{"id":"https://openalex.org/W7162749685","doi":"https://doi.org/10.48550/arxiv.2605.29233","title":"BlockBatch: Multi-Scale Consensus Decoding for Efficient Diffusion Language Model Inference","display_name":"BlockBatch: Multi-Scale Consensus Decoding for Efficient Diffusion Language Model Inference","publication_year":2026,"publication_date":"2026-05-28","ids":{"openalex":"https://openalex.org/W7162749685","doi":"https://doi.org/10.48550/arxiv.2605.29233"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.29233","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29233","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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.2605.29233","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137335287","display_name":"Xiaoyou Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Xiaoyou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137389647","display_name":"Cheng-Jhih Shih","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shih, Cheng-Jhih","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137338073","display_name":"Binfei Ji","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ji, Binfei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137381398","display_name":"Yong Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Yong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137363325","display_name":"Yingyan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Yingyan Celine","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":0.5415999889373779,"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":0.5415999889373779,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.08290000259876251,"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":0.06159999966621399,"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/block","display_name":"Block (permutation group theory)","score":0.6420000195503235},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6251999735832214},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5893999934196472},{"id":"https://openalex.org/keywords/granularity","display_name":"Granularity","score":0.5511999726295471},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5249999761581421},{"id":"https://openalex.org/keywords/complementarity","display_name":"Complementarity (molecular biology)","score":0.45500001311302185},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.447299987077713},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.444599986076355}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7623999714851379},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.6420000195503235},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6251999735832214},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5893999934196472},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.5511999726295471},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5249999761581421},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4885999858379364},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.4666000008583069},{"id":"https://openalex.org/C202269582","wikidata":"https://www.wikidata.org/wiki/Q2644277","display_name":"Complementarity (molecular biology)","level":2,"score":0.45500001311302185},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.447299987077713},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.444599986076355},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.438400000333786},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.3693999946117401},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.36059999465942383},{"id":"https://openalex.org/C41431624","wikidata":"https://www.wikidata.org/wiki/Q1053357","display_name":"Block size","level":3,"score":0.30720001459121704},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3061000108718872},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2962000072002411},{"id":"https://openalex.org/C2779982251","wikidata":"https://www.wikidata.org/wiki/Q25053762","display_name":"Stochastic block model","level":3,"score":0.27489998936653137},{"id":"https://openalex.org/C70437156","wikidata":"https://www.wikidata.org/wiki/Q7228652","display_name":"Pooling","level":2,"score":0.2736999988555908},{"id":"https://openalex.org/C2781142347","wikidata":"https://www.wikidata.org/wiki/Q1366592","display_name":"Hilbert curve","level":2,"score":0.2685000002384186}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.29233","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29233","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.29233","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.29233","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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":{"Diffusion":[0],"language":[1],"models":[2],"(dLLMs)":[3],"generate":[4],"text":[5],"by":[6,115,176],"iteratively":[7],"denoising":[8,41,174],"multiple":[9,128],"token":[10,146],"positions":[11],"in":[12],"parallel,":[13],"offering":[14],"an":[15,99],"attractive":[16],"alternative":[17],"to":[18,159],"strictly":[19],"autoregressive":[20],"decoding.":[21],"In":[22],"practice,":[23],"however,":[24],"block-wise":[25],"dLLM":[26,206],"inference":[27,124],"exposes":[28],"a":[29,63,83,121,136,160,182,198],"difficult":[30],"granularity":[31],"trade-off:":[32],"small":[33],"blocks":[34,45],"preserve":[35],"local":[36,156],"conditioning":[37],"but":[38,49,92],"require":[39],"many":[40],"steps,":[42],"whereas":[43],"large":[44],"expose":[46],"more":[47],"parallelism":[48],"can":[50],"make":[51],"premature":[52],"commitments":[53],"and":[54,107,150,169,180,200],"accumulate":[55],"cache":[56],"error.":[57],"Existing":[58],"acceleration":[59],"methods":[60],"typically":[61],"choose":[62],"single":[64],"block":[65,73,79,88,157],"size":[66,80],"per":[67],"request,":[68],"leaving":[69],"the":[70,132],"complementarity":[71],"among":[72],"sizes":[74,89],"unused.":[75],"We":[76],"show":[77],"that":[78,126,154],"itself":[81],"is":[82],"useful":[84],"branching":[85],"dimension.":[86],"Different":[87],"induce":[90],"related":[91],"non-identical":[93],"KV-cache":[94],"trajectories:":[95],"branches":[96,130,143],"often":[97],"share":[98],"initial":[100],"prefix,":[101],"bifurcate":[102],"at":[103],"semantically":[104],"decisive":[105],"positions,":[106],"later":[108],"agree":[109],"on":[110,178],"syntactically":[111],"lightweight":[112],"tokens.":[113],"Motivated":[114],"this":[116],"structure,":[117],"we":[118],"propose":[119],"BlockBatch,":[120],"training-free":[122],"online":[123],"framework":[125],"executes":[127],"block-size":[129,195],"for":[131,204],"same":[133],"request":[134],"inside":[135],"batched":[137],"forward":[138],"pass.":[139],"BlockBatch":[140,172],"coordinates":[141],"these":[142],"through":[144],"confidence-gated":[145],"merging,":[147],"leader-based":[148],"synchronization,":[149],"periodic":[151],"full-sequence":[152],"refreshes":[153],"re-anchor":[155],"updates":[158],"globally":[161],"consistent":[162],"KV":[163],"state.":[164],"Across":[165],"3":[166],"representative":[167],"dLLMs":[168],"4":[170],"datasets,":[171],"reduces":[173],"NFEs":[175],"26.6\\%":[177],"average":[179,184],"achieves":[181],"1.33$\\times$":[183],"end-to-end":[185],"speedup":[186],"over":[187],"Fast-dLLM":[188],"while":[189],"preserving":[190],"accuracy.":[191],"These":[192],"results":[193],"identify":[194],"diversity":[196],"as":[197],"practical":[199],"previously":[201],"underexplored":[202],"axis":[203],"branch-parallel":[205],"inference.":[207]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-30T00:00:00"}