26579999923706055},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.25589999556541443},{"id":"https://openalex.org/C122041747","wikidata":"https://www.wikidata.org/wiki/Q838611","display_name":"Ball (mathematics)","level":2,"score":0.2542000114917755}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.01476","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01476","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.2606.01476","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01476","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":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.692565381526947}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"On-Policy":[0],"Distillation":[1],"(OPD)":[2],"trains":[3],"a":[4,17,57,78,105,113,134,147,165,204],"student":[5,153,251],"model":[6],"on":[7,77,196,242],"its":[8,156,245],"own":[9],"generative":[10],"trajectories":[11],"under":[12],"dense":[13],"token-level":[14,54,70,210],"feedback":[15],"from":[16,64],"stronger":[18,228],"teacher,":[19],"mitigating":[20],"both":[21,110],"the":[22,31,52,69,129,152,171,188,250,253],"off-policy":[23],"distribution":[24],"shift":[25],"of":[26,35,60,81,173,255],"Supervised":[27],"Fine-Tuning":[28],"(SFT)":[29],"and":[30,87,89,142,164,176,222,234],"sparse":[32],"credit":[33],"assignment":[34],"Reinforcement":[36],"Learning":[37],"(RL).":[38],"However,":[39],"standard":[40,189],"OPD":[41,190],"faces":[42],"two":[43],"coupled":[44],"limitations.":[45],"First,":[46],"it":[47],"requires":[48],"direct":[49],"access":[50],"to":[51,91,194],"teacher's":[53,130],"logits,":[55],"excluding":[56],"broad":[58],"class":[59],"capable":[61],"proprietary":[62],"models":[63],"serving":[65],"as":[66,96,232],"teachers.":[67],"Second,":[68],"logit":[71,121,211],"signal":[72,208],"itself":[73],"is":[74,217],"brittle,":[75],"depending":[76],"narrow":[79],"overlap":[80],"plausible":[82],"next":[83],"tokens":[84],"between":[85],"teacher":[86,247],"student,":[88],"prone":[90],"amplifying":[92],"degenerate":[93],"patterns":[94],"such":[95,231],"repetition":[97],"loops.":[98],"In":[99],"this":[100,144],"paper,":[101],"we":[102],"introduce":[103],"OmniOPD,":[104],"novel":[106],"framework":[107],"that":[108,127,150,199],"addresses":[109],"limitations":[111],"through":[112,133],"logit-free,":[114],"chunk-level":[115,200],"supervision":[116,145],"signal.":[117],"OmniOPD":[118,186,236],"replaces":[119],"deterministic":[120],"matching":[122],"with":[123,227],"Monte":[124],"Carlo":[125],"rollouts":[126],"approximate":[128],"local":[131],"preferences":[132],"continuous":[135],"semantic":[136,201],"similarity":[137],"metric":[138],"over":[139,244],"multi-token":[140],"chunks,":[141],"concentrates":[143],"via":[146],"peak-entropy":[148],"scheduler":[149],"audits":[151],"only":[154],"at":[155],"high-uncertainty":[157],"reasoning":[158],"forks.":[159],"A":[160],"Dirichlet-Multinomial":[161],"Bayesian":[162],"prior":[163],"base-model":[166],"KL":[167],"anchor":[168],"further":[169],"bound":[170],"variance":[172],"discrete":[174],"sampling":[175],"prevent":[177],"policy":[178],"collapse":[179],"across":[180],"unaudited":[181],"tokens.":[182],"Across":[183],"competitive":[184],"benchmarks,":[185],"surpasses":[187],"approach":[191],"by":[192,219],"up":[193],"+28.64%":[195],"math,":[197],"confirming":[198],"verification":[202],"extracts":[203],"more":[205],"reliable":[206],"learning":[207],"than":[209],"matching,":[212],"whose":[213],"high":[214],"information":[215],"density":[216],"offset":[218],"significant":[220],"noise":[221],"brittleness.":[223],"Furthermore,":[224],"when":[225],"paired":[226],"black-box":[229],"teachers":[230],"Claude-4.5-Haiku":[233],"Gemini-2.5-Flash,":[235],"achieves":[237],"an":[238],"additional":[239],"+9.54%":[240],"relative":[241],"math":[243],"open-weight":[246],"counterpart,":[248],"advancing":[249],"past":[252],"performance":[254],"self-exploratory":[256],"RL.":[257]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-03T00:00:00"}