{"id":"https://openalex.org/W7204280198","doi":"https://doi.org/10.48550/arxiv.2608.24267","title":"Cross-Stack Validation of Language-Model Training: A Clinical Fine-Tuning Case Study","display_name":"Cross-Stack Validation of Language-Model Training: A Clinical Fine-Tuning Case Study","publication_year":2026,"publication_date":"2026-08-25","ids":{"openalex":"https://openalex.org/W7204280198","doi":"https://doi.org/10.48550/arxiv.2608.24267"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.24267","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24267","pdf_url":"https://arxiv.org/pdf/2608.24267","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":"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/2608.24267","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5148245028","display_name":"Thang Tran","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tran, Thang","raw_affiliation_strings":["CloudKites AI Lab"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CloudKites AI Lab","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5148397444","display_name":"Lan Dang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I56590836","display_name":"Monash University","ror":"https://ror.org/02bfwt286","country_code":"AU","type":"education","lineage":["https://openalex.org/I56590836"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Dang, Lan","raw_affiliation_strings":["Monash Business School, Monash University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Monash Business School, Monash University","institution_ids":["https://openalex.org/I56590836"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":4,"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/T10743","display_name":"Software Testing and Debugging Techniques","score":0.9976000189781189,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/T10743","display_name":"Software Testing and Debugging Techniques","score":0.9976000189781189,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/T10260","display_name":"Software Engineering Research","score":0.00039999998989515007,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T10126","display_name":"Logic, programming, and type systems","score":0.0003000000142492354,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/differential-testing","display_name":"differential testing","score":0.8980000019073486},{"id":"https://openalex.org/keywords/language-model-fine-tuning","display_name":"language model fine-tuning","score":0.8080000281333923},{"id":"https://openalex.org/keywords/lora","display_name":"LoRa","score":0.6729999780654907},{"id":"https://openalex.org/keywords/qwen3","display_name":"Qwen3","score":0.6399999856948853},{"id":"https://openalex.org/keywords/medical-question-answering","display_name":"medical question answering","score":0.871999979019165},{"id":"https://openalex.org/keywords/pytorch","display_name":"PyTorch","score":0.7789999842643738},{"id":"https://openalex.org/keywords/numbat","display_name":"numbat","score":0.5950000286102295},{"id":"https://openalex.org/keywords/fault-detection","display_name":"fault detection","score":0.5429999828338623},{"id":"https://openalex.org/keywords/memory-models","display_name":"memory models","score":0.3490000069141388},{"id":"https://openalex.org/keywords/text-rendering","display_name":"text rendering","score":0.38499999046325684}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7623000144958496},{"id":"https://openalex.org/C55166926","wikidata":"https://www.wikidata.org/wiki/Q2892946","display_name":"Oracle","level":2,"score":0.5516999959945679},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.4569000005722046},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41339999437332153},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.4018999934196472},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3921000063419342},{"id":"https://openalex.org/C2780615836","wikidata":"https://www.wikidata.org/wiki/Q2471869","display_name":"USable","level":2,"score":0.39070001244544983},{"id":"https://openalex.org/C205711294","wikidata":"https://www.wikidata.org/wiki/Q176953","display_name":"Rendering (computer graphics)","level":2,"score":0.36250001192092896},{"id":"https://openalex.org/C26713055","wikidata":"https://www.wikidata.org/wiki/Q245962","display_name":"Implementation","level":2,"score":0.35019999742507935},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.33889999985694885},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.33809998631477356},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.3276999890804291},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.31790000200271606},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.30480000376701355},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3027999997138977},{"id":"https://openalex.org/C35651441","wikidata":"https://www.wikidata.org/wiki/Q625303","display_name":"Independence (probability theory)","level":2,"score":0.30219998955726624},{"id":"https://openalex.org/C63540848","wikidata":"https://www.wikidata.org/wiki/Q3140932","display_name":"Fault tolerance","level":2,"score":0.287200003862381},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.28349998593330383},{"id":"https://openalex.org/C93226319","wikidata":"https://www.wikidata.org/wiki/Q193137","display_name":"Differential (mechanical device)","level":2,"score":0.27480000257492065},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2669999897480011},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2632000148296356},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.25780001282691956},{"id":"https://openalex.org/C12186640","wikidata":"https://www.wikidata.org/wiki/Q6815743","display_name":"Memory model","level":3,"score":0.25589999556541443},{"id":"https://openalex.org/C143724316","wikidata":"https://www.wikidata.org/wiki/Q312468","display_name":"Series (stratigraphy)","level":2,"score":0.25369998812675476}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.24267","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24267","pdf_url":"https://arxiv.org/pdf/2608.24267","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.24267","doi":"https://doi.org/10.48550/arxiv.2608.24267","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.24267","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:2608.24267","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24267","pdf_url":"https://arxiv.org/pdf/2608.24267","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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204280198.pdf","grobid_xml":"https://content.openalex.org/works/W7204280198.grobid-xml"},"referenced_works_count":12,"referenced_works":["https://openalex.org/W2908510526","https://openalex.org/W2972522091","https://openalex.org/W3104663419","https://openalex.org/W3162922479","https://openalex.org/W3168867926","https://openalex.org/W4221154592","https://openalex.org/W4284686707","https://openalex.org/W4327594616","https://openalex.org/W4389141459","https://openalex.org/W4389500716","https://openalex.org/W4407230117","https://openalex.org/W4414939355"],"related_works":["https://openalex.org/W7212292283","https://openalex.org/W7214384306","https://openalex.org/W7214072487","https://openalex.org/W4414735796","https://openalex.org/W7220716618","https://openalex.org/W7219884174","https://openalex.org/W7220748413","https://openalex.org/W7214573030","https://openalex.org/W4402667112","https://openalex.org/W7136204024"],"abstract_inverted_index":{"Neural":[0],"network":[1],"training":[2,49],"has":[3,269],"an":[4,128,258],"oracle":[5],"problem:":[6],"a":[7,14,57,75,79,95,113,149,205,235,246,252,262],"run":[8],"can":[9,51],"converge":[10],"normally":[11],"and":[12,65,90,94,104,109,125,136,162,272],"yield":[13],"usable":[15],"model":[16,86,199,239],"while":[17],"the":[18,63,91,100,102,105,152,166,193,197,202,222,242,264,273],"software":[19,188],"beneath":[20],"it":[21,111],"computes":[22],"something":[23],"other":[24],"than":[25,62,221],"specified.":[26],"Almost":[27],"all":[28],"such":[29],"work":[30,249],"runs":[31],"on":[32,160,196],"one":[33,171],"stack,":[34,101],"so":[35],"there":[36],"is":[37,275],"rarely":[38],"anything":[39],"independent":[40,129],"to":[41,112,255],"check":[42],"against.":[43],"We":[44,73],"study":[45],"whether":[46],"independently":[47],"implemented":[48],"stacks":[50],"serve":[52],"as":[53],"differential":[54,70],"oracles":[55],"for":[56,187],"whole":[58],"fine-tuning":[59],"pipeline,":[60],"rather":[61],"operators":[64],"inference":[66],"paths":[67],"that":[68,178],"prior":[69],"testing":[71],"targets.":[72],"define":[74],"trajectory-level":[76],"protocol":[77],"--":[78,108],"shared":[80],"specification,":[81],"cross-check":[82],"points":[83],"spanning":[84,148],"arithmetic,":[85],"loading,":[87],"data":[88],"rendering":[89],"learning":[92],"trajectory,":[93],"separation":[96],"of":[97,99,116,170,184],"independence":[98],"orchestration":[103],"language":[106,236],"runtime":[107,274],"apply":[110],"LoRA":[114],"adaptation":[115],"Qwen3-0.6B":[117],"over":[118],"168,574":[119],"clinical":[120,209],"question-answer":[121],"pairs":[122],"under":[123,126],"PyTorch":[124],"numbat,":[127],"framework":[130],"written":[131],"in":[132,207],"Zig,":[133],"driven":[134],"natively":[135],"through":[137],"its":[138],"C":[139],"interface":[140,265],"from":[141,234,241],"six":[142],"languages.":[143],"Across":[144],"42":[145],"paired":[146],"evaluations":[147],"full":[150],"epoch":[151,167],"two":[153,183,244],"stacks'":[154],"held-out":[155,214],"cross-entropy":[156],"differs":[157,240],"by":[158],"0.134%":[159],"average,":[161],"four":[163,229],"implementations":[164],"end":[165],"within":[168],"0.15%":[169],"another.":[172],"The":[173,190],"comparison":[174],"exposed":[175],"17":[176],"faults":[177,224,230],"single-implementation":[179],"development":[180],"had":[181],"missed,":[182],"them":[185],"notable":[186],"engineering.":[189],"fault":[191],"with":[192],"largest":[194],"effect":[195],"trained":[198],"lay":[200],"outside":[201],"numerical":[203],"kernels:":[204],"mismatch":[206],"how":[208],"text":[210],"was":[211],"rendered":[212],"moved":[213],"loss":[215],"0.15,":[216],"some":[217],"500":[218],"times":[219],"more":[220],"arithmetic":[223],"found":[225],"beside":[226],"it.":[227],"And":[228],"were":[231],"reachable":[232],"only":[233],"whose":[237],"memory":[238],"first":[243],"implementations:":[245],"scheduler":[247],"migrating":[248],"across":[250],"threads,":[251],"collector":[253],"blind":[254],"device":[256],"memory,":[257],"ownership":[259],"discipline":[260],"needing":[261],"primitive":[263],"lacked.":[266],"Implementation":[267],"diversity":[268],"several":[270],"axes,":[271],"one.":[276]},"counts_by_year":[{"year":2026,"cited_by_count":4}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-08-27T00:00:00"}