{"id":"https://openalex.org/W7133341940","doi":"https://doi.org/10.48550/arxiv.2603.01331","title":"MetaState: Persistent Working Memory Enhances Reasoning in Discrete Diffusion Language Models","display_name":"MetaState: Persistent Working Memory Enhances Reasoning in Discrete Diffusion Language Models","publication_year":2026,"publication_date":"2026-03-02","ids":{"openalex":"https://openalex.org/W7133341940","doi":"https://doi.org/10.48550/arxiv.2603.01331"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.01331","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01331","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.01331","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5119181802","display_name":"Kejing Xia","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Xia, Kejing","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128001241","display_name":"Mingzhe Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Mingzhe","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112940771","display_name":"Lixuan Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Lixuan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079778367","display_name":"Zhenbang Du","orcid":"https://orcid.org/0000-0002-1386-8381"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Du, Zhenbang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127900303","display_name":"Xiangchi Yuan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuan, Xiangchi","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5127904126","display_name":"Qirui Jin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Dachuan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5110214786","display_name":"Wenke Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jin, Qirui","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":null,"display_name":"Lee, Wenke","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Wenke","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":8,"corresponding_author_ids":["https://openalex.org/A5119181802"],"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":0.21199999749660492,"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":0.21199999749660492,"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":0.17499999701976776,"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/T12090","display_name":"Language and cultural evolution","score":0.06260000169277191,"subfield":{"id":"https://openalex.org/subfields/3316","display_name":"Cultural Studies"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/noise-reduction","display_name":"Noise reduction","score":0.7164000272750854},{"id":"https://openalex.org/keywords/bridging","display_name":"Bridging (networking)","score":0.7139000296592712},{"id":"https://openalex.org/keywords/bottleneck","display_name":"Bottleneck","score":0.691100001335144},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.6237999796867371},{"id":"https://openalex.org/keywords/sequence","display_name":"Sequence (biology)","score":0.4767000079154968},{"id":"https://openalex.org/keywords/diffusion","display_name":"Diffusion","score":0.4219000041484833},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.42160001397132874}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7615000009536743},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.7164000272750854},{"id":"https://openalex.org/C174348530","wikidata":"https://www.wikidata.org/wiki/Q188635","display_name":"Bridging (networking)","level":2,"score":0.7139000296592712},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.691100001335144},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.6237999796867371},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4982999861240387},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.4767000079154968},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.4219000041484833},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.42160001397132874},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4072999954223633},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.36739999055862427},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.35179999470710754},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.336899995803833},{"id":"https://openalex.org/C82687282","wikidata":"https://www.wikidata.org/wiki/Q66221","display_name":"Auxiliary memory","level":2,"score":0.3276999890804291},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.3221000134944916},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3206000030040741},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.3025999963283539},{"id":"https://openalex.org/C189950617","wikidata":"https://www.wikidata.org/wiki/Q937228","display_name":"Property (philosophy)","level":2,"score":0.2816999852657318}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.01331","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01331","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.01331","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.01331","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":"article"},"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":{"Discrete":[0],"diffusion":[1],"language":[2],"models":[3],"(dLLMs)":[4],"generate":[5],"text":[6],"by":[7],"iteratively":[8],"denoising":[9,18,50,78],"a":[10,87,93,107,111,121,130,147],"masked":[11],"sequence.":[12],"However,":[13],"standard":[14],"dLLMs":[15],"condition":[16],"each":[17],"step":[19],"solely":[20],"on":[21,175],"the":[22,40,57,135,140,164],"current":[23],"hard-masked":[24],"sequence,":[25],"while":[26,162],"intermediate":[27,68],"continuous":[28,44],"representations":[29],"are":[30],"discarded":[31],"after":[32],"sampling":[33],"and":[34,52,74,129,167,178],"remasking.":[35],"We":[36,142],"term":[37],"this":[38,82],"bottleneck":[39,60],"\\textbf{Information":[41],"Island}":[42],"issue:":[43],"information":[45,126],"remains":[46],"isolated":[47],"within":[48],"individual":[49],"steps":[51],"fails":[53],"to":[54,71,152],"propagate":[55],"across":[56,76,127,188],"trajectory.":[58],"This":[59],"is":[61],"especially":[62],"harmful":[63],"for":[64],"reasoning,":[65],"which":[66],"requires":[67],"reasoning":[69,170,177],"state":[70],"be":[72],"preserved":[73],"updated":[75,136],"many":[77],"steps.":[79],"To":[80],"address":[81],"limitation,":[83],"we":[84],"introduce":[85],"\\textbf{MetaState},":[86],"lightweight":[88],"recurrent":[89],"augmentation":[90],"that":[91,114,124,133],"equips":[92],"frozen":[94,173],"dLLM":[95],"backbone":[96,116,165],"with":[97,106,146,182],"persistent,":[98],"fixed-size":[99],"working":[100],"memory.":[101],"MetaState":[102,156],"comprises":[103],"three":[104],"modules":[105,145],"shared":[108],"time":[109],"conditioner:":[110],"cross-attention":[112,131],"\\textbf{Mixer}":[113],"reads":[115],"activations":[117],"into":[118,139],"memory":[119,137],"slots,":[120],"GRU-style":[122],"\\textbf{Updater}":[123],"integrates":[125],"steps,":[128],"\\textbf{Injector}":[132],"writes":[134],"back":[138],"backbone.":[141],"train":[143],"these":[144],"dedicated":[148],"$K$-step":[149],"unrolling":[150],"pipeline":[151],"learn":[153],"multi-step":[154],"dynamics.":[155],"adds":[157],"only":[158],"${\\sim}0.6\\%$":[159],"trainable":[160],"parameters":[161],"keeping":[163],"frozen,":[166],"consistently":[168],"improves":[169],"performance":[171],"over":[172],"baselines":[174],"mathematical":[176],"code":[179],"generation":[180],"benchmarks,":[181],"an":[183],"average":[184],"gain":[185],"of":[186],"$4.5\\%$":[187],"all":[189],"evaluations.":[190]},"counts_by_year":[],"updated_date":"2026-04-02T13:48:15.688549","created_date":"2026-03-04T00:00:00"}