{"id":"https://openalex.org/W7220716618","doi":"https://doi.org/10.48550/arxiv.2610.05458","title":"Measuring and Reducing Cross-Vendor Mismatch in Language Models","display_name":"Measuring and Reducing Cross-Vendor Mismatch in Language Models","publication_year":2026,"publication_date":"2026-10-04","ids":{"openalex":"https://openalex.org/W7220716618","doi":"https://doi.org/10.48550/arxiv.2610.05458"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2610.05458","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05458","pdf_url":"https://arxiv.org/pdf/2610.05458","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2610.05458","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5017355779","display_name":"Erland Hilman Fuadi","orcid":"https://orcid.org/0000-0002-9569-2884","observed_orcids":["https://orcid.org/0000-0002-9569-2884"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fuadi, Erland Hilman","raw_affiliation_strings":["University of Artificial Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Artificial Intelligence","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5157709519","display_name":"Chong Tian","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tian, Chong","raw_affiliation_strings":["University of Artificial Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Artificial Intelligence","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083800790","display_name":"Xiaosong Ma","orcid":"https://orcid.org/0000-0003-1261-2496","observed_orcids":["https://orcid.org/0000-0003-1261-2496"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Xiaosong","raw_affiliation_strings":["University of Artificial Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Artificial Intelligence","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5012361506","display_name":"Qirong Ho","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ho, Qirong","raw_affiliation_strings":["University of Artificial Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Artificial Intelligence","institution_ids":[]}]}],"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":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/T11697","display_name":"Numerical Methods and Algorithms","score":0.6255000233650208,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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/T11697","display_name":"Numerical Methods and Algorithms","score":0.6255000233650208,"subfield":{"id":"https://openalex.org/subfields/1703","display_name":"Computational Theory and Mathematics"},"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":0.10239999741315842,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.0982000008225441,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/language-models","display_name":"language models","score":0.6790000200271606},{"id":"https://openalex.org/keywords/mixture-of-experts","display_name":"mixture of experts","score":0.5789999961853027},{"id":"https://openalex.org/keywords/fp16","display_name":"FP16","score":0.6880000233650208},{"id":"https://openalex.org/keywords/low-rank-adapters","display_name":"low-rank adapters","score":0.7279999852180481},{"id":"https://openalex.org/keywords/knowledge-distillation","display_name":"knowledge distillation","score":0.9290000200271606}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6317999958992004},{"id":"https://openalex.org/C134765980","wikidata":"https://www.wikidata.org/wiki/Q879126","display_name":"Bitwise operation","level":2,"score":0.621999979019165},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.4982999861240387},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.4912000000476837},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.478300005197525},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4569000005722046},{"id":"https://openalex.org/C75291252","wikidata":"https://www.wikidata.org/wiki/Q1315756","display_name":"TRACE (psycholinguistics)","level":2,"score":0.4336000084877014},{"id":"https://openalex.org/C140331021","wikidata":"https://www.wikidata.org/wiki/Q1868104","display_name":"Logit","level":2,"score":0.4097000062465668},{"id":"https://openalex.org/C132459708","wikidata":"https://www.wikidata.org/wiki/Q744069","display_name":"Extrapolation","level":2,"score":0.4000999927520752},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3993000090122223},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.37139999866485596},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.3646000027656555},{"id":"https://openalex.org/C96865113","wikidata":"https://www.wikidata.org/wiki/Q2946816","display_name":"Certificate","level":2,"score":0.3598000109195709},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.34209999442100525},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.33799999952316284},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.3361000120639801},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.31679999828338623},{"id":"https://openalex.org/C105795698","wikidata":"https://www.wikidata.org/wiki/Q12483","display_name":"Statistics","level":1,"score":0.310699999332428},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.2987000048160553},{"id":"https://openalex.org/C182306322","wikidata":"https://www.wikidata.org/wiki/Q1779371","display_name":"Order (exchange)","level":2,"score":0.290800005197525},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.29019999504089355},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.27489998936653137},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2703999876976013},{"id":"https://openalex.org/C197947376","wikidata":"https://www.wikidata.org/wiki/Q5155608","display_name":"Comparability","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.26440000534057617},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.26260000467300415},{"id":"https://openalex.org/C2780069185","wikidata":"https://www.wikidata.org/wiki/Q7977945","display_name":"Equivalence (formal languages)","level":2,"score":0.2554999887943268},{"id":"https://openalex.org/C103088060","wikidata":"https://www.wikidata.org/wiki/Q1062839","display_name":"Error detection and correction","level":2,"score":0.25279998779296875},{"id":"https://openalex.org/C114289077","wikidata":"https://www.wikidata.org/wiki/Q3284399","display_name":"Statistical model","level":2,"score":0.2526000142097473}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2610.05458","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05458","pdf_url":"https://arxiv.org/pdf/2610.05458","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.2610.05458","doi":"https://doi.org/10.48550/arxiv.2610.05458","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2610.05458","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2610.05458","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2610.05458","pdf_url":"https://arxiv.org/pdf/2610.05458","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":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7220716618.pdf","grobid_xml":"https://content.openalex.org/works/W7220716618.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7214099313","https://openalex.org/W7214125032","https://openalex.org/W7204280198","https://openalex.org/W7151751846","https://openalex.org/W7214545680","https://openalex.org/W7163193975","https://openalex.org/W7167900054","https://openalex.org/W7221012314","https://openalex.org/W4413059409","https://openalex.org/W7203651398"],"abstract_inverted_index":{"Running":[0],"the":[1,18,24,59,72,82,87,99,104,112,120,125,136,149,156,186,199,205,208],"same":[2,187],"language":[3],"model":[4,19],"on":[5,175,189],"different":[6,14],"graphics":[7],"processing":[8],"unit":[9],"(GPU)":[10],"vendors":[11],"can":[12],"produce":[13],"logits,":[15],"even":[16],"when":[17],"weights":[20],"and":[21,33,51,116,128,132,222],"inputs":[22],"are":[23],"same.":[25],"We":[26,54],"analyze":[27],"cross-vendor":[28,212],"mismatch":[29,60,159],"in":[30,89,135],"two":[31,34],"dense":[32,73,137],"mixture-of-experts":[35],"(MoE)":[36],"models":[37],"with":[38],"five":[39],"metric":[40],"families,":[41],"namely":[42],"bitwise":[43,194],"equality,":[44],"logit":[45,75,126],"differences,":[46],"top-K":[47],"consistency,":[48],"token":[49],"agreement,":[50],"task":[52],"accuracy.":[53],"trace":[55],"one":[56,183],"source":[57],"of":[58,93,103,106,204],"to":[61,69,207],"accumulation":[62],"order":[63],"inside":[64],"vendors'":[65],"matrix":[66],"instructions.":[67],"Upcasting":[68],"FP32":[70,115,192],"reduces":[71],"model's":[74],"error":[76,122,127],"by":[77,214],"43%":[78],"at":[79,96,228],"three":[80],"times":[81,98],"runtime,":[83,100],"yet":[84],"keeping":[85],"only":[86],"MLPs":[88],"BF16":[90],"retains":[91],"94%":[92],"this":[94],"gain":[95],"1.3":[97],"so":[101,210],"most":[102,203],"cost":[105,221],"full":[107],"upcasting":[108],"buys":[109],"little.":[110],"In":[111],"MoE":[113],"models,":[114],"FP16":[117,133],"both":[118,219],"lower":[119],"probability":[121],"but":[123],"raise":[124],"change":[129],"expert":[130],"selection,":[131],"fails":[134],"model.":[138],"An":[139],"output-head":[140],"low-rank":[141],"adapter":[142],"(LoRA)":[143],"does":[144,153],"not":[145,154],"help":[146],"either,":[147],"since":[148],"final":[150],"hidden":[151],"state":[152],"predict":[155],"mismatch.":[157],"The":[158],"also":[160],"carries":[161],"into":[162],"training.":[163],"With":[164],"every":[165],"seed":[166],"fixed,":[167],"a":[168,172,215],"student":[169],"distilled":[170,184],"from":[171,182,185],"teacher":[173,188],"running":[174],"AMD":[176],"answers":[177],"431":[178],"MMLU":[179],"questions":[180],"differently":[181],"NVIDIA.":[190],"Under":[191],"upcasting,":[193],"equality":[195],"barely":[196],"changes":[197],"while":[198],"output":[200],"distributions":[201],"move":[202],"way":[206],"reference,":[209],"judging":[211],"agreement":[213],"single":[216],"measure":[217],"misreads":[218],"its":[220,223],"gains.":[224],"Code":[225],"is":[226],"available":[227],"https://github.com/crova-project/crova.":[229]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-10-07T00:00:00"}