{"id":"https://openalex.org/W7214636339","doi":"https://doi.org/10.48550/arxiv.2609.30846","title":"I-Parakeet: Integer-Only Conformer ASR on Mobile NPU","display_name":"I-Parakeet: Integer-Only Conformer ASR on Mobile NPU","publication_year":2026,"publication_date":"2026-09-25","ids":{"openalex":"https://openalex.org/W7214636339","doi":"https://doi.org/10.48550/arxiv.2609.30846"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.30846","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.30846","pdf_url":"https://arxiv.org/pdf/2609.30846","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","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/2609.30846","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5153321323","display_name":"Taichi Nishimura","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405282746","display_name":"Sony Interactive Entertainment (Japan)","ror":"https://ror.org/00dcvss65","country_code":"JP","type":"company","lineage":["https://openalex.org/I4405282746"]}],"countries":["JP"],"is_corresponding":true,"raw_author_name":"Nishimura, Taichi","raw_affiliation_strings":["Sony Interactive Entertainment , Tokyo , Japan"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Sony Interactive Entertainment , Tokyo , Japan","institution_ids":["https://openalex.org/I4405282746"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5153321323"],"corresponding_institution_ids":["https://openalex.org/I4405282746"],"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.8076000213623047,"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":0.8076000213623047,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.17190000414848328,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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":0.0020000000949949026,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/conformers","display_name":"conformational isomers","score":0.6470000147819519},{"id":"https://openalex.org/keywords/speech-recognition","display_name":"speech recognition","score":0.6340000033378601},{"id":"https://openalex.org/keywords/edge-deployment","display_name":"edge deployment","score":0.8920000195503235},{"id":"https://openalex.org/keywords/batch-normalization","display_name":"batch normalization","score":0.6790000200271606},{"id":"https://openalex.org/keywords/quantization","display_name":"quantization","score":0.4189999997615814},{"id":"https://openalex.org/keywords/real-time-speech-recognition","display_name":"real-time speech recognition","score":0.6019999980926514}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7339000105857849},{"id":"https://openalex.org/C122048520","wikidata":"https://www.wikidata.org/wiki/Q2913954","display_name":"Percentile","level":2,"score":0.6154000163078308},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.5015000104904175},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.49050000309944153},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.48069998621940613},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.47749999165534973},{"id":"https://openalex.org/C97137487","wikidata":"https://www.wikidata.org/wiki/Q729138","display_name":"Integer (computer science)","level":2,"score":0.4763000011444092},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.454800009727478},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.43479999899864197},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.3894999921321869},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.3837999999523163},{"id":"https://openalex.org/C17020691","wikidata":"https://www.wikidata.org/wiki/Q139677","display_name":"Operator (biology)","level":5,"score":0.3698999881744385},{"id":"https://openalex.org/C187691185","wikidata":"https://www.wikidata.org/wiki/Q2020720","display_name":"Grid","level":2,"score":0.3684000074863434},{"id":"https://openalex.org/C141353440","wikidata":"https://www.wikidata.org/wiki/Q182221","display_name":"Fuse (electrical)","level":2,"score":0.3546999990940094},{"id":"https://openalex.org/C186967261","wikidata":"https://www.wikidata.org/wiki/Q5082128","display_name":"Mobile device","level":2,"score":0.3538999855518341},{"id":"https://openalex.org/C28826006","wikidata":"https://www.wikidata.org/wiki/Q33521","display_name":"Applied mathematics","level":1,"score":0.3377000093460083},{"id":"https://openalex.org/C118615104","wikidata":"https://www.wikidata.org/wiki/Q121416","display_name":"Discrete mathematics","level":1,"score":0.33000001311302185},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.31459999084472656},{"id":"https://openalex.org/C99821215","wikidata":"https://www.wikidata.org/wiki/Q1136583","display_name":"Swap (finance)","level":2,"score":0.3116999864578247},{"id":"https://openalex.org/C49154492","wikidata":"https://www.wikidata.org/wiki/Q5300","display_name":"Central processing unit","level":2,"score":0.30790001153945923},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.3061999976634979},{"id":"https://openalex.org/C22029948","wikidata":"https://www.wikidata.org/wiki/Q45089","display_name":"Dice","level":2,"score":0.3059999942779541},{"id":"https://openalex.org/C122383733","wikidata":"https://www.wikidata.org/wiki/Q865920","display_name":"Approximation error","level":2,"score":0.30059999227523804},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.27489998936653137},{"id":"https://openalex.org/C40969351","wikidata":"https://www.wikidata.org/wiki/Q3516228","display_name":"Word error rate","level":2,"score":0.27390000224113464},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.27149999141693115},{"id":"https://openalex.org/C84211073","wikidata":"https://www.wikidata.org/wiki/Q117879","display_name":"Floating point","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C186394612","wikidata":"https://www.wikidata.org/wiki/Q7098942","display_name":"Optimal design","level":2,"score":0.26010000705718994},{"id":"https://openalex.org/C14166107","wikidata":"https://www.wikidata.org/wiki/Q253829","display_name":"Net (polyhedron)","level":2,"score":0.25429999828338623}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.30846","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.30846","pdf_url":"https://arxiv.org/pdf/2609.30846","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.30846","doi":"https://doi.org/10.48550/arxiv.2609.30846","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.30846","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:2609.30846","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.30846","pdf_url":"https://arxiv.org/pdf/2609.30846","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":"public-domain","license_id":"https://openalex.org/licenses/public-domain","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214636339.pdf","grobid_xml":"https://content.openalex.org/works/W7214636339.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4286796096","https://openalex.org/W3142081304","https://openalex.org/W4362598042","https://openalex.org/W4389983270","https://openalex.org/W4413930132","https://openalex.org/W4221138270","https://openalex.org/W7162029275","https://openalex.org/W3201561669","https://openalex.org/W7166497543","https://openalex.org/W7167035629"],"abstract_inverted_index":{"In":[0],"this":[1],"paper,":[2],"we":[3,75,108],"propose":[4],"I-Parakeet,":[5],"an":[6,77,134],"integer-only":[7,105],"implementation":[8],"of":[9,39,80,87,119,128,165],"NVIDIA's":[10],"Parakeet-CTC":[11],"(0.6B":[12],"parameters)":[13],"that":[14,114],"runs":[15],"on":[16,35,153,157],"a":[17,110,124,158,162,170],"smartphone":[18],"NPU":[19,160],"without":[20],"any":[21],"floating-point":[22],"operator":[23],"or":[24],"CPU":[25,171],"fallback.":[26],"Modern":[27],"Conformer":[28],"ASR":[29],"models":[30,44],"are":[31,72],"hard":[32],"to":[33,48],"deploy":[34],"edge":[36],"devices":[37],"because":[38],"their":[40],"size,":[41],"and":[42,100,141],"quantized":[43],"still":[45],"fall":[46],"back":[47],"floating":[49],"point":[50],"for":[51,137,144],"numerically":[52],"sensitive":[53],"operations.":[54,106],"This":[55],"prevents":[56],"them":[57],"from":[58],"fully":[59],"exploiting":[60],"integer":[61,78],"accelerators":[62],"such":[63],"as":[64],"mobile":[65],"NPUs.":[66],"To":[67],"achieve":[68],"this,":[69],"our":[70],"contributions":[71],"threefold.":[73],"First,":[74],"derive":[76],"formulation":[79],"the":[81,85,88,101,116,120,138,145],"relative-positional":[82],"self-attention":[83],"at":[84,161],"core":[86],"Conformer.":[89],"We":[90],"fuse":[91],"its":[92],"two":[93,131],"score":[94],"branches":[95],"with":[96],"different":[97],"quantization":[98],"scales":[99],"relative":[102],"shift":[103],"into":[104],"Second,":[107],"introduce":[109],"minimax-optimized":[111],"Swish":[112,121],"approximation":[113],"minimizes":[115],"maximum":[117],"error":[118],"output.":[122],"Third,":[123],"layer-wise":[125],"range":[126],"analysis":[127],"activations":[129],"yields":[130],"targeted":[132],"remedies:":[133],"INT16":[135],"grid":[136],"BatchNorm":[139],"output":[140],"percentile":[142],"calibration":[143],"heavy-tailed":[146],"pre-encoder":[147],"activations.":[148],"I-Parakeet":[149],"achieves":[150],"4.97%":[151],"WER":[152],"LibriSpeech":[154],"test-other,":[155],"running":[156],"Qualcomm":[159],"real-time":[163],"factor":[164],"0.048,":[166],"7.5x":[167],"faster":[168],"than":[169],"baseline.":[172]},"counts_by_year":[],"updated_date":"2026-11-06T12:05:46.655746","created_date":"2026-09-29T00:00:00"}