{"id":"https://openalex.org/W7204596471","doi":"https://doi.org/10.48550/arxiv.2608.26140","title":"Affix Cache for Diffusion Large Language Models","display_name":"Affix Cache for Diffusion Large Language Models","publication_year":2026,"publication_date":"2026-06-26","ids":{"openalex":"https://openalex.org/W7204596471","doi":"https://doi.org/10.48550/arxiv.2608.26140"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.26140","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.26140","pdf_url":"https://arxiv.org/pdf/2608.26140","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.26140","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124866638","display_name":"Kaihua Liang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Kaihua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124786205","display_name":"An Zhong","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhong, An","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126584693","display_name":"Xin Tan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Tan, Xin","raw_affiliation_strings":["CUHK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CUHK","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125525069","display_name":"Zafar Ayyub Qazi","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I207789805","display_name":"Lahore University of Management Sciences","ror":"https://ror.org/05b5x4a35","country_code":"PK","type":"education","lineage":["https://openalex.org/I207789805"]}],"countries":["PK"],"is_corresponding":false,"raw_author_name":"Qazi, Zafar Ayyub","raw_affiliation_strings":["LUMS"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"LUMS","institution_ids":["https://openalex.org/I207789805"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124816770","display_name":"Hong Xu","orcid":"https://orcid.org/0000-0003-1389-5408","observed_orcids":["https://orcid.org/0000-0003-1389-5408"]},"institutions":[{"id":"https://openalex.org/I177725633","display_name":"Chinese University of Hong Kong","ror":"https://ror.org/00t33hh48","country_code":"HK","type":"education","lineage":["https://openalex.org/I177725633"]}],"countries":["HK"],"is_corresponding":false,"raw_author_name":"Xu, Hong","raw_affiliation_strings":["CUHK"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"CUHK","institution_ids":["https://openalex.org/I177725633"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139358511","display_name":"Jian Weng","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Weng, Jian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5120609513","display_name":"Marco Canini","orcid":"https://orcid.org/0000-0002-5051-4283","observed_orcids":["https://orcid.org/0000-0002-5051-4283"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Canini, Marco","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.6445000171661377,"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":0.6445000171661377,"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":0.13619999587535858,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.12720000743865967,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/diffusion-large-language-models","display_name":"diffusion large language models","score":0.9890000224113464},{"id":"https://openalex.org/keywords/bidirectional-attention","display_name":"bidirectional attention","score":0.6439999938011169},{"id":"https://openalex.org/keywords/non-autoregressive-decoding","display_name":"non-autoregressive decoding","score":0.8579999804496765},{"id":"https://openalex.org/keywords/inference-efficiency","display_name":"inference efficiency","score":0.621999979019165}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8497999906539917},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.7383999824523926},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.6822999715805054},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.5317000150680542},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.4959000051021576},{"id":"https://openalex.org/C2778428490","wikidata":"https://www.wikidata.org/wiki/Q62155","display_name":"Affix","level":2,"score":0.49410000443458557},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.47690001130104065},{"id":"https://openalex.org/C90805587","wikidata":"https://www.wikidata.org/wiki/Q10944557","display_name":"Word (group theory)","level":2,"score":0.35120001435279846},{"id":"https://openalex.org/C130795937","wikidata":"https://www.wikidata.org/wiki/Q2561570","display_name":"Remote direct memory access","level":2,"score":0.34610000252723694},{"id":"https://openalex.org/C70388272","wikidata":"https://www.wikidata.org/wiki/Q5968558","display_name":"IBM","level":2,"score":0.33550000190734863},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.33180001378059387},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.3167000114917755},{"id":"https://openalex.org/C189783530","wikidata":"https://www.wikidata.org/wiki/Q352090","display_name":"CPU cache","level":3,"score":0.3154999911785126},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.29269999265670776},{"id":"https://openalex.org/C2778583558","wikidata":"https://www.wikidata.org/wiki/Q771245","display_name":"Code reuse","level":3,"score":0.27160000801086426},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.26969999074935913},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2653000056743622},{"id":"https://openalex.org/C46637626","wikidata":"https://www.wikidata.org/wiki/Q6693015","display_name":"Low latency (capital markets)","level":2,"score":0.25999999046325684}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.26140","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.26140","pdf_url":"https://arxiv.org/pdf/2608.26140","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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"},{"id":"doi:10.48550/arxiv.2608.26140","doi":"https://doi.org/10.48550/arxiv.2608.26140","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.26140","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.26140","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.26140","pdf_url":"https://arxiv.org/pdf/2608.26140","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"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":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204596471.pdf","grobid_xml":"https://content.openalex.org/works/W7204596471.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7169828316","https://openalex.org/W4415332076","https://openalex.org/W7092289163","https://openalex.org/W4415329327","https://openalex.org/W6904055037","https://openalex.org/W4417105316","https://openalex.org/W7164032507","https://openalex.org/W7139935094","https://openalex.org/W7202215318","https://openalex.org/W4415037641"],"abstract_inverted_index":{"Diffusion":[0],"Large":[1],"Language":[2],"Models":[3],"(DLLMs)":[4],"enable":[5],"non-autoregressive":[6],"decoding,":[7],"but":[8],"efficient":[9],"inference":[10,153],"support":[11],"remains":[12],"immature:":[13],"unlike":[14],"autoregressive":[15],"models,":[16],"whose":[17,155],"requests":[18],"reuse":[19,58,121],"a":[20,31,55,86,151],"shared":[21,32,61,168],"prefix":[22],"key-value":[23],"(KV)":[24],"cache,":[25],"DLLMs":[26],"use":[27],"bidirectional":[28],"attention,":[29],"so":[30],"context's":[33],"KV":[34,97,162,199],"states":[35,98,163],"depend":[36],"on":[37,79,106,122],"the":[38,74,80,101,115,173],"tokens":[39,78],"still":[40],"being":[41],"decoded,":[42],"leaving":[43],"directly":[44],"reused":[45],"caches":[46],"stale":[47],"and":[48,93,131],"full":[49],"recomputation":[50],"necessary.":[51],"We":[52,147],"present":[53],"ACache,":[54],"cross-request":[56,144],"cache":[57,167,200],"mechanism":[59],"for":[60],"spans,":[62],"or":[63,70],"affixes,":[64],"at":[65,132],"any":[66],"position:":[67],"prefix,":[68],"infix,":[69],"suffix.":[71],"ACache":[72,111,149,180],"measures":[73],"influence":[75],"of":[76,114,128],"affix":[77,103,129,166],"masked":[81],"generation":[82],"region":[83],"to":[84,118,186,189,204],"identify":[85],"small":[87],"request-specific":[88],"subset":[89],"as":[90,190,192],"Anchor":[91,161],"Tokens,":[92],"recomputes":[94],"only":[95,177],"their":[96],"while":[99,196],"reusing":[100],"remaining":[102],"cache.":[104],"Built":[105],"state-of-the-art":[107],"intra-request":[108,178],"caching":[109],"mechanisms,":[110],"recovers":[112],"most":[113],"accuracy":[116,137],"lost":[117],"direct":[119],"affix-cache":[120],"average":[123],"when":[124],"recomputing":[125],"around":[126],"20%":[127],"tokens,":[130],"that":[133],"budget":[134],"preserves":[135],"more":[136],"than":[138],"selection":[139],"criteria":[140],"adapted":[141],"from":[142],"prior":[143],"cache-reuse":[145],"systems.":[146],"co-design":[148],"with":[150,176],"modern":[152],"engine,":[154],"attention":[156],"reads":[157],"each":[158],"request's":[159],"recomputed":[160],"alongside":[164],"one":[165],"across":[169],"concurrent":[170],"requests.":[171],"Against":[172],"same":[174],"system":[175],"caching,":[179],"cuts":[181],"recompute":[182],"latency":[183],"by":[184,202],"up":[185,203],"56.7%,":[187],"translating":[188],"much":[191],"1.71$\\times$":[193],"end-to-end":[194],"throughput,":[195],"reducing":[197],"peak":[198],"memory":[201],"45.8%.":[205]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-29T00:00:00"}