{"id":"https://openalex.org/W7225165697","doi":"https://doi.org/10.48550/arxiv.2611.00624","title":"XLA from the Trenches: Correctness on TPU Workloads at Scale","display_name":"XLA from the Trenches: Correctness on TPU Workloads at Scale","publication_year":2026,"publication_date":"2026-10-31","ids":{"openalex":"https://openalex.org/W7225165697","doi":"https://doi.org/10.48550/arxiv.2611.00624"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2611.00624","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2611.00624","pdf_url":"https://arxiv.org/pdf/2611.00624","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2611.00624","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5042000826","display_name":"Ignacio Cano","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cano, Ignacio","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003317570","display_name":"Amir Saadat","orcid":"https://orcid.org/0000-0002-8089-2381","observed_orcids":["https://orcid.org/0000-0002-8089-2381"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saadat, Amir","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130272300","display_name":"Rishipal Singh Bhatia","orcid":"https://orcid.org/0009-0001-6682-8105","observed_orcids":["https://orcid.org/0009-0001-6682-8105"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bhatia, Rishipal Singh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017864929","display_name":"Samuel Ginzburg","orcid":"https://orcid.org/0000-0001-8878-0811","observed_orcids":["https://orcid.org/0000-0001-8878-0811"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ginzburg, Samuel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160410637","display_name":"Patrick C. Toulme","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Toulme, Patrick C.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027146672","display_name":"Ioannis Zarkadas","orcid":"https://orcid.org/0009-0000-2364-3891","observed_orcids":["https://orcid.org/0009-0000-2364-3891"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zarkadas, Ioannis","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049448714","display_name":"Pavanesh Narayanan","orcid":"https://orcid.org/0009-0007-7097-6227","observed_orcids":["https://orcid.org/0009-0007-7097-6227"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Narayanan, Praveen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029139039","display_name":"Pratik Bhatu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bhatu, Pratik","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030478462","display_name":"Nicolas Basile","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Basile, Nicolas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078056312","display_name":"Alexander Dietm\u00fcller","orcid":"https://orcid.org/0000-0003-3769-3958","observed_orcids":["https://orcid.org/0000-0003-3769-3958"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dietm\u00fcller, Alexander","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160527307","display_name":"Ruishan Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Ruishan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037374138","display_name":"Jacob Wahlgren","orcid":"https://orcid.org/0000-0003-1669-7714","observed_orcids":["https://orcid.org/0000-0003-1669-7714"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wahlgren, Jacob","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160526923","display_name":"Victor Barros","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Barros, Victor","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160416067","display_name":"Subhankar Shah","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shah, Subhankar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090492010","display_name":"Michael Kuperstein","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kuperstein, Michael","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045731430","display_name":"Farzin Houshmand","orcid":"https://orcid.org/0000-0001-8516-2401","observed_orcids":["https://orcid.org/0000-0001-8516-2401"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Houshmand, Farzin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160495947","display_name":"Emma Sevastian","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sevastian, Emma","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160457305","display_name":"Peter Gavin","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gavin, Peter","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5110949807","display_name":"Karthik Srinivasa Murthy","orcid":"https://orcid.org/0000-0002-6063-9653","observed_orcids":["https://orcid.org/0000-0002-6063-9653"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Murthy, Karthik Srinivasa","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009468601","display_name":"Berkin Ilbeyi","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ilbeyi, Berkin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5000932612","display_name":"Amit Sabne","orcid":"https://orcid.org/0000-0002-2179-0078","observed_orcids":["https://orcid.org/0000-0002-2179-0078"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sabne, Amit","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5011231117","display_name":"Eli Bendersky","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bendersky, Eli","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":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/T10743","display_name":"Software Testing and Debugging Techniques","score":0.9980000257492065,"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":0.9980000257492065,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.0013000000035390258,"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/T12127","display_name":"Software System Performance and Reliability","score":9.999999747378752e-05,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/x-linked-agammaglobulinemia","display_name":"X-linked agammaglobulinemia","score":0.765999972820282},{"id":"https://openalex.org/keywords/tpu","display_name":"TPU","score":0.6769999861717224},{"id":"https://openalex.org/keywords/compiler-correctness","display_name":"compiler correctness","score":0.8970000147819519},{"id":"https://openalex.org/keywords/fuzzing","display_name":"fuzzing","score":0.7630000114440918},{"id":"https://openalex.org/keywords/differential-testing","display_name":"differential testing","score":0.5960000157356262},{"id":"https://openalex.org/keywords/memory-corruption","display_name":"memory corruption","score":0.6420000195503235},{"id":"https://openalex.org/keywords/large-language-model-training","display_name":"large language model training","score":0.6050000190734863}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8138999938964844},{"id":"https://openalex.org/C169590947","wikidata":"https://www.wikidata.org/wiki/Q47506","display_name":"Compiler","level":2,"score":0.8094000220298767},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.675000011920929},{"id":"https://openalex.org/C111065885","wikidata":"https://www.wikidata.org/wiki/Q1189053","display_name":"Fuzz testing","level":3,"score":0.6593000292778015},{"id":"https://openalex.org/C55439883","wikidata":"https://www.wikidata.org/wiki/Q360812","display_name":"Correctness","level":2,"score":0.5752999782562256},{"id":"https://openalex.org/C190902152","wikidata":"https://www.wikidata.org/wiki/Q1325106","display_name":"Optimizing compiler","level":3,"score":0.4607999920845032},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.4510999917984009},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.41749998927116394},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.38440001010894775},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.37689998745918274},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.37400001287460327},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.32280001044273376},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.3203999996185303},{"id":"https://openalex.org/C40842320","wikidata":"https://www.wikidata.org/wiki/Q19423","display_name":"Buffer overflow","level":2,"score":0.3046000003814697},{"id":"https://openalex.org/C2780589192","wikidata":"https://www.wikidata.org/wiki/Q7285140","display_name":"Raising (metalworking)","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C40305131","wikidata":"https://www.wikidata.org/wiki/Q2616305","display_name":"Obfuscation","level":2,"score":0.2870999872684479},{"id":"https://openalex.org/C2779639559","wikidata":"https://www.wikidata.org/wiki/Q7661178","display_name":"Symbolic execution","level":3,"score":0.2847000062465668},{"id":"https://openalex.org/C2989134064","wikidata":"https://www.wikidata.org/wiki/Q288510","display_name":"Execution time","level":2,"score":0.27570000290870667},{"id":"https://openalex.org/C115903868","wikidata":"https://www.wikidata.org/wiki/Q80993","display_name":"Software engineering","level":1,"score":0.26759999990463257},{"id":"https://openalex.org/C139571649","wikidata":"https://www.wikidata.org/wiki/Q1156793","display_name":"Program optimization","level":3,"score":0.25589999556541443},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.25450000166893005},{"id":"https://openalex.org/C115901376","wikidata":"https://www.wikidata.org/wiki/Q184199","display_name":"Automation","level":2,"score":0.25130000710487366}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2611.00624","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2611.00624","pdf_url":"https://arxiv.org/pdf/2611.00624","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.2611.00624","doi":"https://doi.org/10.48550/arxiv.2611.00624","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2611.00624","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:2611.00624","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2611.00624","pdf_url":"https://arxiv.org/pdf/2611.00624","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":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7225165697.pdf","grobid_xml":"https://content.openalex.org/works/W7225165697.grobid-xml"},"referenced_works_count":20,"referenced_works":["https://openalex.org/W2023035194","https://openalex.org/W2798341898","https://openalex.org/W2954698171","https://openalex.org/W2954903132","https://openalex.org/W2961619211","https://openalex.org/W4221144766","https://openalex.org/W4242126179","https://openalex.org/W4284708930","https://openalex.org/W4313442384","https://openalex.org/W4378465442","https://openalex.org/W4384155658","https://openalex.org/W4388483154","https://openalex.org/W4389158362","https://openalex.org/W4394769544","https://openalex.org/W4403223181","https://openalex.org/W4403223258","https://openalex.org/W4406221679","https://openalex.org/W4411552089","https://openalex.org/W4414988639","https://openalex.org/W7164320489"],"related_works":["https://openalex.org/W7154150252","https://openalex.org/W4403223181","https://openalex.org/W3119322020","https://openalex.org/W7170312332","https://openalex.org/W4221144766","https://openalex.org/W7139989008","https://openalex.org/W7214370008","https://openalex.org/W4318719456","https://openalex.org/W7165134613","https://openalex.org/W7160556403"],"abstract_inverted_index":{"Accelerating":[0],"modern":[1],"foundational":[2],"models":[3],"has":[4,130],"pushed":[5],"domain-specific":[6],"compilers":[7],"toward":[8],"increasingly":[9],"complex,":[10],"multi-tiered":[11],"optimization":[12,17],"pipelines.":[13],"However,":[14],"this":[15,55,107],"heavy":[16],"focus":[18],"introduces":[19],"delicate":[20],"tradeoffs":[21],"between":[22],"peak":[23],"execution":[24],"performance":[25],"and":[26,35,44,64,95],"correctness.":[27],"Subtle":[28],"codegen":[29],"defects":[30],"can":[31],"trigger":[32],"numerical":[33],"degradation":[34],"memory":[36],"or":[37],"buffer":[38],"corruption,":[39],"invalidating":[40],"multi-million-dollar":[41],"training":[42],"runs":[43],"degrading":[45],"live":[46],"serving":[47],"quality":[48],"without":[49],"raising":[50],"obvious":[51],"compiler":[52,68,138],"errors.":[53],"In":[54],"paper,":[56],"we":[57,105,148],"present":[58],"a":[59,109],"battle-tested":[60],"methodology":[61],"for":[62],"verifying":[63],"triaging":[65],"Google's":[66],"XLA:TPU":[67],"at":[69],"hyperscale.":[70],"We":[71],"leverage":[72],"fuzzing":[73],"techniques,":[74],"combining":[75],"synthetic":[76],"program":[77],"generation":[78],"with":[79,108],"real":[80],"production":[81],"traces,":[82],"while":[83],"orchestrating":[84],"AI":[85],"agent":[86],"swarms":[87],"to":[88,118],"autonomously":[89],"generate":[90],"stress":[91],"tests,":[92],"reproduce":[93],"failures,":[94],"draft":[96],"fixes.":[97],"To":[98],"pinpoint":[99],"root":[100],"causes":[101],"in":[102,126],"full":[103],"workloads,":[104],"pair":[106],"hierarchical":[110],"differential":[111],"pipeline":[112],"that":[113],"bisects":[114],"large":[115],"graphs":[116],"down":[117],"individual":[119],"failing":[120],"operations":[121],"inside":[122],"fused":[123],"kernels.":[124],"Deployed":[125],"production,":[127],"our":[128],"framework":[129],"already":[131],"driven":[132],"the":[133],"resolution":[134],"of":[135],"over":[136],"142":[137],"bugs":[139],"(with":[140],"103":[141],"additional":[142],"findings":[143],"under":[144],"review),":[145],"from":[146],"which":[147],"distill":[149],"pragmatic":[150],"lessons":[151],"learned.":[152]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-11-04T00:00:00"}
