{"id":"https://openalex.org/W4409248487","doi":"https://doi.org/10.1109/hpca61900.2025.00112","title":"VQ-LLM: High-performance Code Generation for Vector Quantization Augmented LLM Inference","display_name":"VQ-LLM: High-performance Code Generation for Vector Quantization Augmented LLM Inference","publication_year":2025,"publication_date":"2025-03-01","ids":{"openalex":"https://openalex.org/W4409248487","doi":"https://doi.org/10.1109/hpca61900.2025.00112"},"language":"en","primary_location":{"id":"doi:10.1109/hpca61900.2025.00112","is_oa":false,"landing_page_url":"https://doi.org/10.1109/hpca61900.2025.00112","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Symposium on High Performance Computer Architecture (HPCA)","raw_type":"proceedings-article"},"type":"article","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2503.02236","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100623100","display_name":"Zihan Liu","orcid":"https://orcid.org/0000-0002-0874-0682"},"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":true,"raw_author_name":"Zihan Liu","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113192926","display_name":"X. Luo","orcid":null},"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":"Xinhao Luo","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056134380","display_name":"Junxian Guo","orcid":null},"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":"Junxian Guo","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5082388816","display_name":"Wentao Ni","orcid":"https://orcid.org/0000-0002-0687-5196"},"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":"Wentao Ni","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078261857","display_name":"Yangjie Zhou","orcid":"https://orcid.org/0000-0002-3652-5437"},"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":"Yangjie Zhou","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114626347","display_name":"Yue Guan","orcid":"https://orcid.org/0000-0001-7773-4588"},"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":"Yue Guan","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5081230016","display_name":"Christine C. Guo","orcid":"https://orcid.org/0000-0003-1530-0172"},"institutions":[{"id":"https://openalex.org/I170897317","display_name":"Duke University","ror":"https://ror.org/00py81415","country_code":"US","type":"education","lineage":["https://openalex.org/I170897317"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Cong Guo","raw_affiliation_strings":["Duke University"],"affiliations":[{"raw_affiliation_string":"Duke University","institution_ids":["https://openalex.org/I170897317"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008837660","display_name":"Weihao Cui","orcid":"https://orcid.org/0000-0002-6646-5260"},"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":"Weihao Cui","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100956305","display_name":"Yu Feng","orcid":"https://orcid.org/0009-0009-3524-0681"},"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":"Yu Feng","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039318240","display_name":"Minyi Guo","orcid":"https://orcid.org/0000-0003-0034-2302"},"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":"Minyi Guo","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023684582","display_name":"Yuhao Zhu","orcid":"https://orcid.org/0000-0002-2802-0578"},"institutions":[{"id":"https://openalex.org/I5388228","display_name":"University of Rochester","ror":"https://ror.org/022kthw22","country_code":"US","type":"education","lineage":["https://openalex.org/I5388228"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuhao Zhu","raw_affiliation_strings":["University of Rochester"],"affiliations":[{"raw_affiliation_string":"University of Rochester","institution_ids":["https://openalex.org/I5388228"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077768924","display_name":"Minjia Zhang","orcid":"https://orcid.org/0000-0002-8165-166X"},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Minjia Zhang","raw_affiliation_strings":["University of Illinois Urbana-Champaign"],"affiliations":[{"raw_affiliation_string":"University of Illinois Urbana-Champaign","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Chen Jin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen Jin","raw_affiliation_strings":["Magik Compute"],"affiliations":[{"raw_affiliation_string":"Magik Compute","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5003939279","display_name":"Jingwen Leng","orcid":"https://orcid.org/0000-0002-5660-5493"},"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":"Jingwen Leng","raw_affiliation_strings":["Shanghai Jiao Tong University"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":14,"corresponding_author_ids":["https://openalex.org/A5100623100"],"corresponding_institution_ids":["https://openalex.org/I183067930"],"apc_list":null,"apc_paid":null,"fwci":17.8265,"has_fulltext":false,"cited_by_count":8,"citation_normalized_percentile":{"value":0.98958059,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":98,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"1496","last_page":"1509"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.9232000112533569,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.9232000112533569,"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/vector-quantization","display_name":"Vector quantization","score":0.797134518623352},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5965761542320251},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5580483078956604},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.43807539343833923},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.4302406311035156},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.4135058522224426},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.35313189029693604},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3027346134185791},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.15295526385307312}],"concepts":[{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.797134518623352},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5965761542320251},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5580483078956604},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.43807539343833923},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.4302406311035156},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.4135058522224426},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.35313189029693604},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3027346134185791},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.15295526385307312},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/hpca61900.2025.00112","is_oa":false,"landing_page_url":"https://doi.org/10.1109/hpca61900.2025.00112","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Symposium on High Performance Computer Architecture (HPCA)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2503.02236","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2503.02236","pdf_url":"https://arxiv.org/pdf/2503.02236","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2503.02236","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2503.02236","pdf_url":"https://arxiv.org/pdf/2503.02236","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":[{"score":0.4399999976158142,"id":"https://metadata.un.org/sdg/7","display_name":"Affordable and clean energy"}],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":61,"referenced_works":["https://openalex.org/W1976046733","https://openalex.org/W1977182282","https://openalex.org/W2002555321","https://openalex.org/W2077815765","https://openalex.org/W2124509324","https://openalex.org/W2207050309","https://openalex.org/W2556145167","https://openalex.org/W2962834855","https://openalex.org/W3007718266","https://openalex.org/W3117512411","https://openalex.org/W3129831491","https://openalex.org/W3205706264","https://openalex.org/W4214512541","https://openalex.org/W4234854340","https://openalex.org/W4239893399","https://openalex.org/W4280568654","https://openalex.org/W4280633999","https://openalex.org/W4308083739","https://openalex.org/W4327930469","https://openalex.org/W4360831828","https://openalex.org/W4366341968","https://openalex.org/W4382809751","https://openalex.org/W4385245566","https://openalex.org/W4385585365","https://openalex.org/W4387321091","https://openalex.org/W4388979610","https://openalex.org/W4392182282","https://openalex.org/W4394998892","https://openalex.org/W4395020674","https://openalex.org/W4406650295","https://openalex.org/W4408847395","https://openalex.org/W6676912532","https://openalex.org/W6685526727","https://openalex.org/W6751349269","https://openalex.org/W6767440493","https://openalex.org/W6789646864","https://openalex.org/W6799372109","https://openalex.org/W6810610777","https://openalex.org/W6811340617","https://openalex.org/W6811928498","https://openalex.org/W6838322825","https://openalex.org/W6846164622","https://openalex.org/W6847478871","https://openalex.org/W6849058233","https://openalex.org/W6850625674","https://openalex.org/W6852581948","https://openalex.org/W6854308872","https://openalex.org/W6856696905","https://openalex.org/W6857288518","https://openalex.org/W6859466964","https://openalex.org/W6859846904","https://openalex.org/W6860493679","https://openalex.org/W6860936625","https://openalex.org/W6861393904","https://openalex.org/W6861839547","https://openalex.org/W6866033557","https://openalex.org/W6866496068","https://openalex.org/W6868895731","https://openalex.org/W6870125966","https://openalex.org/W6875634212","https://openalex.org/W6893640197"],"related_works":["https://openalex.org/W1539956819","https://openalex.org/W2093152993","https://openalex.org/W1489554298","https://openalex.org/W2378212145","https://openalex.org/W2055243143","https://openalex.org/W2400327593","https://openalex.org/W1988309121","https://openalex.org/W2100120615","https://openalex.org/W1904724140","https://openalex.org/W3209251257"],"abstract_inverted_index":{"Vector":[0],"quantization":[1,28,267],"(VQ),":[2],"which":[3],"treats":[4],"a":[5,8,42,53,153],"vector":[6,107],"as":[7],"compression":[9],"unit,":[10],"gains":[11],"increasing":[12],"research":[13],"interests":[14],"for":[15,131],"its":[16,81],"potential":[17],"to":[18,25,125,159,228,235,251,257,287],"accelerate":[19],"large":[20],"language":[21],"models":[22],"(LLMs).":[23],"Compared":[24],"conventional":[26],"element-wise":[27,266],"methods,":[29],"VQ":[30,54,74,93,103,146,169,237],"algorithms":[31,94,104],"can":[32],"compress":[33],"weight":[34],"and":[35,67,78,95,109,112,119,140,164,192,220,271],"KV":[36],"cache":[37,158,175],"tensors":[38],"in":[39,85,92,232],"LLMs":[40],"with":[41,170,264],"greater":[43,294],"ratio":[44],"while":[45],"maintaining":[46],"the":[47,59,69,86,100,166,181,196,217,242],"high":[48],"model":[49],"accuracy.":[50,295],"However,":[51],"translating":[52],"algorithm\u2019s":[55],"memory":[56,183,207],"reduction":[57,244],"into":[58,75],"actual":[60],"latency":[61,243],"improvement":[62],"is":[63,277],"challenging.":[64],"We":[65,150],"profile":[66],"analyze":[68],"current":[70],"approach":[71],"of":[72,90,102,168,245],"integrating":[73],"computation":[76,97,114,203],"kernels":[77,115],"show":[79],"that":[80,205,274],"major":[82],"inefficiency":[83],"lies":[84],"poor":[87],"access":[88,162],"efficiency":[89,163],"codebooks":[91],"uncoordinated":[96],"dataflow.":[98],"Meanwhile,":[99],"diversity":[101],"(e.g.,":[105],"different":[106,178],"sizes":[108],"entry":[110],"counts)":[111],"LLMs,":[113],"(e.g":[116],"matrix-matrix/vector":[117],"multiplication":[118],"attention":[120],"computation)":[121],"makes":[122],"it":[123],"impractical":[124],"manually":[126],"craft":[127],"efficient":[128,144,202],"kernel":[129,147],"implementations":[130],"each":[132],"specific":[133],"case.":[134],"In":[135],"this":[136],"work,":[137],"we":[138,199,224],"design":[139,200],"implement":[141],"VQ-LLM,":[142],"an":[143,201],"fused":[145],"generation":[148],"framework.":[149],"first":[151],"introduce":[152],"software":[154],"abstraction":[155],"called":[156],"codebook":[157,161,174,197],"optimize":[160],"support":[165],"integration":[167],"various":[171],"computations.":[172],"The":[173],"adaptively":[176],"stores":[177],"entries":[179],"across":[180],"GPU\u2019s":[182],"hierarchy,":[184],"including":[185],"off-chip":[186],"global":[187],"memory,":[188,191],"on-chip":[189],"shared":[190],"registers.":[193],"Centered":[194],"around":[195],"cache,":[198],"engine":[204,215],"optimizes":[206],"traffic":[208],"during":[209],"computations":[210],"involving":[211],"codebooks.":[212],"This":[213],"compute":[214],"adopts":[216],"codebook-centric":[218],"dataflow":[219],"fusion":[221],"optimizations.":[222],"Additionally,":[223],"provide":[225],"adaptive":[226],"heuristics":[227],"tailor":[229],"parameter":[230],"selection":[231],"our":[233,275],"optimizations":[234,240],"diverse":[236],"configurations.":[238],"Our":[239],"achieve":[241],"$\\mathbf{6":[246],"4.":[247],"3":[248],"6":[249],"\\%}$":[250,255],"$\\mathbf{9":[252],"9.":[253],"1":[254],"compared":[256],"existing":[258],"open-source":[259],"implementations.":[260],"A":[261],"final":[262],"comparison":[263],"state-of-the-art":[265],"methods":[268],"like":[269],"AWQ":[270],"QoQ":[272],"shows":[273],"VQ-LLM":[276],"practically":[278],"viable,":[279],"achieving":[280],"latencies":[281,286],"close":[282],"or":[283],"even":[284],"better":[285],"those":[288],"at":[289],"equivalent":[290],"bit-widths,":[291],"potentially":[292],"offering":[293]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":6}],"updated_date":"2026-04-21T08:09:41.155169","created_date":"2025-10-10T00:00:00"}