{"id":"https://openalex.org/W4414199202","doi":"https://doi.org/10.1109/dac63849.2025.11132862","title":"MILLION: MasterIng Long-Context LLM Inference Via Outlier-Immunized KV Product QuaNtization","display_name":"MILLION: MasterIng Long-Context LLM Inference Via Outlier-Immunized KV Product QuaNtization","publication_year":2025,"publication_date":"2025-06-22","ids":{"openalex":"https://openalex.org/W4414199202","doi":"https://doi.org/10.1109/dac63849.2025.11132862"},"language":"en","primary_location":{"id":"doi:10.1109/dac63849.2025.11132862","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11132862","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 62nd ACM/IEEE Design Automation Conference (DAC)","raw_type":"proceedings-article"},"type":"article","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/A5107269495","display_name":"Zongwu Wang","orcid":"https://orcid.org/0009-0003-2157-4927"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zongwu Wang","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018569941","display_name":"Peng Xu","orcid":"https://orcid.org/0000-0003-0822-700X"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Peng Xu","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017670541","display_name":"Fangxin Liu","orcid":"https://orcid.org/0000-0002-8769-293X"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Fangxin Liu","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101443061","display_name":"Yiwei Hu","orcid":"https://orcid.org/0000-0002-6713-4148"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yiwei Hu","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060999547","display_name":"Qingxiao Sun","orcid":"https://orcid.org/0000-0003-2927-362X"},"institutions":[{"id":"https://openalex.org/I204553293","display_name":"China University of Petroleum, Beijing","ror":"https://ror.org/041qf4r12","country_code":"CN","type":"education","lineage":["https://openalex.org/I204553293"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Qingxiao Sun","raw_affiliation_strings":["China University of Petroleum-Beijing"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"China University of Petroleum-Beijing","institution_ids":["https://openalex.org/I204553293"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009211176","display_name":"Gezi Li","orcid":"https://orcid.org/0000-0002-4101-1043"},"institutions":[{"id":"https://openalex.org/I4210160618","display_name":"Huawei Technologies (United Kingdom)","ror":"https://ror.org/056gzgs71","country_code":"GB","type":"company","lineage":["https://openalex.org/I2250955327","https://openalex.org/I4210160618"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Gezi Li","raw_affiliation_strings":["Huawei Technologies Co., Ltd"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Technologies Co., Ltd","institution_ids":["https://openalex.org/I4210160618"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5106084673","display_name":"Cheng Li","orcid":"https://orcid.org/0000-0002-8123-2190"},"institutions":[{"id":"https://openalex.org/I4210160618","display_name":"Huawei Technologies (United Kingdom)","ror":"https://ror.org/056gzgs71","country_code":"GB","type":"company","lineage":["https://openalex.org/I2250955327","https://openalex.org/I4210160618"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Cheng Li","raw_affiliation_strings":["Huawei Technologies Co., Ltd"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Technologies Co., Ltd","institution_ids":["https://openalex.org/I4210160618"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100328984","display_name":"Xuan Wang","orcid":"https://orcid.org/0000-0002-0790-5641"},"institutions":[{"id":"https://openalex.org/I4210160618","display_name":"Huawei Technologies (United Kingdom)","ror":"https://ror.org/056gzgs71","country_code":"GB","type":"company","lineage":["https://openalex.org/I2250955327","https://openalex.org/I4210160618"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Xuan Wang","raw_affiliation_strings":["Huawei Technologies Co., Ltd"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Technologies Co., Ltd","institution_ids":["https://openalex.org/I4210160618"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5071012979","display_name":"Li Jiang","orcid":"https://orcid.org/0000-0003-2724-0605"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Li Jiang","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5049487451","display_name":"Haibing Guan","orcid":"https://orcid.org/0000-0002-4714-7400"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Haibing Guan","raw_affiliation_strings":["Shanghai Jiao Tong University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":10,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.11456743,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"7"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.96670001745224,"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"}},"topics":[{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.96670001745224,"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/inference","display_name":"Inference","score":0.6762999892234802},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.6668000221252441},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.5282999873161316},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.3799000084400177},{"id":"https://openalex.org/keywords/cpu-cache","display_name":"CPU cache","score":0.34290000796318054},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.32580000162124634},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.3188999891281128}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7735000252723694},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6762999892234802},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.6668000221252441},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.5282999873161316},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.43380001187324524},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.42250001430511475},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.3799000084400177},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.34290000796318054},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.33309999108314514},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.32580000162124634},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3188999891281128},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.31119999289512634},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.298799991607666},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2831000089645386},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.28200000524520874},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.27790001034736633},{"id":"https://openalex.org/C2781357197","wikidata":"https://www.wikidata.org/wiki/Q5757597","display_name":"High memory","level":2,"score":0.27549999952316284},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.27469998598098755},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.27219998836517334},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.26429998874664307},{"id":"https://openalex.org/C174348530","wikidata":"https://www.wikidata.org/wiki/Q188635","display_name":"Bridging (networking)","level":2,"score":0.2540000081062317}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/dac63849.2025.11132862","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11132862","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 62nd ACM/IEEE Design Automation Conference (DAC)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320309612","display_name":"Natural Science Foundation of Shanghai","ror":null},{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320335777","display_name":"National Key Research and Development Program of China","ror":null}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":6,"referenced_works":["https://openalex.org/W2124509324","https://openalex.org/W3189877953","https://openalex.org/W4389518760","https://openalex.org/W4389524555","https://openalex.org/W4402670433","https://openalex.org/W4402671766"],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,16],"(LLMs)":[3],"are":[4],"increasingly":[5],"utilized":[6],"for":[7,110,205],"complex":[8],"tasks":[9],"requiring":[10],"longer":[11],"context":[12,78,245],"lengths,":[13],"with":[14,77,198,229],"some":[15],"supporting":[17],"up":[18],"to":[19,90,114],"128":[20],"K":[21,244],"or":[22],"1":[23],"M":[24],"tokens.":[25],"This":[26],"trend,":[27],"however,":[28],"presents":[29],"significant":[30,124],"challenges":[31],"in":[32,41,131],"inference":[33,44,196,216],"speed":[34],"and":[35,98,121,202,211,232,235],"memory":[36,72,99],"management.":[37],"The":[38],"primary":[39],"bottleneck":[40],"long-context":[42],"LLM":[43,82,96],"is":[45,248],"the":[46,92,166],"quadratic":[47],"computational":[48],"complexity":[49],"of":[50,129,161,168],"attention":[51,200],"mechanisms,":[52],"causing":[53,123],"substantial":[54],"slowdowns":[55],"as":[56,86],"sequence":[57],"length":[58],"increases.":[59],"KV":[60,111,132,150,162],"cache":[61,151,163],"mechanism":[62],"alleviates":[63],"this":[64,139],"issue":[65],"by":[66],"storing":[67],"pre-computed":[68],"data,":[69],"but":[70],"introduces":[71],"requirements":[73],"that":[74,207,222],"scale":[75],"linearly":[76],"length,":[79],"hindering":[80],"efficient":[81,199],"deployment.":[83],"Quantization":[84],"emerges":[85],"a":[87,144,158,175,193],"promising":[88],"approach":[89],"address":[91],"widening":[93],"gap":[94],"between":[95],"size":[97],"capacity.":[100],"However,":[101],"traditional":[102],"quantization":[103,120,146,170,177,228],"schemes":[104],"often":[105],"yield":[106],"suboptimal":[107],"compression":[108],"results":[109,220],"caches":[112],"due":[113],"two":[115],"key":[116],"factors:":[117],"i)":[118],"On-the-fly":[119],"de-quantization,":[122],"performance":[125,240],"overhead;":[126],"ii)":[127],"Prevalence":[128],"outliers":[130],"values,":[133],"challenging":[134],"low-bitwidth":[135,149],"uniform":[136],"quantization.":[137,154],"To":[138],"end,":[140],"we":[141,156,173,191],"propose":[142],"MILLION,":[143],"novel":[145],"framework":[147,197],"achieving":[148],"through":[152],"product":[153,181],"First,":[155],"conduct":[157],"thorough":[159],"analysis":[160],"distribution,":[164],"revealing":[165],"limitations":[167],"existing":[169],"schemes.":[171],"Second,":[172],"introduce":[174],"non-uniform":[176],"algorithm":[178],"based":[179],"on":[180],"quantization,":[182,213],"which":[183],"efficiently":[184],"compresses":[185],"data":[186],"while":[187],"preserving":[188],"accuracy.":[189],"Third,":[190],"develop":[192],"high-performance":[194],"GPU":[195],"kernel":[201],"pipeline":[203],"design":[204],"MILLION":[206,223],"leverages":[208],"sparse":[209],"computation":[210],"asynchronous":[212],"significantly":[214],"enhancing":[215],"speed.":[217],"Comprehensive":[218],"evaluation":[219],"demonstrate":[221],"can":[224],"achieve":[225,236],"4":[226],"bits":[227],"trivial":[230],"perplexity":[231],"accuracy":[233],"loss,":[234],"2.09":[237],"x":[238],"end-to-end":[239],"gains":[241],"at":[242,250],"32":[243],"length.":[246],"Code":[247],"released":[249],"https://github.com/ZongwuWang/MILLION.":[251]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}