{"id":"https://openalex.org/W7214699424","doi":"https://doi.org/10.48550/arxiv.2609.34738","title":"Beyond Token Alignment: Event Completion for Cross-Tokenizer On-Policy Distillation","display_name":"Beyond Token Alignment: Event Completion for Cross-Tokenizer On-Policy Distillation","publication_year":2026,"publication_date":"2026-09-28","ids":{"openalex":"https://openalex.org/W7214699424","doi":"https://doi.org/10.48550/arxiv.2609.34738"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.34738","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.34738","pdf_url":"https://arxiv.org/pdf/2609.34738","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/2609.34738","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5153429588","display_name":"Jiacheng Liu","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":"Liu, Jiacheng","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153511770","display_name":"Jingwei Song","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":"Song, Jingwei","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102662097","display_name":"Qituan Zhang","orcid":"https://orcid.org/0009-0008-1985-5582","observed_orcids":["https://orcid.org/0009-0008-1985-5582"]},"institutions":[{"id":"https://openalex.org/I24943067","display_name":"Fudan University","ror":"https://ror.org/013q1eq08","country_code":"CN","type":"education","lineage":["https://openalex.org/I24943067"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhang, Qituan","raw_affiliation_strings":["Fudan University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University","institution_ids":["https://openalex.org/I24943067"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5153478852","display_name":"Siheng Chen","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":"Chen, Siheng","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5153462549","display_name":"Linfeng Zhang","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":"Zhang, Linfeng","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":1,"institutions_distinct_count":2,"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":0.7285000085830688,"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":0.7285000085830688,"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":0.121799997985363,"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":0.10760000348091125,"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":0.9900000095367432},{"id":"https://openalex.org/keywords/event-completion","display_name":"event completion","score":0.7749999761581421},{"id":"https://openalex.org/keywords/token-alignment","display_name":"token alignment","score":0.5270000100135803},{"id":"https://openalex.org/keywords/code-generation","display_name":"code generation","score":0.40799999237060547},{"id":"https://openalex.org/keywords/mathematical-reasoning","display_name":"mathematical reasoning","score":0.7680000066757202}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7907999753952026},{"id":"https://openalex.org/C2779662365","wikidata":"https://www.wikidata.org/wiki/Q5416694","display_name":"Event (particle physics)","level":2,"score":0.7860000133514404},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.717199981212616},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.6200000047683716},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.5411999821662903},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.4747999906539917},{"id":"https://openalex.org/C159423971","wikidata":"https://www.wikidata.org/wiki/Q177251","display_name":"Associative property","level":2,"score":0.40790000557899475},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.3456000089645386},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3443000018596649},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3357999920845032},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.33309999108314514},{"id":"https://openalex.org/C197096303","wikidata":"https://www.wikidata.org/wiki/Q869887","display_name":"Probability mass function","level":3,"score":0.32829999923706055},{"id":"https://openalex.org/C2779955035","wikidata":"https://www.wikidata.org/wiki/Q4686785","display_name":"Advice (programming)","level":2,"score":0.3147999942302704},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.267300009727478},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.26489999890327454},{"id":"https://openalex.org/C133112747","wikidata":"https://www.wikidata.org/wiki/Q7251931","display_name":"Protocol analysis","level":2,"score":0.26429998874664307}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.34738","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.34738","pdf_url":"https://arxiv.org/pdf/2609.34738","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.2609.34738","doi":"https://doi.org/10.48550/arxiv.2609.34738","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.34738","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:2609.34738","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.34738","pdf_url":"https://arxiv.org/pdf/2609.34738","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":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214699424.pdf","grobid_xml":"https://content.openalex.org/works/W7214699424.grobid-xml"},"referenced_works_count":22,"referenced_works":["https://openalex.org/W1821462560","https://openalex.org/W2963736842","https://openalex.org/W2973727699","https://openalex.org/W3101005014","https://openalex.org/W4380993498","https://openalex.org/W4382173266","https://openalex.org/W4385570255","https://openalex.org/W4391156874","https://openalex.org/W4391631876","https://openalex.org/W4399837985","https://openalex.org/W4404781935","https://openalex.org/W4409362851","https://openalex.org/W4414939355","https://openalex.org/W6891914185","https://openalex.org/W7116060278","https://openalex.org/W7126272460","https://openalex.org/W7134813685","https://openalex.org/W7160849202","https://openalex.org/W7171395983","https://openalex.org/W7196953391","https://openalex.org/W7204879293","https://openalex.org/W7208838824"],"related_works":["https://openalex.org/W7160849202","https://openalex.org/W7220922622","https://openalex.org/W7223199191","https://openalex.org/W7164013384","https://openalex.org/W4416445216","https://openalex.org/W7171395983","https://openalex.org/W7162112067","https://openalex.org/W7214782337","https://openalex.org/W7138850368","https://openalex.org/W7153091404"],"abstract_inverted_index":{"On-policy":[0],"distillation":[1,171],"(OPD)":[2],"transfers":[3],"knowledge":[4,226],"between":[5],"language":[6],"models":[7],"through":[8],"teacher":[9,19,69,82,116,175,201],"supervision":[10,222],"on":[11,232],"student-generated":[12],"trajectories.":[13],"With":[14],"different":[15],"tokenizers,":[16,166],"a":[17,55,65,68,173,177],"single":[18],"token":[20],"may":[21,74],"require":[22],"multiple":[23,71],"student":[24,63,126,138,150],"tokens":[25,44,73],"to":[26,48,148,168,176],"generate,":[27],"creating":[28],"intermediate":[29],"states":[30],"where":[31],"the":[32,62,76,81,85,120,149,189,208],"event":[33,205,215,218],"is":[34],"entered":[35],"but":[36,80],"not":[37],"yet":[38],"completed.":[39],"Existing":[40],"cross-tokenizer":[41,107,225],"methods":[42],"align":[43],"or":[45],"text":[46],"spans":[47],"construct":[49],"comparable":[50],"prediction":[51],"targets.":[52],"We":[53,99],"study":[54],"complementary":[56,221],"problem":[57],"after":[58,203],"partial":[59,204],"generation:":[60],"once":[61],"produces":[64],"prefix":[66],"of":[67,123,198],"token,":[70],"next":[72],"complete":[75],"same":[77],"remaining":[78],"bytes,":[79],"only":[83],"specifies":[84],"required":[86],"completion":[87,185,194,219],"rather":[88],"than":[89],"how":[90],"probability":[91,108,122,130],"should":[92],"be":[93,230],"divided":[94],"among":[95,132],"these":[96],"valid":[97],"continuations.":[98],"introduce":[100],"Event-Set":[101],"Completion":[102],"Distillation":[103],"(ESCD),":[104],"which":[105],"complements":[106],"alignment":[109],"with":[110],"completion-set":[111],"supervision.":[112],"ESCD":[113],"aggregates":[114],"prefix-related":[115],"events":[117],"and":[118,140,159,165,217],"supervises":[119],"total":[121],"byte-compatible":[124],"one-step":[125,193],"completions,":[127],"avoiding":[128],"tokenizer-dependent":[129],"splits":[131],"individual":[133],"tokens.":[134],"The":[135],"method":[136],"reuses":[137],"trajectories":[139],"predictions,":[141],"requiring":[142],"neither":[143],"additional":[144],"rollouts":[145],"nor":[146],"changes":[147],"vocabulary.":[151],"Experiments":[152],"demonstrate":[153],"consistent":[154],"gains":[155],"in":[156,207],"mathematics,":[157],"code,":[158],"scientific":[160],"reasoning":[161],"across":[162],"model":[163],"families":[164],"extending":[167],"large-scale":[169],"MoE":[170],"from":[172],"1T":[174],"35B":[178],"student.":[179],"Local":[180],"analyses":[181],"show":[182],"that":[183],"retaining":[184],"sets":[186],"better":[187],"matches":[188],"reference":[190],"supervision,":[191],"while":[192],"covers":[195],"over":[196],"99%":[197],"observed":[199],"compatible":[200],"mass":[202],"entry":[206,216],"studied":[209],"tokenizer":[210],"pairs.":[211],"These":[212],"findings":[213],"support":[214],"as":[220],"targets":[223],"for":[224],"transfer.":[227],"Code":[228],"will":[229],"released":[231],"GitHub.":[233]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-09-30T00:00:00"}