{"id":"https://openalex.org/W7128364940","doi":"https://doi.org/10.48550/arxiv.2602.06412","title":"Stopping Computation for Converged Tokens in Masked Diffusion-LM Decoding","display_name":"Stopping Computation for Converged Tokens in Masked Diffusion-LM Decoding","publication_year":2026,"publication_date":"2026-02-06","ids":{"openalex":"https://openalex.org/W7128364940","doi":"https://doi.org/10.48550/arxiv.2602.06412"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2602.06412","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.06412","pdf_url":"https://arxiv.org/pdf/2602.06412","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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/2602.06412","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5055651881","display_name":"Daisuke Oba","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4400009020","display_name":"Institute of Science Tokyo","ror":"https://ror.org/05dqf9946","country_code":"JP","type":"education","lineage":["https://openalex.org/I4400009020"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Oba, Daisuke","raw_affiliation_strings":["Institute of Science Tokyo"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Science Tokyo","institution_ids":["https://openalex.org/I4400009020"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120058124","display_name":"Danushka Bollegala","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]},{"id":"https://openalex.org/I146655781","display_name":"University of Liverpool","ror":"https://ror.org/04xs57h96","country_code":"GB","type":"education","lineage":["https://openalex.org/I146655781"]}],"countries":["GB","US"],"is_corresponding":false,"raw_author_name":"Bollegala, Danushka","raw_affiliation_strings":["Amazon","University of Liverpool"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon","institution_ids":["https://openalex.org/I1311688040"]},{"raw_affiliation_string":"University of Liverpool","institution_ids":["https://openalex.org/I146655781"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125422774","display_name":"Masahiro Kaneko","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210113480","display_name":"Mohamed bin Zayed University of Artificial Intelligence","ror":"https://ror.org/0258gkt32","country_code":"AE","type":"education","lineage":["https://openalex.org/I4210113480"]}],"countries":["AE"],"is_corresponding":false,"raw_author_name":"Kaneko, Masahiro","raw_affiliation_strings":["MBZUAI"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MBZUAI","institution_ids":["https://openalex.org/I4210113480"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5107642698","display_name":"Naoaki Okazaki","orcid":"https://orcid.org/0000-0001-7635-6175","observed_orcids":["https://orcid.org/0000-0001-7635-6175"]},"institutions":[{"id":"https://openalex.org/I4400009020","display_name":"Institute of Science Tokyo","ror":"https://ror.org/05dqf9946","country_code":"JP","type":"education","lineage":["https://openalex.org/I4400009020"]}],"countries":["JP"],"is_corresponding":false,"raw_author_name":"Okazaki, Naoaki","raw_affiliation_strings":["Institute of Science Tokyo"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Science Tokyo","institution_ids":["https://openalex.org/I4400009020"]}]}],"institutions":[],"countries_distinct_count":4,"institutions_distinct_count":4,"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":0.5121999979019165,"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":0.5121999979019165,"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":0.45829999446868896,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.003599999938160181,"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/masked-diffusion-language-models","display_name":"masked diffusion language models","score":0.9750000238418579},{"id":"https://openalex.org/keywords/computation-reduction","display_name":"computation reduction","score":0.3919999897480011},{"id":"https://openalex.org/keywords/posterior-stabilization","display_name":"posterior stabilization","score":0.5559999942779541},{"id":"https://openalex.org/keywords/iterative-sampling","display_name":"iterative sampling","score":0.4560000002384186},{"id":"https://openalex.org/keywords/llada","display_name":"LLaDA","score":0.5649999976158142},{"id":"https://openalex.org/keywords/flops-reduction","display_name":"FLOPs reduction","score":0.5070000290870667}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.8410000205039978},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7174999713897705},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.6656000018119812},{"id":"https://openalex.org/C198082294","wikidata":"https://www.wikidata.org/wiki/Q3399648","display_name":"Position (finance)","level":2,"score":0.6485999822616577},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.579200029373169},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.567799985408783},{"id":"https://openalex.org/C174839445","wikidata":"https://www.wikidata.org/wiki/Q1134386","display_name":"Lock (firearm)","level":2,"score":0.5496000051498413},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.5392000079154968},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.47760000824928284},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.4456999897956848},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.39980000257492065},{"id":"https://openalex.org/C57493831","wikidata":"https://www.wikidata.org/wiki/Q3134666","display_name":"Projection (relational algebra)","level":2,"score":0.38040000200271606},{"id":"https://openalex.org/C2778067643","wikidata":"https://www.wikidata.org/wiki/Q166507","display_name":"Interval (graph theory)","level":2,"score":0.3594000041484833},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.32440000772476196},{"id":"https://openalex.org/C75291252","wikidata":"https://www.wikidata.org/wiki/Q1315756","display_name":"TRACE (psycholinguistics)","level":2,"score":0.31450000405311584},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.3061000108718872},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.3009999990463257},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.29350000619888306},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.2824000120162964},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.2700999975204468},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.25769999623298645},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.25049999356269836}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2602.06412","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.06412","pdf_url":"https://arxiv.org/pdf/2602.06412","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2602.06412","doi":"https://doi.org/10.48550/arxiv.2602.06412","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.06412","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":"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2602.06412","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2602.06412","pdf_url":"https://arxiv.org/pdf/2602.06412","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G2682864155","display_name":"\u300c\u6559\u80b2\u7684\u4fa1\u5024\u300d\u306e\u9ad8\u3044\u65e5\u672c\u8a9e\u30b3\u30fc\u30d1\u30b9\u306e\u69cb\u7bc9\u306b\u3088\u308b\u5c0f\u898f\u6a21\u8a00\u8a9e\u30e2\u30c7\u30eb\u306e\u7814\u7a76","funder_award_id":"25H01137","funder_id":"https://openalex.org/F4320334764","funder_display_name":"Japan Society for the Promotion of Science"}],"funders":[{"id":"https://openalex.org/F4320334764","display_name":"Japan Society for the Promotion of Science","ror":"https://ror.org/00hhkn466"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7128364940.pdf","grobid_xml":"https://content.openalex.org/works/W7128364940.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7134844364","https://openalex.org/W7164202783","https://openalex.org/W4416156711","https://openalex.org/W7166638395","https://openalex.org/W7220962774","https://openalex.org/W7163149319","https://openalex.org/W7166852146","https://openalex.org/W4414682260","https://openalex.org/W7150718528","https://openalex.org/W7155194316"],"abstract_inverted_index":{"Masked":[0],"Diffusion":[1],"Language":[2],"Models":[3],"generate":[4],"sequences":[5],"via":[6],"iterative":[7],"sampling":[8],"that":[9,63],"progressively":[10],"unmasks":[11],"tokens.":[12],"However,":[13],"they":[14],"still":[15],"recompute":[16],"the":[17,48,93,105,110,119,127,143,161,168,172,178],"attention":[18,78],"and":[19,71,80,116],"feed-forward":[20,72],"blocks":[21],"for":[22],"every":[23,27],"token":[24,114,182],"position":[25,53,64],"at":[26,50,171,189],"step":[28,174],"--":[29,65,74],"even":[30],"when":[31,47],"many":[32],"unmasked":[33,52],"tokens":[34],"are":[35],"essentially":[36],"fixed,":[37],"resulting":[38],"in":[39,42,180],"substantial":[40,131],"waste":[41],"compute.":[43],"We":[44,153],"propose":[45],"SureLock:":[46,165],"posterior":[49],"an":[51],"has":[54],"stabilized":[55],"across":[56],"steps":[57],"(our":[58],"sure":[59],"condition),":[60],"we":[61],"lock":[62,173],"thereafter":[66],"skipping":[67],"its":[68,77],"query":[69],"projection":[70],"sublayers":[73],"while":[75,148],"caching":[76],"keys":[79],"values":[81],"so":[82],"other":[83],"positions":[84],"can":[85],"continue":[86],"to":[87,89,100,142,159,176],"attend":[88],"it.":[90],"This":[91],"reduces":[92,136],"dominant":[94],"per-iteration":[95],"computational":[96],"cost":[97],"from":[98],"$O(N^2d)$":[99],"$O(MNd)$":[101],"where":[102],"$N$":[103],"is":[104,109,118,187],"sequence":[106],"length,":[107],"$M$":[108,124],"number":[111],"of":[112,164],"unlocked":[113],"positions,":[115],"$d$":[117],"model":[120],"dimension.":[121],"In":[122],"practice,":[123],"decreases":[125],"as":[126],"iteration":[128],"progresses,":[129],"yielding":[130],"savings.":[132],"On":[133],"LLaDA-8B,":[134],"SureLock":[135],"algorithmic":[137],"FLOPs":[138],"by":[139],"30--50%":[140],"relative":[141],"same":[144],"sampler":[145],"without":[146],"locking,":[147],"maintaining":[149],"comparable":[150],"generation":[151],"quality.":[152],"also":[154],"provide":[155],"a":[156],"theoretical":[157],"analysis":[158],"justify":[160],"design":[162],"rationale":[163],"monitoring":[166],"only":[167],"local":[169],"KL":[170],"suffices":[175],"bound":[177],"deviation":[179],"final":[181],"probabilities.":[183],"Our":[184],"project":[185],"page":[186],"available":[188],"https://daioba.github.io/surelock":[190],".":[191]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-02-10T00:00:00"}