{"id":"https://openalex.org/W7155203048","doi":"https://doi.org/10.48550/arxiv.2604.19079","title":"Reducing the Offline-Streaming Gap for Unified ASR Transducer with Consistency Regularization","display_name":"Reducing the Offline-Streaming Gap for Unified ASR Transducer with Consistency Regularization","publication_year":2026,"publication_date":"2026-04-21","ids":{"openalex":"https://openalex.org/W7155203048","doi":"https://doi.org/10.48550/arxiv.2604.19079"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.19079","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.19079","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":null,"license_id":null,"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.2604.19079","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5054253945","display_name":"Andrei Andrusenko","orcid":"https://orcid.org/0000-0002-8697-832X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Andrusenko, Andrei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134318510","display_name":"Vladimir Bataev","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bataev, Vladimir","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134227412","display_name":"Lilit Grigoryan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Grigoryan, Lilit","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134316467","display_name":"Nune Tadevosyan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tadevosyan, Nune","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026088310","display_name":"Vitaly Lavrukhin","orcid":"https://orcid.org/0009-0006-7866-8301"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lavrukhin, Vitaly","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5134267566","display_name":"Boris Ginsburg","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ginsburg, Boris","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/T10201","display_name":"Speech Recognition and Synthesis","score":0.7103000283241272,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.7103000283241272,"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/T10860","display_name":"Speech and Audio Processing","score":0.26179999113082886,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T10403","display_name":"Phonetics and Phonology Research","score":0.0024999999441206455,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.6614000201225281},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.5593000054359436},{"id":"https://openalex.org/keywords/latency","display_name":"Latency (audio)","score":0.5339999794960022},{"id":"https://openalex.org/keywords/unification","display_name":"Unification","score":0.5333999991416931},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.4156000018119812},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.3955000042915344},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.3379000127315521}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7623999714851379},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.6614000201225281},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.5593000054359436},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.5339999794960022},{"id":"https://openalex.org/C96146094","wikidata":"https://www.wikidata.org/wiki/Q609057","display_name":"Unification","level":2,"score":0.5333999991416931},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.42649999260902405},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.4156000018119812},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.3955000042915344},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3822999894618988},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3587999939918518},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3379000127315521},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.3280999958515167},{"id":"https://openalex.org/C45493050","wikidata":"https://www.wikidata.org/wiki/Q7884934","display_name":"Unified Model","level":2,"score":0.3221000134944916},{"id":"https://openalex.org/C56318395","wikidata":"https://www.wikidata.org/wiki/Q215928","display_name":"Transducer","level":2,"score":0.3084000051021576},{"id":"https://openalex.org/C46637626","wikidata":"https://www.wikidata.org/wiki/Q6693015","display_name":"Low latency (capital markets)","level":2,"score":0.29280000925064087},{"id":"https://openalex.org/C157138929","wikidata":"https://www.wikidata.org/wiki/Q570","display_name":"Loudspeaker","level":2,"score":0.289000004529953},{"id":"https://openalex.org/C196921405","wikidata":"https://www.wikidata.org/wiki/Q786431","display_name":"Online algorithm","level":2,"score":0.28529998660087585},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2849000096321106},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2759000062942505},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.27549999952316284},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.2721000015735626},{"id":"https://openalex.org/C2780102126","wikidata":"https://www.wikidata.org/wiki/Q10928179","display_name":"Online and offline","level":2,"score":0.2687999904155731},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.25589999556541443}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.19079","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.19079","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.19079","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.19079","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions","score":0.545724630355835}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Unification":[0],"of":[1,76],"automatic":[2],"speech":[3],"recognition":[4],"(ASR)":[5],"systems":[6],"reduces":[7],"development":[8],"and":[9,23,43,56,67,104,110,118],"maintenance":[10],"costs,":[11],"but":[12],"training":[13,38,86,111],"a":[14,31,47],"single":[15,48],"model":[16,108,121],"to":[17,106],"perform":[18],"well":[19],"in":[20],"both":[21,41],"offline":[22,42,66,102],"low-latency":[24],"streaming":[25,44,68,95],"settings":[26],"remains":[27],"challenging.":[28],"We":[29],"present":[30],"Unified":[32,115],"ASR":[33,116],"framework":[34,117],"for":[35,79],"Transducer":[36],"(RNNT)":[37],"that":[39,90],"supports":[40],"decoding":[45],"within":[46],"model,":[49],"using":[50],"chunk-limited":[51],"attention":[52],"with":[53],"right":[54],"context":[55],"dynamic":[57],"chunked":[58],"convolutions.":[59],"To":[60],"further":[61],"close":[62],"the":[63,91,119],"gap":[64],"between":[65],"performance,":[69],"we":[70],"introduce":[71],"an":[72],"efficient":[73],"Triton":[74],"implementation":[75],"mode-consistency":[77],"regularization":[78],"RNNT":[80],"(MCR-RNNT),":[81],"which":[82],"encourages":[83],"agreement":[84],"across":[85],"modes.":[87],"Experiments":[88],"show":[89],"proposed":[92,114],"approach":[93],"improves":[94],"accuracy":[96],"at":[97],"low":[98],"latency":[99],"while":[100],"preserving":[101],"performance":[103],"scaling":[105],"larger":[107],"sizes":[109],"datasets.":[112],"The":[113],"English":[120],"checkpoint":[122],"are":[123],"open-sourced.":[124]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-23T00:00:00"}