{"id":"https://openalex.org/W7163715852","doi":"https://doi.org/10.48550/arxiv.2606.06060","title":"ReCache: Learning Budget-Aware Caching Schedules for Diffusion Models via REINFORCE","display_name":"ReCache: Learning Budget-Aware Caching Schedules for Diffusion Models via REINFORCE","publication_year":2026,"publication_date":"2026-06-04","ids":{"openalex":"https://openalex.org/W7163715852","doi":"https://doi.org/10.48550/arxiv.2606.06060"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.06060","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.06060","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.2606.06060","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5110463679","display_name":"Mishan Aliev","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Aliev, Mishan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120282846","display_name":"Eva Neudachina","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Neudachina, Eva","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029787618","display_name":"Il'ya Bykov","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bykov, Ilya","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137975482","display_name":"Aleksandr Oganov","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Oganov, Aleksandr","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5050892798","display_name":"Kirill Struminsky","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Struminsky, Kirill","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137989595","display_name":"Aibek Alanov","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Alanov, Aibek","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5051726909","display_name":"Denis Rakitin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rakitin, Denis","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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.5040000081062317,"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":0.5040000081062317,"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/T11304","display_name":"Advanced Neuroimaging Techniques and Applications","score":0.07750000059604645,"subfield":{"id":"https://openalex.org/subfields/2741","display_name":"Radiology, Nuclear Medicine and Imaging"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.06970000267028809,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.682699978351593},{"id":"https://openalex.org/keywords/scheduling","display_name":"Scheduling (production processes)","score":0.5644000172615051},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.498199999332428},{"id":"https://openalex.org/keywords/redundancy","display_name":"Redundancy (engineering)","score":0.49399998784065247},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.4916999936103821},{"id":"https://openalex.org/keywords/schedule","display_name":"Schedule","score":0.48399999737739563},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.4345000088214874},{"id":"https://openalex.org/keywords/noise-reduction","display_name":"Noise reduction","score":0.42080000042915344},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.41999998688697815},{"id":"https://openalex.org/keywords/train","display_name":"Train","score":0.4180999994277954}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7791000008583069},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.682699978351593},{"id":"https://openalex.org/C206729178","wikidata":"https://www.wikidata.org/wiki/Q2271896","display_name":"Scheduling (production processes)","level":2,"score":0.5644000172615051},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.498199999332428},{"id":"https://openalex.org/C152124472","wikidata":"https://www.wikidata.org/wiki/Q1204361","display_name":"Redundancy (engineering)","level":2,"score":0.49399998784065247},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.4916999936103821},{"id":"https://openalex.org/C68387754","wikidata":"https://www.wikidata.org/wiki/Q7271585","display_name":"Schedule","level":2,"score":0.48399999737739563},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.4345000088214874},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.42080000042915344},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.41999998688697815},{"id":"https://openalex.org/C190839683","wikidata":"https://www.wikidata.org/wiki/Q2448197","display_name":"Train","level":2,"score":0.4180999994277954},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4163999855518341},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.4108000099658966},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.37070000171661377},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.37049999833106995},{"id":"https://openalex.org/C5465570","wikidata":"https://www.wikidata.org/wiki/Q5326898","display_name":"Early stopping","level":3,"score":0.365200012922287},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.359499990940094},{"id":"https://openalex.org/C143587482","wikidata":"https://www.wikidata.org/wiki/Q1543216","display_name":"Iterative and incremental development","level":2,"score":0.35249999165534973},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.350600004196167},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.3495999872684479},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.3479999899864197},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.32580000162124634},{"id":"https://openalex.org/C155032097","wikidata":"https://www.wikidata.org/wiki/Q798503","display_name":"Backpropagation","level":3,"score":0.31630000472068787},{"id":"https://openalex.org/C52740198","wikidata":"https://www.wikidata.org/wiki/Q1539564","display_name":"Importance sampling","level":3,"score":0.3100000023841858},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3041999936103821},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.29980000853538513},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2896000146865845},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.28949999809265137},{"id":"https://openalex.org/C148483581","wikidata":"https://www.wikidata.org/wiki/Q446488","display_name":"Feature selection","level":2,"score":0.28870001435279846},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.28619998693466187},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.2831000089645386},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.27720001339912415},{"id":"https://openalex.org/C8505890","wikidata":"https://www.wikidata.org/wiki/Q605095","display_name":"Budget constraint","level":2,"score":0.27459999918937683},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.2736000120639801},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.27230000495910645},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.27140000462532043},{"id":"https://openalex.org/C52622490","wikidata":"https://www.wikidata.org/wiki/Q1026626","display_name":"Feature extraction","level":2,"score":0.262800008058548},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.25929999351501465},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.257099986076355},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.2540000081062317}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.06060","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.06060","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.2606.06060","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.06060","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":{"Modern":[0],"diffusion":[1,127],"models":[2],"generate":[3],"high-quality":[4],"images":[5],"and":[6,129,159,217],"videos,":[7],"but":[8],"their":[9],"iterative":[10],"denoising":[11,28,50],"process":[12],"makes":[13],"inference":[14,137,174],"expensive.":[15],"Feature":[16],"caching":[17,46,154],"accelerates":[18],"sampling":[19],"by":[20,191,211,222],"reusing":[21],"or":[22,63],"predicting":[23],"intermediate":[24],"activations":[25],"across":[26,170],"neighboring":[27],"steps,":[29],"exploiting":[30],"the":[31,36,45,73,87,104,219],"redundancy":[32],"of":[33,80],"computations":[34],"along":[35],"reverse":[37],"trajectory.":[38],"In":[39],"this":[40],"work,":[41],"we":[42],"focus":[43],"on":[44,186,200],"schedule:":[47],"selecting":[48],"which":[49,94],"steps":[51],"should":[52],"be":[53],"fully":[54],"recomputed.":[55],"Existing":[56],"schedules":[57],"are":[58],"either":[59],"fixed":[60],"(e.g.":[61],"uniform)":[62],"chosen":[64],"adaptively":[65],"from":[66,135,226],"per-step":[67],"error":[68],"heuristics;":[69],"in":[70],"both":[71],"cases,":[72],"actual":[74],"compute":[75,112],"cost":[76],"is":[77,150,234],"a":[78,85,98,114,144,165,182,204],"side-effect":[79],"hand-tuned":[81],"thresholds":[82],"rather":[83],"than":[84],"quantity":[86],"user":[88],"can":[89],"specify.":[90],"We":[91],"propose":[92],"ReCache,":[93],"inverts":[95],"this:":[96],"given":[97],"target":[99],"budget":[100],"k,":[101],"it":[102,188,208],"learns":[103],"recomputation":[105],"schedule":[106],"that":[107],"maximizes":[108],"generation":[109,147],"quality,":[110],"turning":[111],"into":[113],"directly":[115],"controllable":[116],"input.":[117],"ReCache":[118,149,176],"trains":[119],"via":[120],"policy":[121,168],"gradients,":[122],"sidestepping":[123],"backpropagation":[124],"through":[125],"full":[126],"inference,":[128],"uses":[130],"no":[131],"labelled":[132],"data.":[133],"Generations":[134],"uncached":[136],"serve":[138],"as":[139],"matching":[140],"targets,":[141],"paired":[142],"with":[143,152],"reward":[145],"for":[146,162],"quality.":[148],"compatible":[151],"any":[153],"mechanism,":[155,164],"including":[156],"feature":[157,160],"reuse":[158],"forecasting;":[161],"each":[163],"single":[166],"trained":[167],"adapts":[169],"computational":[171],"budgets":[172],"at":[173,203,236],"time.":[175],"consistently":[177],"outperforms":[178],"scheduling":[179],"baselines:":[180],"under":[181],"$\\times5.04$":[183],"FLOPs":[184],"reduction":[185],"FLUX,":[187],"reduces":[189],"LPIPS":[190,210],"31%":[192],"(from":[193,213],"0.456":[194],"to":[195,198,215,228],"0.316)":[196],"compared":[197],"DiCache;":[199],"Wan":[201],"2.1":[202],"$\\sim":[205],"\\times2.6$":[206],"speedup,":[207],"drops":[209],"65%":[212],"0.480":[214],"0.169)":[216],"boosts":[218],"VBench":[220],"score":[221],"7%":[223],"(5.6":[224],"points,":[225],"70.4":[227],"76.0)":[229],"over":[230],"uniform":[231],"HiCache.":[232],"Code":[233],"available":[235],"https://github.com/thecrazymage/ReCache.":[237]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-06-06T00:00:00"}