{"id":"https://openalex.org/W7161780335","doi":"https://doi.org/10.48550/arxiv.2605.19726","title":"Efficient Long-Context Modeling in Diffusion Language Models via Block Approximate Sparse Attention","display_name":"Efficient Long-Context Modeling in Diffusion Language Models via Block Approximate Sparse Attention","publication_year":2026,"publication_date":"2026-05-19","ids":{"openalex":"https://openalex.org/W7161780335","doi":"https://doi.org/10.48550/arxiv.2605.19726"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.19726","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19726","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.19726","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136580260","display_name":"Wenhu Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Wenhu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136546042","display_name":"Yiming Wu","orcid":"https://orcid.org/0009-0008-9936-6737"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Yiming","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136595486","display_name":"Huanyu Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Huanyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136602198","display_name":"Yaoyang Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Yaoyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5000675851","display_name":"Huanzhang Dou","orcid":"https://orcid.org/0000-0002-3277-5906"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dou, Huanzhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5066489766","display_name":"Senqiao Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Senqiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136589965","display_name":"Sitong Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Sitong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136511022","display_name":"Hanbin Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Hanbin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136552324","display_name":"Jiaya Jia","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jia, Jiaya","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.24879999458789825,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.24879999458789825,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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":0.19589999318122864,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.16459999978542328,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/autoregressive-model","display_name":"Autoregressive model","score":0.5922999978065491},{"id":"https://openalex.org/keywords/oracle","display_name":"Oracle","score":0.5609999895095825},{"id":"https://openalex.org/keywords/block","display_name":"Block (permutation group theory)","score":0.5579000115394592},{"id":"https://openalex.org/keywords/salient","display_name":"Salient","score":0.4611000120639801},{"id":"https://openalex.org/keywords/sampling","display_name":"Sampling (signal processing)","score":0.4496000111103058},{"id":"https://openalex.org/keywords/covariance","display_name":"Covariance","score":0.41370001435279846},{"id":"https://openalex.org/keywords/operator","display_name":"Operator (biology)","score":0.38589999079704285},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.3831999897956848}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7337999939918518},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.5922999978065491},{"id":"https://openalex.org/C55166926","wikidata":"https://www.wikidata.org/wiki/Q2892946","display_name":"Oracle","level":2,"score":0.5609999895095825},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.5579000115394592},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5105999708175659},{"id":"https://openalex.org/C2780719617","wikidata":"https://www.wikidata.org/wiki/Q1030752","display_name":"Salient","level":2,"score":0.4611000120639801},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.4496000111103058},{"id":"https://openalex.org/C178650346","wikidata":"https://www.wikidata.org/wiki/Q201984","display_name":"Covariance","level":2,"score":0.41370001435279846},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.40209999680519104},{"id":"https://openalex.org/C17020691","wikidata":"https://www.wikidata.org/wiki/Q139677","display_name":"Operator (biology)","level":5,"score":0.38589999079704285},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3831999897956848},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3675000071525574},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.358599990606308},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.3425000011920929},{"id":"https://openalex.org/C2780428219","wikidata":"https://www.wikidata.org/wiki/Q16952335","display_name":"Cover (algebra)","level":2,"score":0.3321000039577484},{"id":"https://openalex.org/C130367717","wikidata":"https://www.wikidata.org/wiki/Q189791","display_name":"Diagonal","level":2,"score":0.3312000036239624},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.29249998927116394},{"id":"https://openalex.org/C186644900","wikidata":"https://www.wikidata.org/wiki/Q194152","display_name":"Parsing","level":2,"score":0.2854999899864197},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.2833999991416931},{"id":"https://openalex.org/C185142706","wikidata":"https://www.wikidata.org/wiki/Q1134404","display_name":"Covariance matrix","level":2,"score":0.2815000116825104},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.27639999985694885},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.2685000002384186},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2639000117778778},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.2623000144958496}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.19726","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19726","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.19726","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.19726","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","score":0.56512850522995,"display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Diffusion":[0],"Language":[1],"Models":[2],"(DLMs)":[3],"enable":[4],"globally":[5],"coherent,":[6],"bidirectional,":[7],"and":[8,55,104,111,124,155,169,176],"controllable":[9],"text":[10],"generation,":[11],"offering":[12],"advantages":[13],"over":[14,36,150],"traditional":[15],"autoregressive":[16],"LLMs,":[17],"while":[18],"scaling":[19],"to":[20,147],"ultra-long":[21],"sequences":[22],"remains":[23],"costly.":[24],"Many":[25],"existing":[26],"block-sparse":[27],"attention":[28,39,102,153],"methods":[29],"select":[30],"blocks":[31],"by":[32],"fixed":[33],"sampling":[34,50],"patterns":[35],"the":[37,66,106],"high-resolution":[38],"space,":[40,85],"such":[41],"as":[42],"tail":[43],"regions":[44,80],"or":[45],"anti-diagonal":[46],"stripes.":[47],"Such":[48],"prior-driven":[49],"can":[51],"miss":[52],"salient":[53],"tokens":[54],"introduce":[56,119],"instability":[57],"under":[58],"distribution":[59],"shifts.":[60],"In":[61],"this":[62,116],"paper,":[63],"we":[64,97,118],"propose":[65],"Block":[67],"Approximate":[68],"Sparse":[69],"Attention":[70],"framework":[71],"(BA-Att)":[72],"with":[73],"block-wise":[74],"pre-downsampled":[75],"operation,":[76],"which":[77],"identifies":[78],"informative":[79],"within":[81],"a":[82,120,125],"compact":[83],"downsampled":[84],"avoiding":[86],"reliance":[87],"on":[88,115],"brittle":[89],"positional":[90],"priors.":[91],"To":[92],"analyze":[93],"its":[94],"theoretical":[95],"behavior,":[96],"define":[98],"an":[99],"oracle":[100],"post-downsample":[101,112],"map":[103],"formalize":[105],"approximation":[107],"error":[108],"between":[109],"pre-":[110],"schemes.":[113],"Based":[114],"insight,":[117],"lightweight":[121],"norm-sorting":[122],"module":[123],"covariance-compensated":[126],"correction":[127],"that":[128,142],"approximates":[129],"full":[130],"covariance":[131],"using":[132],"diagonal":[133],"QK":[134],"variances,":[135],"reducing":[136],"computational":[137],"complexity.":[138],"Extensive":[139],"experiments":[140],"show":[141],"our":[143],"operator":[144],"achieves":[145],"up":[146],"6.95x":[148],"acceleration":[149],"FlashAttention":[151],"in":[152],"computation,":[154],"maintains":[156],"near":[157],"full-attention":[158],"performance":[159],"at":[160],"50%":[161],"sparsity":[162],"across":[163],"language":[164,167],"models,":[165,168,172],"multimodal":[166],"video":[170],"generation":[171],"demonstrating":[173],"strong":[174],"efficiency":[175],"generalization.":[177]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-21T00:00:00"}