{"id":"https://openalex.org/W4415594528","doi":"https://doi.org/10.1109/tc.2025.3626150","title":"MATA: A Memory-Efficient Attention Accelerator for LLMs Exploiting Look-Back KV Cache Pruning","display_name":"MATA: A Memory-Efficient Attention Accelerator for LLMs Exploiting Look-Back KV Cache Pruning","publication_year":2025,"publication_date":"2025-10-27","ids":{"openalex":"https://openalex.org/W4415594528","doi":"https://doi.org/10.1109/tc.2025.3626150"},"language":null,"primary_location":{"id":"doi:10.1109/tc.2025.3626150","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2025.3626150","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","0016-9340","1557-9956","2326-3814"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Computers","raw_type":"journal-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/A5101196275","display_name":"Zeyu Zhu","orcid":"https://orcid.org/0009-0001-5405-0295"},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zeyu Zhu","raw_affiliation_strings":["Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0003-0709-2438","affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]},{"raw_affiliation_string":"C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100438627","display_name":"Gang Li","orcid":"https://orcid.org/0000-0001-7101-9532"},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Gang Li","raw_affiliation_strings":["Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-7835-4739","affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]},{"raw_affiliation_string":"C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001331259","display_name":"Tielong Liu","orcid":"https://orcid.org/0009-0001-5500-4338"},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tielong Liu","raw_affiliation_strings":["Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0006-9675-4465","affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]},{"raw_affiliation_string":"C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5012344182","display_name":"Zitao Mo","orcid":"https://orcid.org/0000-0001-8623-5465"},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zitao Mo","raw_affiliation_strings":["Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0001-8623-5465","affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]},{"raw_affiliation_string":"C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5056632010","display_name":"Xiaoyao Liang","orcid":"https://orcid.org/0000-0002-2790-5884"},"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":"Xiaoyao Liang","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0002-2790-5884","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5112149353","display_name":"Jian Cheng","orcid":"https://orcid.org/0000-0003-1289-2758"},"institutions":[{"id":"https://openalex.org/I19820366","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35","country_code":"CN","type":"government","lineage":["https://openalex.org/I19820366"]},{"id":"https://openalex.org/I4210094879","display_name":"Shandong Institute of Automation","ror":"https://ror.org/00qdtba35","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4210094879","https://openalex.org/I4210142748"]},{"id":"https://openalex.org/I4210112150","display_name":"Institute of Automation","ror":"https://ror.org/022c3hy66","country_code":"CN","type":"facility","lineage":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jian Cheng","raw_affiliation_strings":["Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China"],"raw_orcid":"https://orcid.org/0000-0003-1289-2758","affiliations":[{"raw_affiliation_string":"Chinese Academy of Sciences, C<sup>2</sup>DL, Institute of Automation, Beijing, China","institution_ids":["https://openalex.org/I4210094879","https://openalex.org/I4210112150"]},{"raw_affiliation_string":"C2DL, Institute of Automation, Chinese Academy of Sciences, Beijing, China","institution_ids":["https://openalex.org/I19820366","https://openalex.org/I4210112150"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.6688,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.75909299,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":"75","issue":"8","first_page":"2660","last_page":"2672"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11181","display_name":"Advanced Data Storage Technologies","score":0.9901000261306763,"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"}},"topics":[{"id":"https://openalex.org/T11181","display_name":"Advanced Data Storage Technologies","score":0.9901000261306763,"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"}},{"id":"https://openalex.org/T11478","display_name":"Caching and Content Delivery","score":0.953499972820282,"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"}},{"id":"https://openalex.org/T12702","display_name":"Brain Tumor Detection and Classification","score":0.9495999813079834,"subfield":{"id":"https://openalex.org/subfields/2808","display_name":"Neurology"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.6273000240325928},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.4787999987602234},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4122999906539917},{"id":"https://openalex.org/keywords/dram","display_name":"Dram","score":0.4065000116825104},{"id":"https://openalex.org/keywords/compiler","display_name":"Compiler","score":0.3970000147819519},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.35989999771118164},{"id":"https://openalex.org/keywords/sorting","display_name":"Sorting","score":0.35040000081062317},{"id":"https://openalex.org/keywords/cpu-cache","display_name":"CPU cache","score":0.3495999872684479}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7996000051498413},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.6273000240325928},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.5321000218391418},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.4787999987602234},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4122999906539917},{"id":"https://openalex.org/C7366592","wikidata":"https://www.wikidata.org/wiki/Q1255620","display_name":"Dram","level":2,"score":0.4065000116825104},{"id":"https://openalex.org/C169590947","wikidata":"https://www.wikidata.org/wiki/Q47506","display_name":"Compiler","level":2,"score":0.3970000147819519},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.35989999771118164},{"id":"https://openalex.org/C111696304","wikidata":"https://www.wikidata.org/wiki/Q2303697","display_name":"Sorting","level":2,"score":0.35040000081062317},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.3495999872684479},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.34860000014305115},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.33880001306533813},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.336899995803833},{"id":"https://openalex.org/C132943942","wikidata":"https://www.wikidata.org/wiki/Q2562511","display_name":"Footprint","level":2,"score":0.33629998564720154},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.3224000036716461},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.3109000027179718},{"id":"https://openalex.org/C187691185","wikidata":"https://www.wikidata.org/wiki/Q2020720","display_name":"Grid","level":2,"score":0.3057999908924103},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.30250000953674316},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.2888999879360199},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.28200000524520874},{"id":"https://openalex.org/C74366991","wikidata":"https://www.wikidata.org/wiki/Q2755335","display_name":"Network processor","level":3,"score":0.26750001311302185},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.2615000009536743},{"id":"https://openalex.org/C108094655","wikidata":"https://www.wikidata.org/wiki/Q181593","display_name":"Sorting algorithm","level":3,"score":0.2563999891281128},{"id":"https://openalex.org/C134835016","wikidata":"https://www.wikidata.org/wiki/Q690265","display_name":"Lookup table","level":2,"score":0.2533999979496002},{"id":"https://openalex.org/C2776650193","wikidata":"https://www.wikidata.org/wiki/Q264661","display_name":"Obstacle","level":2,"score":0.25270000100135803},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2515999972820282}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/tc.2025.3626150","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tc.2025.3626150","pdf_url":null,"source":{"id":"https://openalex.org/S157670870","display_name":"IEEE Transactions on Computers","issn_l":"0018-9340","issn":["0018-9340","0016-9340","1557-9956","2326-3814"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Computers","raw_type":"journal-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G3352541487","display_name":null,"funder_award_id":"BK20243051","funder_id":"https://openalex.org/F4320322769","funder_display_name":"Natural Science Foundation of Jiangsu Province"}],"funders":[{"id":"https://openalex.org/F4320322769","display_name":"Natural Science Foundation of Jiangsu Province","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":26,"referenced_works":["https://openalex.org/W2034861439","https://openalex.org/W2725159389","https://openalex.org/W2807925339","https://openalex.org/W2964110616","https://openalex.org/W3017024317","https://openalex.org/W3100985894","https://openalex.org/W3159727696","https://openalex.org/W3189877953","https://openalex.org/W3206453033","https://openalex.org/W3211525823","https://openalex.org/W4214686755","https://openalex.org/W4281660701","https://openalex.org/W4285676448","https://openalex.org/W4294974598","https://openalex.org/W4308083513","https://openalex.org/W4308083526","https://openalex.org/W4318541578","https://openalex.org/W4366341968","https://openalex.org/W4380881077","https://openalex.org/W4384561707","https://openalex.org/W4389159031","https://openalex.org/W4391494845","https://openalex.org/W4402684050","https://openalex.org/W7133224090","https://openalex.org/W7133227460","https://openalex.org/W7133244816"],"related_works":[],"abstract_inverted_index":{"Transformer-based":[0],"Large":[1],"Language":[2],"Models":[3],"(LLMs)":[4],"have":[5,31],"sparked":[6],"a":[7,101,112,121,136],"new":[8],"wave":[9],"of":[10,96,161,177],"AI":[11],"applications.":[12],"However,":[13],"their":[14],"large":[15],"computational":[16],"complexity":[17],"and":[18,80,180,191],"memory":[19],"footprint":[20],"pose":[21],"significant":[22],"challenges":[23],"for":[24,41,76],"real-world":[25],"deployment.":[26],"Although":[27],"dedicated":[28],"transformer":[29,188],"accelerators":[30,189],"been":[32],"widely":[33],"explored,":[34],"we":[35,72,99,134,150],"observe":[36],"that":[37,44,55,141],"they":[38],"are":[39],"unefficient":[40],"decoder-only":[42],"LLMs":[43,77],"feature":[45],"autoregressive":[46],"computations":[47],"with":[48,131],"KV":[49,61,97,118],"Cache.":[50],"Our":[51],"in-depth":[52],"analysis":[53],"reveals":[54],"DRAM":[56],"accesses":[57],"induced":[58],"by":[59,92],"the":[60,64,84,89,93,124,152,166],"Cache":[62,119],"dominate":[63],"overall":[65],"attention":[66],"process.":[67],"To":[68],"address":[69],"this":[70],"issue,":[71],"propose":[73,100],"aMemory-efficientATtentionAccelerator":[74],"(MATA)":[75],"through":[78,111],"algorithm":[79,85],"hardware":[81,125],"co-design.":[82],"Specifically,at":[83],"level,":[86,126],"to":[87,120,127],"mitigate":[88],"overhead":[90],"caused":[91],"linear":[94],"increase":[95],"Cache,":[98],"post-training":[102],"Look-Back":[103],"pruning":[104],"method.":[105],"It":[106],"dynamically":[107],"discards":[108],"unimportant":[109],"tokens":[110,130],"comprehensive":[113],"scoring":[114],"scheme,":[115],"thereby":[116],"restricting":[117],"constant":[122],"volume.At":[123],"identify":[128],"important":[129],"low":[132],"latency,":[133],"design":[135],"Slice":[137],"Top-K":[138],"(STK)":[139],"engine":[140],"can":[142,174],"complete":[143],"top-k-based":[144],"sorting":[145],"withO(N)":[146],"time":[147],"complexity.":[148],"Moreover,":[149],"present":[151],"Adaptive":[153],"Dataflow,":[154],"which":[155],"adaptively":[156],"performs":[157],"different":[158],"inference":[159],"phases":[160],"LLMs,":[162],"thus":[163],"significantly":[164],"enhancing":[165],"PE":[167],"array":[168],"utilization.":[169],"On":[170],"average,":[171],"our":[172],"MATA":[173],"achieve":[175],"speedups":[176],"3.56\u00d7,":[178],"2.23\u00d7":[179],"2.04\u00d7,":[181],"1.56\u00d7":[182],"energy":[183],"savings":[184],"over":[185],"two":[186],"state-of-the-art":[187],"SpAtten":[190],"FACT,":[192],"respectively.":[193]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-29T09:40:50.615796","created_date":"2025-10-28T00:00:00"}
