{"id":"https://openalex.org/W7161977113","doi":"https://doi.org/10.48550/arxiv.2605.20813","title":"PulseCol: Periodically Refreshed Column-Sparse Attention for Accelerating Diffusion Language Models","display_name":"PulseCol: Periodically Refreshed Column-Sparse Attention for Accelerating Diffusion Language Models","publication_year":2026,"publication_date":"2026-05-20","ids":{"openalex":"https://openalex.org/W7161977113","doi":"https://doi.org/10.48550/arxiv.2605.20813"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.20813","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20813","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.2605.20813","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136633194","display_name":"Yanyi Lyu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lyu, Yanyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136626285","display_name":"Letian Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Letian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113124197","display_name":"Futing Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Futing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136626719","display_name":"Miao Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Miao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136635356","display_name":"Weili Guan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guan, Weili","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136628541","display_name":"Liqiang Nie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nie, Liqiang","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/T10028","display_name":"Topic Modeling","score":0.1891999989748001,"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":0.1891999989748001,"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":0.10080000013113022,"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/T11948","display_name":"Machine Learning in Materials Science","score":0.07050000131130219,"subfield":{"id":"https://openalex.org/subfields/2505","display_name":"Materials Chemistry"},"field":{"id":"https://openalex.org/fields/25","display_name":"Materials Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.8927000164985657},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6638000011444092},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.6456999778747559},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.6047000288963318},{"id":"https://openalex.org/keywords/sparse-matrix","display_name":"Sparse matrix","score":0.49140000343322754},{"id":"https://openalex.org/keywords/noise-reduction","display_name":"Noise reduction","score":0.4537999927997589},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.4212000072002411}],"concepts":[{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.8927000164985657},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8111000061035156},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6638000011444092},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.6456999778747559},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.6047000288963318},{"id":"https://openalex.org/C56372850","wikidata":"https://www.wikidata.org/wiki/Q1050404","display_name":"Sparse matrix","level":3,"score":0.49140000343322754},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.4537999927997589},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4336000084877014},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.4212000072002411},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4203999936580658},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.3427000045776367},{"id":"https://openalex.org/C147297375","wikidata":"https://www.wikidata.org/wiki/Q6674930","display_name":"Look-ahead","level":2,"score":0.3285999894142151},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.32190001010894775},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.3192000091075897},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.3052000105381012},{"id":"https://openalex.org/C68710425","wikidata":"https://www.wikidata.org/wiki/Q5275442","display_name":"Diffusion process","level":3,"score":0.2750999927520752},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.2705000042915344},{"id":"https://openalex.org/C124066611","wikidata":"https://www.wikidata.org/wiki/Q28684319","display_name":"Sparse approximation","level":2,"score":0.2660999894142151},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.25920000672340393},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2526000142097473}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.20813","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20813","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.2605.20813","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20813","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Inference":[0],"in":[1,43,60],"diffusion":[2,101],"large":[3],"language":[4,102],"models":[5],"(dLLMs)":[6],"is":[7,40,49],"computationally":[8],"expensive,":[9],"as":[10],"full":[11],"self-attention":[12],"must":[13],"be":[14,73,119],"repeatedly":[15],"executed":[16],"at":[17,137,151],"each":[18],"step":[19,141],"of":[20,155,162],"the":[21,138,160],"denoising":[22,140],"process":[23],"without":[24],"KV":[25],"cache.":[26],"Recent":[27],"sparse":[28,54,135,163,181],"attention":[29,97,116,164,182],"methods":[30,183],"for":[31,99,184,195],"dLLMs":[32],"mitigate":[33],"this":[34,88,129],"cost":[35],"via":[36],"block-sparse":[37],"computation,":[38],"which":[39],"applied":[41,74],"only":[42,150],"later":[44],"iterations":[45,77],"when":[46],"model":[47,188],"performance":[48],"less":[50],"sensitive":[51],"to":[52,118,158,201],"coarse-grained":[53],"approximation,":[55],"but":[56],"yields":[57],"limited":[58],"improvements":[59],"computational":[61],"efficiency":[62,85],"and":[63,78,142,175],"acceleration.":[64],"This":[65],"motivates":[66],"a":[67,93,110,152],"finer-grained":[68,111],"sparsification":[69],"strategy":[70],"that":[71,170],"can":[72],"from":[75],"earlier":[76],"leverages":[79],"reusable":[80],"sparsity":[81,108,174],"patterns,":[82],"enabling":[83],"further":[84,133],"gains.":[86],"In":[87],"work,":[89],"we":[90],"introduce":[91],"PulseCol,":[92],"periodically":[94],"refreshed":[95],"column-sparse":[96,112,196],"method":[98],"accelerating":[100],"models.":[103],"PulseCol":[104,132,171,198],"replaces":[105],"coarse":[106],"block-level":[107],"with":[109],"structure,":[113],"allowing":[114],"important":[115],"interactions":[117],"retained":[120],"more":[121],"precisely":[122],"while":[123,186],"exposing":[124],"greater":[125,176],"sparsity.":[126],"Built":[127],"on":[128],"column-level":[130],"formulation,":[131],"identifies":[134],"patterns":[136,165],"early":[139],"reuses":[143],"them":[144,149],"across":[145,207],"subsequent":[146],"iterations,":[147],"refreshing":[148],"small":[153],"number":[154],"intermediate":[156],"steps":[157],"track":[159],"evolution":[161],"during":[166],"denoising.":[167],"Experiments":[168],"show":[169],"achieves":[172],"higher":[173],"practical":[177],"speedup":[178,204],"than":[179],"prior":[180],"dLLMs,":[185],"maintaining":[187],"quality.":[189],"Enabled":[190],"by":[191],"optimized":[192],"GPU":[193],"kernels":[194],"attention,":[197],"delivers":[199],"up":[200],"1.95$\\times$":[202],"end-to-end":[203],"over":[205],"FlashAttention":[206],"several":[208],"context":[209],"lengths.":[210]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-22T00:00:00"}