{"id":"https://openalex.org/W7225326490","doi":"https://doi.org/10.5281/zenodo.23971226","title":"thox-edge-ops v2: fused decode kernels for small models on THOX edge hardware","display_name":"thox-edge-ops v2: fused decode kernels for small models on THOX edge hardware","publication_year":2026,"publication_date":"2026-11-05","ids":{"openalex":"https://openalex.org/W7225326490","doi":"https://doi.org/10.5281/zenodo.23971226"},"language":"en","primary_location":{"id":"doi:10.5281/zenodo.23971226","doi":"https://doi.org/10.5281/zenodo.23971226","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23971226","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.5281/zenodo.23971226","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5160373893","display_name":"Tommy Xaypanya","orcid":"https://orcid.org/0009-0000-4425-9565","observed_orcids":["https://orcid.org/0009-0000-4425-9565"]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Xaypanya, Tommy","raw_affiliation_strings":["THOX.ai LLC"],"raw_orcid":"https://orcid.org/0009-0000-4425-9565","affiliations":[{"raw_affiliation_string":"THOX.ai LLC","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5160373893"],"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.7602999806404114,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.7602999806404114,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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":0.16820000112056732,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.01489999983459711,"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/small-language-models","display_name":"small language models","score":0.8820000290870667},{"id":"https://openalex.org/keywords/edge-hardware","display_name":"edge hardware","score":0.7250000238418579},{"id":"https://openalex.org/keywords/rmsnorm","display_name":"RMSNorm","score":0.6119999885559082},{"id":"https://openalex.org/keywords/rotary-positional-embeddings","display_name":"rotary positional embeddings","score":0.5640000104904175},{"id":"https://openalex.org/keywords/silu-activation","display_name":"SiLU activation","score":0.7639999985694885},{"id":"https://openalex.org/keywords/gelu-activation","display_name":"GELU activation","score":0.7429999709129333},{"id":"https://openalex.org/keywords/cpu-inference","display_name":"CPU-only inference","score":0.6579999923706055},{"id":"https://openalex.org/keywords/jetson-orin","display_name":"Jetson Orin","score":0.6809999942779541},{"id":"https://openalex.org/keywords/apple-silicon","display_name":"Apple Silicon","score":0.6330000162124634}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7465999722480774},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5263000130653381},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.5217999815940857},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.4717999994754791},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.4616999924182892},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.4472000002861023},{"id":"https://openalex.org/C56951928","wikidata":"https://www.wikidata.org/wiki/Q3539213","display_name":"Trimming","level":2,"score":0.4372999966144562},{"id":"https://openalex.org/C79581498","wikidata":"https://www.wikidata.org/wiki/Q1367530","display_name":"Suite","level":2,"score":0.4318999946117401},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.4106999933719635},{"id":"https://openalex.org/C67953723","wikidata":"https://www.wikidata.org/wiki/Q192525","display_name":"Workstation","level":2,"score":0.3846000134944916},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.38269999623298645},{"id":"https://openalex.org/C78371743","wikidata":"https://www.wikidata.org/wiki/Q1672829","display_name":"Optical proximity correction","level":3,"score":0.3788999915122986},{"id":"https://openalex.org/C121684516","wikidata":"https://www.wikidata.org/wiki/Q7600677","display_name":"Computer graphics (images)","level":1,"score":0.3732999861240387},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.36730000376701355},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.3628999888896942},{"id":"https://openalex.org/C169590947","wikidata":"https://www.wikidata.org/wiki/Q47506","display_name":"Compiler","level":2,"score":0.3440000116825104},{"id":"https://openalex.org/C168065819","wikidata":"https://www.wikidata.org/wiki/Q845566","display_name":"Debugging","level":2,"score":0.34139999747276306},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.32710000872612},{"id":"https://openalex.org/C113174947","wikidata":"https://www.wikidata.org/wiki/Q2859736","display_name":"Tree (set theory)","level":2,"score":0.3221000134944916},{"id":"https://openalex.org/C150552126","wikidata":"https://www.wikidata.org/wiki/Q339387","display_name":"SIMD","level":2,"score":0.3206000030040741},{"id":"https://openalex.org/C459310","wikidata":"https://www.wikidata.org/wiki/Q117801","display_name":"Computational science","level":1,"score":0.3176000118255615},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.3091999888420105},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.30090001225471497},{"id":"https://openalex.org/C54732982","wikidata":"https://www.wikidata.org/wiki/Q1415345","display_name":"Modulo","level":2,"score":0.3003000020980835},{"id":"https://openalex.org/C9395851","wikidata":"https://www.wikidata.org/wiki/Q177929","display_name":"Stack (abstract data type)","level":2,"score":0.29679998755455017},{"id":"https://openalex.org/C2780870223","wikidata":"https://www.wikidata.org/wiki/Q1004415","display_name":"Runtime system","level":2,"score":0.29319998621940613},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.2689000070095062},{"id":"https://openalex.org/C157202957","wikidata":"https://www.wikidata.org/wiki/Q1659609","display_name":"Image warping","level":2,"score":0.2669999897480011},{"id":"https://openalex.org/C80899671","wikidata":"https://www.wikidata.org/wiki/Q1304193","display_name":"Vertex (graph theory)","level":3,"score":0.2660999894142151},{"id":"https://openalex.org/C3018397939","wikidata":"https://www.wikidata.org/wiki/Q3644502","display_name":"Open source","level":3,"score":0.2630000114440918},{"id":"https://openalex.org/C135257023","wikidata":"https://www.wikidata.org/wiki/Q691358","display_name":"Domain-specific language","level":2,"score":0.26249998807907104},{"id":"https://openalex.org/C139983466","wikidata":"https://www.wikidata.org/wiki/Q17194","display_name":"Video Graphics Array","level":3,"score":0.2621999979019165},{"id":"https://openalex.org/C854659","wikidata":"https://www.wikidata.org/wiki/Q1859284","display_name":"Message passing","level":2,"score":0.25760000944137573},{"id":"https://openalex.org/C45347329","wikidata":"https://www.wikidata.org/wiki/Q5166604","display_name":"Convolution (computer science)","level":3,"score":0.2574999928474426},{"id":"https://openalex.org/C119701452","wikidata":"https://www.wikidata.org/wiki/Q5165881","display_name":"Control reconfiguration","level":2,"score":0.2556000053882599},{"id":"https://openalex.org/C531996449","wikidata":"https://www.wikidata.org/wiki/Q486487","display_name":"Windows NT","level":3,"score":0.2540000081062317},{"id":"https://openalex.org/C76970557","wikidata":"https://www.wikidata.org/wiki/Q1869750","display_name":"Loop unrolling","level":3,"score":0.250900000333786}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.5281/zenodo.23971226","doi":"https://doi.org/10.5281/zenodo.23971226","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23971226","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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"},{"id":"doi:10.5281/zenodo.23971227","doi":"https://doi.org/10.5281/zenodo.23971227","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23971227","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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":"doi:10.5281/zenodo.23971226","doi":"https://doi.org/10.5281/zenodo.23971226","is_oa":true,"landing_page_url":"https://doi.org/10.5281/zenodo.23971226","pdf_url":null,"source":{"id":"https://openalex.org/S4306400562","display_name":"Zenodo (CERN European Organization for Nuclear Research)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I67311998","host_organization_name":"European Organization for Nuclear Research","host_organization_lineage":["https://openalex.org/I67311998"],"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":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7222698571","https://openalex.org/W7112220246","https://openalex.org/W7165171290","https://openalex.org/W7128704250","https://openalex.org/W7222954758","https://openalex.org/W7223018399","https://openalex.org/W7162426210","https://openalex.org/W7165882840","https://openalex.org/W7171319592","https://openalex.org/W3017598548"],"abstract_inverted_index":{"Small":[0],"language":[1],"models":[2],"on":[3,118,124],"edge":[4],"hardware":[5,160],"spend":[6],"their":[7],"decode":[8,95],"step":[9],"in":[10],"launch":[11],"overhead":[12],"and":[13,37,44,50,62,67,108,115,130,150,158,166,177,196],"memory":[14],"traffic,":[15],"not":[16,202],"arithmetic.":[17],"thox-edge-ops":[18],"v2":[19,195],"is":[20,169],"a":[21,30,51,140,159],"set":[22],"of":[23],"six":[24],"fused":[25,40,52,113],"inference":[26],"operators":[27],"(RMSNorm":[28],"with":[29,58,81,188],"weight":[31],"offset":[32],"that":[33],"covers":[34],"the":[35,71,82,92,97,112,125,144,153,155,163,178],"Llama":[36],"Gemma":[38],"conventions,":[39],"residual-add":[41],"RMSNorm,":[42],"SiLU":[43],"tanh-GELU":[45],"gated":[46],"activations,":[47],"rotary":[48,131],"embedding,":[49],"per-head":[53],"QK-norm":[54,114],"plus":[55],"RoPE":[56,116],"prologue)":[57],"one":[59],"source":[60],"tree":[61],"three":[63],"backends:":[64],"CUDA,":[65],"Metal":[66],"CPU,":[68,126],"distributed":[69],"through":[70],"Hugging":[72],"Face":[73],"Kernel":[74],"Hub":[75,191],"as":[76],"23":[77],"ABI-checked":[78],"x86_64":[79,164],"builds,":[80],"full":[83],"438-test":[84],"suite":[85],"passing":[86],"against":[87],"3":[88],"builds":[89],"downloaded":[90],"from":[91],"Hub.":[93],"At":[94],"shapes":[96],"GPU":[98],"kernels":[99],"run":[100],"6.90x":[101],"faster":[102,110],"than":[103],"eager":[104],"PyTorch":[105],"for":[106,111,171,194],"RMSNorm":[107,127],"24.43x":[109],"prologue":[117],"an":[119],"NVIDIA":[120],"RTX":[121],"4060":[122],"Ti;":[123],"runs":[128],"2.42x":[129],"embedding":[132],"6.75x":[133],"faster.":[134],"The":[135],"revised":[136],"public":[137],"edition":[138],"adds":[139],"deployment":[141],"section":[142],"covering":[143],"THOX":[145],"stack":[146],"(edge":[147],"hardware,":[148],"ThoxOS":[149],"ThoxInux,":[151],"MeshStack,":[152],"kernels,":[154],"small":[156],"models)":[157],"qualification":[161],"matrix:":[162],"workstation":[165],"hub":[167],"class":[168,181],"MEASURED":[170],"v2;":[172],"Jetson":[173],"Orin,":[174],"Apple":[175],"silicon":[176],"Arm":[179,198],"device":[180],"(ThoxAir,":[182],"ThoxMini,":[183],"ThoxClip)":[184],"are":[185],"targets":[186],"only,":[187],"no":[189,197],"aarch64":[190],"build":[192],"published":[193],"measurement":[199],"claimed.":[200],"Preprint,":[201],"peer":[203],"reviewed.":[204],"THOX.ai":[205],"Research.":[206]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-11-06T00:00:00"}