{"id":"https://openalex.org/W7162508617","doi":"https://doi.org/10.48550/arxiv.2605.27075","title":"BeCARE: Budgeted Cache Refresh for Diffusion Transformer Acceleration","display_name":"BeCARE: Budgeted Cache Refresh for Diffusion Transformer Acceleration","publication_year":2026,"publication_date":"2026-05-26","ids":{"openalex":"https://openalex.org/W7162508617","doi":"https://doi.org/10.48550/arxiv.2605.27075"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.27075","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27075","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,"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":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.27075","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137186664","display_name":"Yuhang Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yuhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102990843","display_name":"Junxiang Qiu","orcid":"https://orcid.org/0000-0003-1257-1268"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qiu, Junxiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137188606","display_name":"Huixia Ben","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ben, Huixia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137117152","display_name":"Zhenhua Tang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tang, Zhenhua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5144149231","display_name":"Lin Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Lin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5144107185","display_name":"Shuo Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Shuo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5043525989","display_name":"Yanbin Hao","orcid":"https://orcid.org/0000-0002-0695-1566"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hao, Yanbin","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/T10502","display_name":"Advanced Memory and Neural Computing","score":0.2069000005722046,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10502","display_name":"Advanced Memory and Neural Computing","score":0.2069000005722046,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11019","display_name":"Image Enhancement Techniques","score":0.18790000677108765,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.07880000025033951,"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/transformer","display_name":"Transformer","score":0.5537999868392944},{"id":"https://openalex.org/keywords/control-theory","display_name":"Control theory (sociology)","score":0.5490999817848206},{"id":"https://openalex.org/keywords/acceleration","display_name":"Acceleration","score":0.44780001044273376},{"id":"https://openalex.org/keywords/diffusion-process","display_name":"Diffusion process","score":0.41029998660087585},{"id":"https://openalex.org/keywords/fixed-point","display_name":"Fixed point","score":0.4090000092983246},{"id":"https://openalex.org/keywords/cache","display_name":"Cache","score":0.3889999985694885},{"id":"https://openalex.org/keywords/iterative-and-incremental-development","display_name":"Iterative and incremental development","score":0.3831999897956848}],"concepts":[{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5537999868392944},{"id":"https://openalex.org/C47446073","wikidata":"https://www.wikidata.org/wiki/Q5165890","display_name":"Control theory (sociology)","level":3,"score":0.5490999817848206},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.49779999256134033},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.44780001044273376},{"id":"https://openalex.org/C68710425","wikidata":"https://www.wikidata.org/wiki/Q5275442","display_name":"Diffusion process","level":3,"score":0.41029998660087585},{"id":"https://openalex.org/C61445026","wikidata":"https://www.wikidata.org/wiki/Q217608","display_name":"Fixed point","level":2,"score":0.4090000092983246},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.3889999985694885},{"id":"https://openalex.org/C143587482","wikidata":"https://www.wikidata.org/wiki/Q1543216","display_name":"Iterative and incremental development","level":2,"score":0.3831999897956848},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.3540000021457672},{"id":"https://openalex.org/C3017489831","wikidata":"https://www.wikidata.org/wiki/Q2393193","display_name":"Running time","level":2,"score":0.3244999945163727},{"id":"https://openalex.org/C155386361","wikidata":"https://www.wikidata.org/wiki/Q1649571","display_name":"Process control","level":3,"score":0.323199987411499},{"id":"https://openalex.org/C159694833","wikidata":"https://www.wikidata.org/wiki/Q2321565","display_name":"Iterative method","level":2,"score":0.32100000977516174},{"id":"https://openalex.org/C44154836","wikidata":"https://www.wikidata.org/wiki/Q45045","display_name":"Simulation","level":1,"score":0.3188000023365021},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.295199990272522},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.2840000092983246},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.2766999900341034},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.26660001277923584},{"id":"https://openalex.org/C17500928","wikidata":"https://www.wikidata.org/wiki/Q959968","display_name":"Control system","level":2,"score":0.258899986743927},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.2522999942302704}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.27075","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27075","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,"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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.27075","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27075","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,"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":"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":{"Training-free":[0],"feature":[1],"caching":[2,116,206],"accelerates":[3],"diffusion":[4],"transformer":[5],"(DiT)":[6],"inference":[7],"by":[8,225],"reusing":[9],"or":[10],"forecasting":[11],"intermediate":[12],"features.":[13],"However,":[14],"fixed":[15,152],"schedules":[16],"make":[17],"compute":[18,35],"predictable":[19],"but":[20,32],"ignore":[21],"prompt-":[22],"and":[23,102,131,158,166,176,195],"timestep-dependent":[24],"risk,":[25],"whereas":[26],"hand-tuned":[27],"error":[28],"thresholds":[29],"adapt":[30],"locally":[31],"leave":[33],"realized":[34],"difficult":[36],"to":[37,63,82,105,125,181,212],"control.":[38],"To":[39],"address":[40],"these":[41],"limitations,":[42],"we":[43,71,120],"present":[44],"Budgeted":[45],"Cache":[46],"Refresh":[47],"(BeCARE),":[48],"a":[49,54,112,154,159,178],"training-free":[50,205],"framework":[51],"that,":[52],"given":[53],"user-specified":[55],"cap":[56],"on":[57,190],"full":[58],"computations,":[59],"adaptively":[60],"determines":[61],"when":[62,114],"refresh":[64,113,134,142,174],"the":[65,78,84,122,133,140,145,215,229,243],"cache":[66,103,246],"during":[67],"accelerated":[68],"inference.":[69],"Specifically,":[70],"first":[72],"derive":[73],"an":[74,107,127],"error-amplification":[75],"profile":[76,97,124],"from":[77,210],"sampler's":[79],"noise":[80],"schedule":[81],"characterize":[83],"varying":[85],"impact":[86],"of":[87],"approximation":[88],"errors":[89],"across":[90,183,208],"denoising":[91],"timesteps.":[92],"We":[93],"then":[94],"combine":[95],"this":[96],"with":[98,192],"prompt-specific":[99],"extrapolation":[100,165],"residuals":[101],"age":[104],"form":[106],"accumulated":[108],"risk":[109],"score,":[110],"triggering":[111],"continued":[115],"becomes":[117],"harmful.":[118],"Meanwhile,":[119],"use":[121],"same":[123,244],"construct":[126],"analytic":[128],"spending":[129],"reference":[130],"adjust":[132],"threshold":[135],"through":[136],"feedback,":[137],"thereby":[138],"distributing":[139],"limited":[141],"budget":[143,168,184],"over":[144,228,248],"sampling":[146],"trajectory.":[147],"Finally,":[148],"structural":[149],"safeguards":[150],"(a":[151],"warmup,":[153],"budget-derived":[155],"late-stage":[156],"reserve,":[157],"maximum":[160],"cache-reuse":[161],"length)":[162],"prevent":[163],"unreliable":[164],"premature":[167],"exhaustion.":[169],"These":[170],"designs":[171],"enable":[172],"prompt-adaptive":[173],"placement":[175],"allow":[177],"single":[179],"calibration":[180],"transfer":[182],"tiers":[185],"without":[186],"per-tier":[187],"tuning.":[188],"Experiments":[189],"FLUX.1-dev":[191],"200":[193],"prompts":[194],"measured":[196],"FLOPs":[197],"show":[198],"that":[199],"our":[200],"method":[201],"consistently":[202],"outperforms":[203],"representative":[204],"baselines":[207],"speedups":[209],"3x":[211],"6x.":[213],"At":[214],"main":[216],"operating":[217],"point":[218],"(3.3x":[219],"acceleration),":[220],"it":[221,239],"improves":[222,242],"paired":[223],"PSNR":[224],"2.7":[226],"dB":[227],"strongest":[230],"baseline":[231],"using":[232],"no":[233],"more":[234],"compute.":[235,253],"On":[236],"SD3.5":[237],"Large,":[238],"also":[240],"substantially":[241],"Taylor":[245],"engine":[247],"fixed-interval":[249],"scheduling":[250],"at":[251],"matched":[252]},"counts_by_year":[],"updated_date":"2026-08-01T09:00:35.917206","created_date":"2026-05-28T00:00:00"}
