{"id":"https://openalex.org/W7166307650","doi":"https://doi.org/10.1145/3801487.3801812","title":"MKA: Memory-Keyed Attention for Efficient Long-Context Reasoning","display_name":"MKA: Memory-Keyed Attention for Efficient Long-Context Reasoning","publication_year":2026,"publication_date":"2026-05-19","ids":{"openalex":"https://openalex.org/W7166307650","doi":"https://doi.org/10.1145/3801487.3801812"},"language":null,"primary_location":{"id":"doi:10.1145/3801487.3801812","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3801487.3801812","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 23rd ACM International Conference on Computing Frontiers","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3801487.3801812","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5126241617","display_name":"Dong Liu","orcid":null},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Dong Liu","raw_affiliation_strings":["University of California - Los Angeles, Los Angeles, California, USA"],"raw_orcid":"https://orcid.org/0009-0009-6815-8297","affiliations":[{"raw_affiliation_string":"University of California - Los Angeles, Los Angeles, California, USA","institution_ids":["https://openalex.org/I161318765"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5106342688","display_name":"Yang Yu","orcid":"https://orcid.org/0000-0002-6286-0490"},"institutions":[{"id":"https://openalex.org/I78577930","display_name":"Columbia University","ror":"https://ror.org/00hj8s172","country_code":"US","type":"education","lineage":["https://openalex.org/I78577930"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yanxuan Yu","raw_affiliation_strings":["Columbia University, New York, New York, USA"],"raw_orcid":"https://orcid.org/0009-0001-7096-2438","affiliations":[{"raw_affiliation_string":"Columbia University, New York, New York, USA","institution_ids":["https://openalex.org/I78577930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5068931679","display_name":"Benjamin J. Lengerich","orcid":"https://orcid.org/0000-0001-8690-9554"},"institutions":[{"id":"https://openalex.org/I135310074","display_name":"University of Wisconsin\u2013Madison","ror":"https://ror.org/01y2jtd41","country_code":"US","type":"education","lineage":["https://openalex.org/I135310074"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ben Lengerich","raw_affiliation_strings":["University of Wisconsin-Madison, Madison, Wisconsin, USA"],"raw_orcid":"https://orcid.org/0000-0001-8690-9554","affiliations":[{"raw_affiliation_string":"University of Wisconsin-Madison, Madison, Wisconsin, USA","institution_ids":["https://openalex.org/I135310074"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5042765290","display_name":"Ying Wu","orcid":"https://orcid.org/0000-0001-8029-3664"},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ying Nian Wu","raw_affiliation_strings":["University of California - Los Angeles, Los Angeles, California, USA"],"raw_orcid":"https://orcid.org/0000-0001-8029-3664","affiliations":[{"raw_affiliation_string":"University of California - Los Angeles, Los Angeles, California, USA","institution_ids":["https://openalex.org/I161318765"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"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":0.93718247,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"280","last_page":"289"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T13702","display_name":"Machine Learning in Healthcare","score":0.16940000653266907,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.16940000653266907,"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"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.12430000305175781,"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"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.1120000034570694,"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/bottleneck","display_name":"Bottleneck","score":0.7738999724388123},{"id":"https://openalex.org/keywords/perplexity","display_name":"Perplexity","score":0.6974999904632568},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.4862000048160553},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.46880000829696655},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.4366999864578247},{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.4275999963283539}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8152999877929688},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.7738999724388123},{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.6974999904632568},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4902999997138977},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.4862000048160553},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.46880000829696655},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.4366999864578247},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.4275999963283539},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.4052000045776367},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.37070000171661377},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.34689998626708984},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.30480000376701355},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.2745000123977661},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2727999985218048},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2597000002861023}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3801487.3801812","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3801487.3801812","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 23rd ACM International Conference on Computing Frontiers","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3801487.3801812","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3801487.3801812","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 23rd ACM International Conference on Computing Frontiers","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":6,"referenced_works":["https://openalex.org/W2530887700","https://openalex.org/W2964110616","https://openalex.org/W3131922516","https://openalex.org/W4389518760","https://openalex.org/W4404782921","https://openalex.org/W7160210308"],"related_works":[],"abstract_inverted_index":{"As":[0],"long-context":[1,146],"language":[2],"modeling":[3],"becomes":[4],"increasingly":[5],"important,":[6],"the":[7],"cost":[8],"of":[9],"maintaining":[10],"and":[11,27,37,74,128,141],"attending":[12],"to":[13,77,117,122],"large":[14],"Key/Value":[15],"(KV)":[16],"caches":[17],"grows":[18],"rapidly,":[19],"becoming":[20],"a":[21,64,89,111,139],"major":[22],"bottleneck":[23],"in":[24],"both":[25],"training":[26,126],"inference.":[28],"While":[29],"prior":[30],"works":[31],"such":[32],"as":[33,138],"Multi-Query":[34],"Attention":[35,39,62],"(MQA)":[36],"Multi-Latent":[38],"(MLA)":[40],"reduce":[41],"memory":[42,94],"by":[43],"sharing":[44],"or":[45,55],"compressing":[46],"KV":[47,71],"features,":[48],"they":[49],"often":[50],"trade":[51],"off":[52],"representation":[53],"quality":[54],"incur":[56],"runtime":[57],"overhead.":[58],"We":[59,83],"propose":[60],"Memory-Keyed":[61],"(MKA),":[63],"hierarchical":[65],"attention":[66,79,97],"mechanism":[67],"that":[68,92,108],"integrates":[69],"multi-level":[70],"caches\u2014local,":[72],"session,":[73],"long-term\u2014and":[75],"learns":[76],"route":[78],"across":[80],"them":[81],"dynamically.":[82],"further":[84],"introduce":[85],"Route-Fused":[86],"MKA":[87,137],"(FastMKA),":[88],"broadcast-routed":[90],"variant":[91],"fuses":[93],"sources":[95],"before":[96],"computation":[98],"for":[99,144],"enhanced":[100],"efficiency.":[101],"Experiments":[102],"on":[103],"different":[104],"sequence":[105],"lengths":[106],"show":[107],"FastMKA":[109],"achieves":[110],"favorable":[112],"accuracy-efficiency":[113],"trade-off:":[114],"comparable":[115],"perplexity":[116],"MLA":[118],"while":[119],"achieving":[120],"up":[121],"5":[123],"\u00d7":[124,130],"faster":[125],"throughput":[127],"1.8":[129],"lower":[131],"evaluation":[132],"latency.":[133],"These":[134],"results":[135],"highlight":[136],"practical":[140],"extensible":[142],"framework":[143],"efficient":[145],"attention.":[147]},"counts_by_year":[],"updated_date":"2026-06-28T06:20:29.262167","created_date":"2026-06-28T00:00:00"}