{"id":"https://openalex.org/W4414198863","doi":"https://doi.org/10.1109/dac63849.2025.11132542","title":"KVO-LLM: Boosting Long-Context Generation Throughput for Batched LLM Inference","display_name":"KVO-LLM: Boosting Long-Context Generation Throughput for Batched LLM Inference","publication_year":2025,"publication_date":"2025-06-22","ids":{"openalex":"https://openalex.org/W4414198863","doi":"https://doi.org/10.1109/dac63849.2025.11132542"},"language":"en","primary_location":{"id":"doi:10.1109/dac63849.2025.11132542","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11132542","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/A5100330770","display_name":"Zhenyu Li","orcid":"https://orcid.org/0000-0002-9959-1124"},"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":"Zhenyu Li","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066692868","display_name":"Dongxu Lyu","orcid":"https://orcid.org/0000-0001-6826-2670"},"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":"Dongxu Lyu","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100618174","display_name":"Gang Wang","orcid":"https://orcid.org/0000-0002-3099-5913"},"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":"Gang Wang","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101463703","display_name":"Yuzhou Chen","orcid":"https://orcid.org/0000-0003-0227-3152"},"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":"Yuzhou Chen","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5073749486","display_name":"Liyan Chen","orcid":"https://orcid.org/0000-0001-9693-5377"},"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":"Liyan Chen","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100408982","display_name":"Wenjie Li","orcid":"https://orcid.org/0000-0002-5214-0449"},"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":"Wenjie Li","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008057183","display_name":"Jianfei Jiang","orcid":"https://orcid.org/0000-0002-5521-6197"},"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":"Jianfei Jiang","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102022581","display_name":"Yanan Sun","orcid":"https://orcid.org/0000-0001-8281-9121"},"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":"Yanan Sun","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5054897331","display_name":"Guanghui He","orcid":"https://orcid.org/0000-0002-0486-6421"},"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":"Guanghui He","raw_affiliation_strings":["Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University,State Key Laboratory of Micro/Nano Engineering Science, School of Electronic Information and Electrical Engineering,Shanghai,China","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":9,"corresponding_author_ids":["https://openalex.org/A5100330770"],"corresponding_institution_ids":["https://openalex.org/I183067930"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.12631249,"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/T10028","display_name":"Topic Modeling","score":0.9146000146865845,"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/T10028","display_name":"Topic Modeling","score":0.9146000146865845,"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/cache","display_name":"Cache","score":0.7814000248908997},{"id":"https://openalex.org/keywords/cache-algorithms","display_name":"Cache algorithms","score":0.46320000290870667},{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.42750000953674316},{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.41839998960494995},{"id":"https://openalex.org/keywords/memory-bandwidth","display_name":"Memory bandwidth","score":0.41110000014305115},{"id":"https://openalex.org/keywords/interleaving","display_name":"Interleaving","score":0.38989999890327454},{"id":"https://openalex.org/keywords/efficient-energy-use","display_name":"Efficient energy use","score":0.3799999952316284},{"id":"https://openalex.org/keywords/smart-cache","display_name":"Smart Cache","score":0.3720000088214874},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.33899998664855957}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8608999848365784},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.7814000248908997},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4997999966144562},{"id":"https://openalex.org/C38556500","wikidata":"https://www.wikidata.org/wiki/Q13404475","display_name":"Cache algorithms","level":4,"score":0.46320000290870667},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.42750000953674316},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.41839998960494995},{"id":"https://openalex.org/C188045654","wikidata":"https://www.wikidata.org/wiki/Q17148339","display_name":"Memory bandwidth","level":2,"score":0.41110000014305115},{"id":"https://openalex.org/C28034677","wikidata":"https://www.wikidata.org/wiki/Q17092530","display_name":"Interleaving","level":2,"score":0.38989999890327454},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.3799999952316284},{"id":"https://openalex.org/C167713795","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"Smart Cache","level":5,"score":0.3720000088214874},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.37119999527931213},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.33899998664855957},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.33730000257492065},{"id":"https://openalex.org/C59687516","wikidata":"https://www.wikidata.org/wiki/Q5015938","display_name":"Cache-oblivious algorithm","level":5,"score":0.32899999618530273},{"id":"https://openalex.org/C28362024","wikidata":"https://www.wikidata.org/wiki/Q2067413","display_name":"Thrashing","level":2,"score":0.32710000872612},{"id":"https://openalex.org/C7366592","wikidata":"https://www.wikidata.org/wiki/Q1255620","display_name":"Dram","level":2,"score":0.32030001282691956},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.31470000743865967},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.31290000677108765},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.30320000648498535},{"id":"https://openalex.org/C118524514","wikidata":"https://www.wikidata.org/wiki/Q173212","display_name":"Computer architecture","level":1,"score":0.3012999892234802},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.29170000553131104},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.28200000524520874},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.2775999903678894},{"id":"https://openalex.org/C2776257435","wikidata":"https://www.wikidata.org/wiki/Q1576430","display_name":"Bandwidth (computing)","level":2,"score":0.2678000032901764},{"id":"https://openalex.org/C101468663","wikidata":"https://www.wikidata.org/wiki/Q1620158","display_name":"Modular design","level":2,"score":0.26420000195503235},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.2632000148296356},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.2547999918460846},{"id":"https://openalex.org/C25536678","wikidata":"https://www.wikidata.org/wiki/Q5015977","display_name":"Cache invalidation","level":5,"score":0.2531000077724457},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.25040000677108765}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/dac63849.2025.11132542","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11132542","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/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":8,"referenced_works":["https://openalex.org/W2105102111","https://openalex.org/W4392427708","https://openalex.org/W4394998968","https://openalex.org/W4401211590","https://openalex.org/W4401211806","https://openalex.org/W4401211807","https://openalex.org/W4402671766","https://openalex.org/W4404133534"],"related_works":[],"abstract_inverted_index":{"With":[0],"the":[1,34,42,63,102,145,164,190],"widespread":[2],"deployment":[3],"of":[4,67,147],"long-context":[5,97],"large":[6],"language":[7],"models":[8],"(LLMs),":[9],"efficient":[10],"and":[11,22,30,37,118,123,178,184,203],"high-quality":[12],"generation":[13,28,196],"is":[14,94],"becoming":[15],"increasingly":[16],"important.":[17],"Modern":[18],"LLMs":[19],"employ":[20],"batching":[21],"key-value":[23],"(KV)":[24],"cache":[25,44,75,109,149,160],"to":[26,62,157,189,200],"improve":[27],"throughput":[29,197],"quality.":[31],"However,":[32],"as":[33],"context":[35],"length":[36],"batch":[38],"size":[39],"rise":[40],"drastically,":[41],"KV":[43,70,74,108,121,148,159],"incurs":[45],"extreme":[46],"external":[47],"memory":[48],"access":[49],"(EMA)":[50],"issues.":[51],"Recent":[52],"LLM":[53,99,192],"accelerators":[54],"face":[55],"substantial":[56,137],"processing":[57],"element":[58],"(PE)":[59],"under-utilization":[60],"due":[61],"low":[64],"arithmetic":[65],"intensity":[66],"attention":[68,126],"with":[69,79,152],"cache,":[71],"while":[72],"existing":[73],"compression":[76,161],"algorithms":[77],"struggle":[78],"hardware":[80,140],"inefficiency":[81],"or":[82],"significant":[83,153],"accuracy":[84,154],"degradation.":[85],"To":[86],"address":[87],"these":[88],"issues,":[89],"an":[90],"algorithm-architecture":[91],"co-optimization,":[92],"KVO-LLM,":[93],"proposed":[95],"for":[96],"batched":[98],"generation.":[100],"At":[101,163],"algorithm":[103,143],"level,":[104,166],"we":[105,167],"propose":[106,168],"a":[107,169],"quantization-aware":[110],"pruning":[111,128],"method":[112],"that":[113,174],"first":[114],"adopts":[115,175],"salient-token-aware":[116],"quantization":[117],"then":[119],"prunes":[120],"channels":[122],"tokens":[124,132],"by":[125,198],"guided":[127],"based":[129],"on":[130,139],"salient":[131],"identified":[133],"during":[134],"quantization.":[135],"Achieving":[136],"savings":[138],"overhead,":[141],"our":[142],"reduces":[144],"EMA":[146],"over":[150],"91%":[151],"advantages":[155],"compared":[156],"previous":[158],"algorithms.":[162],"architecture":[165],"multi-core":[170],"jointly":[171],"optimized":[172],"accelerator":[173],"operator":[176],"fusion":[177],"cross-batch":[179],"interleaving":[180],"strategy,":[181],"maximizing":[182],"PE":[183],"DRAM":[185],"bandwidth":[186],"utilization.":[187],"Compared":[188],"state-of-the-art":[191],"accelerators,":[193],"KVO-LLM":[194],"improves":[195],"up":[199],"$7.32":[201],"\\times$,":[202],"attains":[204],"$5.52":[205],"\\sim":[206],"8.38":[207],"\\times$":[208],"better":[209],"energy":[210],"efficiency.":[211]},"counts_by_year":[],"updated_date":"2026-04-09T08:11:56.329763","created_date":"2025-10-10T00:00:00"}