{"id":"https://openalex.org/W7160489755","doi":"https://doi.org/10.48550/arxiv.2605.04215","title":"Predict-then-Diffuse: Adaptive Response Length for Compute-Budgeted Inference in Diffusion LLMs","display_name":"Predict-then-Diffuse: Adaptive Response Length for Compute-Budgeted Inference in Diffusion LLMs","publication_year":2026,"publication_date":"2026-05-05","ids":{"openalex":"https://openalex.org/W7160489755","doi":"https://doi.org/10.48550/arxiv.2605.04215"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.04215","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.04215","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.04215","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135560816","display_name":"Michael Rottoli","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Rottoli, Michael","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037398393","display_name":"Subhankar Roy","orcid":"https://orcid.org/0009-0008-2395-8111"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Roy, Subhankar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5026292704","display_name":"Stefano Paraboschi","orcid":"https://orcid.org/0000-0003-0399-1738"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Paraboschi, Stefano","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5135560816"],"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.26429998874664307,"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"}},"topics":[{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.26429998874664307,"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/T10028","display_name":"Topic Modeling","score":0.12219999730587006,"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/T14347","display_name":"Big Data and Digital Economy","score":0.07460000365972519,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/inference","display_name":"Inference","score":0.7488999962806702},{"id":"https://openalex.org/keywords/padding","display_name":"Padding","score":0.6945000290870667},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5182999968528748},{"id":"https://openalex.org/keywords/truncation","display_name":"Truncation (statistics)","score":0.5019999742507935},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4593000113964081},{"id":"https://openalex.org/keywords/throughput","display_name":"Throughput","score":0.45570001006126404},{"id":"https://openalex.org/keywords/response-time","display_name":"Response time","score":0.4375}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.7488999962806702},{"id":"https://openalex.org/C165435473","wikidata":"https://www.wikidata.org/wiki/Q1509884","display_name":"Padding","level":2,"score":0.6945000290870667},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6689000129699707},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5182999968528748},{"id":"https://openalex.org/C106195933","wikidata":"https://www.wikidata.org/wiki/Q7847935","display_name":"Truncation (statistics)","level":2,"score":0.5019999742507935},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4593000113964081},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.45570001006126404},{"id":"https://openalex.org/C19012869","wikidata":"https://www.wikidata.org/wiki/Q578372","display_name":"Response time","level":2,"score":0.4375},{"id":"https://openalex.org/C82876162","wikidata":"https://www.wikidata.org/wiki/Q17096504","display_name":"Latency (audio)","level":2,"score":0.4162999987602234},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3935000002384186},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.364300012588501},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.36000001430511475},{"id":"https://openalex.org/C32407928","wikidata":"https://www.wikidata.org/wiki/Q2733833","display_name":"Best response","level":3,"score":0.29170000553131104},{"id":"https://openalex.org/C2779438525","wikidata":"https://www.wikidata.org/wiki/Q5255048","display_name":"Demand response","level":3,"score":0.2791000008583069},{"id":"https://openalex.org/C2779793024","wikidata":"https://www.wikidata.org/wiki/Q17299941","display_name":"Indirect Inference","level":3,"score":0.26750001311302185},{"id":"https://openalex.org/C158600405","wikidata":"https://www.wikidata.org/wiki/Q5054566","display_name":"Causal inference","level":2,"score":0.26409998536109924},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.257099986076355},{"id":"https://openalex.org/C185429906","wikidata":"https://www.wikidata.org/wiki/Q1130160","display_name":"Estimator","level":2,"score":0.25699999928474426}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.04215","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.04215","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.04215","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.04215","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":"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":{"Diffusion-based":[0],"Large":[1],"Language":[2],"Models":[3],"(D-LLMs)":[4],"represent":[5],"a":[6,42,53,90,138,149,154,160,168],"promising":[7],"frontier":[8],"in":[9,60],"generative":[10],"AI,":[11],"offering":[12],"fully":[13],"parallel":[14],"token":[15],"generation":[16],"that":[17,78,95,191],"can":[18],"lead":[19],"to":[20,47,112,199,208],"significant":[21],"throughput":[22],"advantages":[23],"and":[24,92,108,145],"superior":[25],"GPU":[26],"utilization":[27],"over":[28],"the":[29,39,105,129,142,164,179,200],"traditional":[30],"autoregressive":[31],"paradigm.":[32],"However,":[33],"this":[34,85],"parallelism":[35],"is":[36],"constrained":[37],"by":[38,102],"requirement":[40],"of":[41,163],"fixed-size":[43],"response":[44,57,70,106,131,143],"length":[45,58,71,107,132,144],"prior":[46],"generation.":[48],"This":[49],"architectural":[50],"limitation":[51],"imposes":[52],"severe":[54],"trade-off:":[55],"oversized":[56],"results":[59],"computational":[61,195],"waste":[62],"on":[63,159,175,187],"semantically":[64],"meaningless":[65],"padding":[66,176],"tokens,":[67,177],"while":[68,205],"undersized":[69],"causes":[72],"output":[73,183],"truncation":[74],"requiring":[75],"costly":[76],"re-computations":[77],"introduce":[79,153],"unpredictable":[80],"latency":[81],"spikes.":[82],"To":[83],"tackle":[84],"issue,":[86],"we":[87,152],"propose":[88],"Predict-then-Diffuse,":[89],"simple":[91],"model-agnostic":[93],"framework":[94,171],"enables":[96],"compute-budgeted":[97],"inference":[98,114,147,203],"per":[99],"input":[100,135],"query":[101],"first":[103],"estimating":[104],"then":[109],"using":[110],"it":[111],"run":[113],"with":[115,148],"D-LLM.":[116],"At":[117],"its":[118],"core":[119],"lies":[120],"an":[121,134],"Adaptive":[122],"Response":[123],"Length":[124],"Predictor":[125],"(AdaRLP),":[126],"which":[127],"estimates":[128],"optimal":[130],"given":[133],"query.":[136],"As":[137,167],"measure":[139],"against":[140],"under-estimating":[141],"re-running":[146],"higher":[150],"value,":[151],"data-driven":[155],"safety":[156],"mechanism":[157],"based":[158],"small":[161],"increase":[162],"predicted":[165],"length.":[166],"whole,":[169],"our":[170],"avoids":[172],"wasting":[173],"computation":[174],"at":[178],"same":[180],"time":[181],"preserving":[182],"quality.":[184],"Experimental":[185],"validation":[186],"multiple":[188],"datasets":[189],"demonstrates":[190],"Predict-then-Diffuse":[192],"significantly":[193],"reduces":[194],"costs":[196],"(FLOP)":[197],"compared":[198],"default":[201],"D-LLM":[202],"mechanism,":[204],"being":[206],"robust":[207],"skewed":[209],"data":[210],"distributions.":[211]},"counts_by_year":[],"updated_date":"2026-05-16T06:04:12.930555","created_date":"2026-05-08T00:00:00"}