{"id":"https://openalex.org/W4413320520","doi":"https://doi.org/10.1109/iolts65288.2025.11116834","title":"When Silicon Fails Silently: Characterizing Hardware-Induced Corruption in LLM Training","display_name":"When Silicon Fails Silently: Characterizing Hardware-Induced Corruption in LLM Training","publication_year":2025,"publication_date":"2025-07-07","ids":{"openalex":"https://openalex.org/W4413320520","doi":"https://doi.org/10.1109/iolts65288.2025.11116834"},"language":"en","primary_location":{"id":"doi:10.1109/iolts65288.2025.11116834","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iolts65288.2025.11116834","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 31st International Symposium on On-Line Testing and Robust System Design (IOLTS)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5090886418","display_name":"Jeffrey Ma","orcid":"https://orcid.org/0000-0002-3646-3547"},"institutions":[{"id":"https://openalex.org/I2801851002","display_name":"Harvard University Press","ror":"https://ror.org/006v7bf86","country_code":"US","type":"other","lineage":["https://openalex.org/I136199984","https://openalex.org/I2801851002"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Jeffrey Ma","raw_affiliation_strings":["Harvard University,Department of Computer Science,Cambridge,MA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Harvard University,Department of Computer Science,Cambridge,MA,USA","institution_ids":["https://openalex.org/I2801851002"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036100753","display_name":"Hengzhi Pei","orcid":"https://orcid.org/0000-0001-7036-2996"},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Hengzhi Pei","raw_affiliation_strings":["Amazon Web Services,Santa Clara,CA,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Web Services,Santa Clara,CA,USA","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066453165","display_name":"Leonard Lausen","orcid":null},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Leonard Lausen","raw_affiliation_strings":["Amazon Web Services,New York,NY,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Web Services,New York,NY,USA","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5082384108","display_name":"George Karypis","orcid":"https://orcid.org/0000-0003-2753-1437"},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"George Karypis","raw_affiliation_strings":["Amazon Web Services,Minneapolis,MN,USA"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon Web Services,Minneapolis,MN,USA","institution_ids":["https://openalex.org/I1311688040"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":3.4568,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.89796392,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"4"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10558","display_name":"Advancements in Semiconductor Devices and Circuit Design","score":0.9908999800682068,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10558","display_name":"Advancements in Semiconductor Devices and Circuit Design","score":0.9908999800682068,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11032","display_name":"VLSI and Analog Circuit Testing","score":0.9279000163078308,"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/computer-science","display_name":"Computer science","score":0.5751602053642273},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.56130450963974},{"id":"https://openalex.org/keywords/language-change","display_name":"Language change","score":0.47917166352272034},{"id":"https://openalex.org/keywords/silicon","display_name":"Silicon","score":0.4636823534965515},{"id":"https://openalex.org/keywords/computer-hardware","display_name":"Computer hardware","score":0.34569889307022095},{"id":"https://openalex.org/keywords/embedded-system","display_name":"Embedded system","score":0.3424263596534729},{"id":"https://openalex.org/keywords/optoelectronics","display_name":"Optoelectronics","score":0.1375105082988739},{"id":"https://openalex.org/keywords/materials-science","display_name":"Materials science","score":0.11246132850646973}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5751602053642273},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.56130450963974},{"id":"https://openalex.org/C2780027415","wikidata":"https://www.wikidata.org/wiki/Q524648","display_name":"Language change","level":2,"score":0.47917166352272034},{"id":"https://openalex.org/C544956773","wikidata":"https://www.wikidata.org/wiki/Q670","display_name":"Silicon","level":2,"score":0.4636823534965515},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.34569889307022095},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.3424263596534729},{"id":"https://openalex.org/C49040817","wikidata":"https://www.wikidata.org/wiki/Q193091","display_name":"Optoelectronics","level":1,"score":0.1375105082988739},{"id":"https://openalex.org/C192562407","wikidata":"https://www.wikidata.org/wiki/Q228736","display_name":"Materials science","level":0,"score":0.11246132850646973},{"id":"https://openalex.org/C142362112","wikidata":"https://www.wikidata.org/wiki/Q735","display_name":"Art","level":0,"score":0.0},{"id":"https://openalex.org/C124952713","wikidata":"https://www.wikidata.org/wiki/Q8242","display_name":"Literature","level":1,"score":0.0},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/iolts65288.2025.11116834","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iolts65288.2025.11116834","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 31st International Symposium on On-Line Testing and Robust System Design (IOLTS)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions","score":0.7699999809265137}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":8,"referenced_works":["https://openalex.org/W2767260595","https://openalex.org/W2964108906","https://openalex.org/W4280619532","https://openalex.org/W4312121086","https://openalex.org/W4384026302","https://openalex.org/W4387321084","https://openalex.org/W4388624644","https://openalex.org/W4412944830"],"related_works":["https://openalex.org/W230091440","https://openalex.org/W2233261550","https://openalex.org/W2810751659","https://openalex.org/W258997015","https://openalex.org/W2997094352","https://openalex.org/W4412817058","https://openalex.org/W3216976533","https://openalex.org/W100620283","https://openalex.org/W2495260952","https://openalex.org/W4366179611"],"abstract_inverted_index":{"As":[0],"the":[1,32,36,67,76,104,133,148,177],"scale":[2],"of":[3,35,38,106,135],"training":[4,43,56,127,178],"large":[5],"language":[6],"models":[7,163],"(LLMs)":[8],"increases,":[9],"one":[10],"emergent":[11],"failure":[12,25],"is":[13],"silent":[14],"data":[15],"corruption":[16],"(SDC),":[17],"where":[18],"hardware":[19],"produces":[20],"incorrect":[21],"computations":[22],"without":[23],"explicit":[24],"signals.":[26],"In":[27,50],"this":[28],"work,":[29],"we":[30,53,74,100],"summarize":[31],"first":[33],"investigation":[34],"impact":[37,105,134],"real-world":[39],"SDCs":[40,136,151,160],"on":[41,109,137,140,152],"LLM":[42],"presented":[44],"in":[45,145,176],"our":[46,51,96],"ACL":[47],"work":[48],"[1].":[49],"investigation,":[52],"compare":[54],"model":[55],"between":[57],"healthy":[58],"production":[59,84],"nodes":[60,63,78,111],"and":[61,95,102,124,155,172],"unhealthy":[62,77,142],"exhibiting":[64],"SDCs.":[65],"With":[66],"help":[68],"from":[69,83,150],"a":[70,120,126],"cloud":[71],"computing":[72],"platform,":[73],"access":[75],"that":[79,132],"were":[80],"swept":[81],"out":[82],"by":[85],"automated":[86],"fleet":[87],"management.":[88],"Using":[89],"deterministic":[90],"execution":[91],"via":[92],"XLA":[93],"compiler":[94],"proposed":[97],"synchronization":[98],"mechanisms,":[99],"isolate":[101],"analyze":[103],"SDC":[107],"errors":[108],"these":[110],"at":[112,115,119,125],"three":[113],"levels:":[114],"each":[116],"submodule":[117,153],"computation,":[118],"single":[121],"optimizer":[122],"step,":[123],"period.":[128],"Our":[129],"results":[130],"reveal":[131],"computation":[138,154],"varies":[139],"different":[141,167,170],"nodes.":[143],"Although":[144],"most":[146],"cases":[147],"perturbations":[149],"gradients":[156],"are":[157],"relatively":[158],"small,":[159],"can":[161],"lead":[162],"to":[164,166],"converge":[165],"optima":[168],"with":[169],"weights":[171],"even":[173],"cause":[174],"spikes":[175],"loss.":[179]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-08-01T09:00:35.917206","created_date":"2025-10-10T00:00:00"}