{"id":"https://openalex.org/W7214969512","doi":"https://doi.org/10.48550/arxiv.2609.37924","title":"Time-Anchored Diffusion Language Models: Latent-Space Caching for Fast Generation","display_name":"Time-Anchored Diffusion Language Models: Latent-Space Caching for Fast Generation","publication_year":2026,"publication_date":"2026-09-29","ids":{"openalex":"https://openalex.org/W7214969512","doi":"https://doi.org/10.48550/arxiv.2609.37924"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.37924","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37924","pdf_url":"https://arxiv.org/pdf/2609.37924","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.37924","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5153603604","display_name":"Joel Anto Paul","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Paul, Joel Anto","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052553707","display_name":"Litu Rout","orcid":"https://orcid.org/0000-0002-5054-5899","observed_orcids":["https://orcid.org/0000-0002-5054-5899"]},"institutions":[{"id":"https://openalex.org/I86519309","display_name":"The University of Texas at Austin","ror":"https://ror.org/00hj54h04","country_code":"US","type":"education","lineage":["https://openalex.org/I86519309"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Rout, Litu","raw_affiliation_strings":["UT Austin"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UT Austin","institution_ids":["https://openalex.org/I86519309"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035329776","display_name":"Aditya Akella","orcid":"https://orcid.org/0000-0002-5920-170X","observed_orcids":["https://orcid.org/0000-0002-5920-170X"]},"institutions":[{"id":"https://openalex.org/I86519309","display_name":"The University of Texas at Austin","ror":"https://ror.org/00hj54h04","country_code":"US","type":"education","lineage":["https://openalex.org/I86519309"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Akella, Aditya","raw_affiliation_strings":["UT Austin"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UT Austin","institution_ids":["https://openalex.org/I86519309"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5028903768","display_name":"Sanjay Shakkottai","orcid":"https://orcid.org/0000-0002-4325-9050","observed_orcids":["https://orcid.org/0000-0002-4325-9050"]},"institutions":[{"id":"https://openalex.org/I86519309","display_name":"The University of Texas at Austin","ror":"https://ror.org/00hj54h04","country_code":"US","type":"education","lineage":["https://openalex.org/I86519309"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Shakkottai, Sanjay","raw_affiliation_strings":["UT Austin"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UT Austin","institution_ids":["https://openalex.org/I86519309"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"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/T10028","display_name":"Topic Modeling","score":0.7026000022888184,"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":0.7026000022888184,"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":0.19020000100135803,"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":0.04830000177025795,"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/diffusion-denoising","display_name":"diffusion denoising","score":0.6809999942779541},{"id":"https://openalex.org/keywords/fusion-module","display_name":"fusion module","score":0.6759999990463257}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8317999839782715},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.609499990940094},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5565000176429749},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.5562999844551086},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5109999775886536},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.5044999718666077},{"id":"https://openalex.org/C48103436","wikidata":"https://www.wikidata.org/wiki/Q599031","display_name":"State (computer science)","level":2,"score":0.49059998989105225},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.4828999936580658},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.45649999380111694},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4156000018119812},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.4002000093460083},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3978999853134155},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.39329999685287476},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.3474000096321106},{"id":"https://openalex.org/C72434380","wikidata":"https://www.wikidata.org/wiki/Q230930","display_name":"State space","level":2,"score":0.3398999869823456},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.33570000529289246},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.3278999924659729},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.323199987411499},{"id":"https://openalex.org/C193415008","wikidata":"https://www.wikidata.org/wiki/Q639681","display_name":"Network architecture","level":2,"score":0.30149999260902405},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.2906999886035919},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2709999978542328}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.37924","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37924","pdf_url":"https://arxiv.org/pdf/2609.37924","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.37924","doi":"https://doi.org/10.48550/arxiv.2609.37924","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.37924","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:2609.37924","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.37924","pdf_url":"https://arxiv.org/pdf/2609.37924","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[{"id":"https://openalex.org/G2186497367","display_name":"AI Institute for Future Edge Networks and Distributed Intelligence (AI-EDGE)","funder_award_id":"2112471","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G2721915658","display_name":"National Science Foundation Expeditions in Computing: Learning Directed Operating System -- A Clean-Slate Paradigm for Operating Systems Design and Implementation","funder_award_id":"2326576","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G853533796","display_name":"Institute for Foundations of Machine Learning","funder_award_id":"2505865","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320310620","display_name":"University of Texas at Austin","ror":"https://ror.org/00hj54h04"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214969512.pdf","grobid_xml":"https://content.openalex.org/works/W7214969512.grobid-xml"},"referenced_works_count":9,"referenced_works":["https://openalex.org/W3210904070","https://openalex.org/W4392822190","https://openalex.org/W4407679583","https://openalex.org/W4414939355","https://openalex.org/W4415081427","https://openalex.org/W4415796274","https://openalex.org/W4416051103","https://openalex.org/W7160095038","https://openalex.org/W7172402159"],"related_works":["https://openalex.org/W4414581322","https://openalex.org/W7204099981","https://openalex.org/W7128914173","https://openalex.org/W7166879454","https://openalex.org/W7106708636","https://openalex.org/W7202215318","https://openalex.org/W7221835644","https://openalex.org/W7127100793","https://openalex.org/W4415037641","https://openalex.org/W7163024699"],"abstract_inverted_index":{"Recent":[0],"work":[1],"on":[2,176],"anchored":[3],"diffusion":[4,78],"language":[5],"models":[6],"improves":[7,169],"denoising":[8,103],"by":[9,171,193],"shaping":[10],"an":[11],"intermediate":[12,58],"latent":[13,31,97,110],"space":[14],"with":[15,112],"supervised":[16],"important-token":[17],"targets.":[18,36],"In":[19],"this":[20,149],"work,":[21],"we":[22],"introduce":[23],"time-based":[24,161],"(self-supervised)":[25],"anchoring,":[26],"which":[27,153,159],"learns":[28,160],"and":[29,100,157,180],"reuses":[30],"anchors":[32,42,162],"without":[33],"requiring":[34],"such":[35,50],"Our":[37],"key":[38],"observation":[39],"is":[40,81,134,141],"that":[41,94,105],"encode":[43],"persistent":[44],"properties":[45],"of":[46,88],"the":[47,67,96,108,113,131],"clean":[48],"sequence,":[49],"as":[51,66,151],"its":[52,138],"semantic":[53,72],"intent,":[54],"global":[55],"structure,":[56],"or":[57],"plan.":[59],"Although":[60],"their":[61,71],"hidden":[62],"representations":[63],"become":[64],"stale":[65],"token":[68],"canvas":[69],"evolves,":[70],"content":[73],"remains":[74],"useful":[75],"across":[76,143],"nearby":[77],"times.":[79],"This":[80,124],"implemented":[82],"through":[83],"a":[84,89,101,121,127,199],"two-stage":[85],"architecture":[86],"consisting":[87],"relatively":[90],"expensive":[91],"anchor":[92,132],"network":[93,104,133],"generates":[95],"cache":[98],"state":[99,111,115],"lightweight":[102],"intelligently":[106],"combines":[107],"cached":[109,139],"current":[114],"at":[116],"each":[117],"reverse":[118,145],"step":[119],"using":[120],"fusion":[122],"module.":[123],"gives":[125],"anchoring":[126],"latent-space":[128],"caching":[129],"interpretation:":[130],"evaluated":[135],"periodically,":[136],"while":[137],"representation":[140],"reused":[142],"multiple":[144],"steps.":[146],"We":[147],"instantiate":[148],"framework":[150],"TADM:Post-train,":[152],"time-anchorizes":[154],"pretrained":[155],"DLMs,":[156],"TADM:Pretraining,":[158],"during":[163],"pretraining.":[164],"Applied":[165],"to":[166,174,195,198,205],"DiffusionGemma-26B,":[167],"TADM:Post-train":[168],"throughput":[170,209],"approximately":[172],"49%":[173],"79%":[175],"several":[177],"math,":[178],"code,":[179],"STEM":[181],"benchmarks":[182],"(GSM8K,":[183],"AIME26,":[184],"GPQA-Diamond,":[185],"LiveCodeBench-v6,":[186],"HumanEval,":[187],"MMLU-Pro).":[188],"TADM:Pretraining":[189],"reduces":[190],"Transformer-layer":[191],"computation":[192],"up":[194,204],"38%":[196],"relative":[197],"standard":[200],"single-stage":[201],"DLM,":[202],"achieves":[203],"73%":[206],"higher":[207],"measured":[208],"than":[210],"ADLM.":[211]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-01T00:00:00"}