{"id":"https://openalex.org/W4415088691","doi":"https://doi.org/10.1145/3746027.3755331","title":"SpeCa: Accelerating Diffusion Transformers with Speculative Feature Caching","display_name":"SpeCa: Accelerating Diffusion Transformers with Speculative Feature Caching","publication_year":2025,"publication_date":"2025-10-25","ids":{"openalex":"https://openalex.org/W4415088691","doi":"https://doi.org/10.1145/3746027.3755331"},"language":"en","primary_location":{"id":"doi:10.1145/3746027.3755331","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3746027.3755331","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 33rd ACM International Conference on Multimedia","raw_type":"proceedings-article"},"type":"preprint","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2509.11628","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Jiacheng Liu","orcid":"https://orcid.org/0009-0002-8759-8641"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Jiacheng Liu","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China and Shandong University, Weihai, China"],"raw_orcid":"https://orcid.org/0009-0002-8759-8641","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China and Shandong University, Weihai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5113844313","display_name":"Chang Zou","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chang Zou","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China and University of Electronic Science and Technology of China, Chengdu, China"],"raw_orcid":"https://orcid.org/0009-0001-2663-2204","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China and University of Electronic Science and Technology of China, Chengdu, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5114188106","display_name":"Yuanhuiyi Lyu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuanhuiyi Lyu","raw_affiliation_strings":["The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China"],"raw_orcid":"https://orcid.org/0009-0004-1450-811X","affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"Fei Ren","orcid":"https://orcid.org/0009-0006-3119-8429"},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Fei Ren","raw_affiliation_strings":["Tsinghua University, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0006-3119-8429","affiliations":[{"raw_affiliation_string":"Tsinghua University, Beijing, China","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5107930049","display_name":"Shaobo Wang","orcid":"https://orcid.org/0000-0001-8156-7081"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shaobo Wang","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0001-8156-7081","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Kaixin Li","orcid":"https://orcid.org/0009-0007-9842-3552"},"institutions":[{"id":"https://openalex.org/I165932596","display_name":"National University of Singapore","ror":"https://ror.org/01tgyzw49","country_code":"SG","type":"education","lineage":["https://openalex.org/I165932596"]}],"countries":["SG"],"is_corresponding":false,"raw_author_name":"Kaixin Li","raw_affiliation_strings":["National University of Singapore, Singapore, Singapore"],"raw_orcid":"https://orcid.org/0009-0007-9842-3552","affiliations":[{"raw_affiliation_string":"National University of Singapore, Singapore, Singapore","institution_ids":["https://openalex.org/I165932596"]}]},{"author_position":"last","author":{"id":null,"display_name":"Linfeng Zhang","orcid":"https://orcid.org/0000-0002-3341-183X"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Linfeng Zhang","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0002-3341-183X","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":7,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I183067930"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.26405522,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"10024","last_page":"10033"},"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":0.9994999766349792,"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":0.9994999766349792,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9923999905586243,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9914000034332275,"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/computation","display_name":"Computation","score":0.6869000196456909},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.6797999739646912},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6549000144004822},{"id":"https://openalex.org/keywords/granularity","display_name":"Granularity","score":0.5198000073432922},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5059999823570251},{"id":"https://openalex.org/keywords/acceleration","display_name":"Acceleration","score":0.4927999973297119},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.3952000141143799},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.3301999866962433}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7784000039100647},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6869000196456909},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6797999739646912},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6549000144004822},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.5198000073432922},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5059999823570251},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.4927999973297119},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4244000017642975},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3952000141143799},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.34540000557899475},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3440000116825104},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.3301999866962433},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.31929999589920044},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.3190999925136566},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3124000132083893},{"id":"https://openalex.org/C78766204","wikidata":"https://www.wikidata.org/wiki/Q555032","display_name":"Multi-core processor","level":2,"score":0.29330000281333923},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.2865999937057495},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.2858000099658966},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.272599995136261},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.2678000032901764},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.265500009059906},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.2635999917984009},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.2621999979019165},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.2540999948978424},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2524000108242035}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1145/3746027.3755331","is_oa":false,"landing_page_url":"https://doi.org/10.1145/3746027.3755331","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 33rd ACM International Conference on Multimedia","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2509.11628","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2509.11628","pdf_url":"https://arxiv.org/pdf/2509.11628","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2509.11628","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2509.11628","pdf_url":"https://arxiv.org/pdf/2509.11628","source":{"id":"https://openalex.org/S4393918464","display_name":"ArXiv.org","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Diffusion":[0],"models":[1,19],"have":[2,197],"revolutionized":[3],"high-fidelity":[4],"image":[5],"and":[6,29,157],"video":[7],"synthesis,":[8],"yet":[9],"their":[10],"computational":[11,108],"demands":[12],"remain":[13],"prohibitive":[14],"for":[15,75,127,134,164,181],"real-time":[16,97],"applications.":[17],"These":[18],"face":[20],"two":[21],"fundamental":[22],"challenges:":[23],"strict":[24],"temporal":[25],"dependencies":[26],"preventing":[27],"parallelization,":[28],"computationally":[30],"intensive":[31,132],"forward":[32],"passes":[33],"required":[34],"at":[35,161,191],"each":[36,103],"denoising":[37],"step.":[38],"Drawing":[39],"inspiration":[40],"from":[41],"speculative":[42],"decoding":[43],"in":[44,65,200],"large":[45],"language":[46],"models,":[47,71],"we":[48],"present":[49],"SpeCa,":[50],"a":[51,87,178],"novel":[52],"'Forecast-then-verify'":[53],"acceleration":[54,140,163,193],"framework":[55],"that":[56,91,116],"effectively":[57],"addresses":[58],"both":[59],"limitations.":[60],"SpeCa's":[61],"core":[62],"innovation":[63],"lies":[64],"introducing":[66],"Speculative":[67],"Sampling":[68],"to":[69,99],"diffusion":[70,183],"predicting":[72],"intermediate":[73],"features":[74],"subsequent":[76],"timesteps":[77],"based":[78,120],"on":[79,121,141,151],"fully":[80],"computed":[81],"reference":[82],"timesteps.":[83],"Our":[84,195],"approach":[85],"implements":[86],"parameter-free":[88],"verification":[89,167],"mechanism":[90,168],"efficiently":[92],"evaluates":[93],"prediction":[94,104],"reliability,":[95],"enabling":[96],"decisions":[98],"accept":[100],"or":[101],"reject":[102],"while":[105,130,153,186],"incurring":[106],"negligible":[107],"overhead.":[109],"Furthermore,":[110],"SpeCa":[111],"introduces":[112],"sample-adaptive":[113],"computation":[114,126],"allocation":[115],"dynamically":[117],"modulates":[118],"resources":[119],"generation":[122,155,188],"complexity,":[123],"allocating":[124],"reduced":[125],"simpler":[128],"samples":[129],"preserving":[131,154],"processing":[133],"complex":[135],"instances.":[136],"Experiments":[137],"demonstrate":[138],"6.34x":[139],"FLUX":[142],"with":[143],"minimal":[144,170],"quality":[145,189],"degradation":[146],"(5.5%":[147],"drop),":[148],"7.3x":[149],"speedup":[150],"DiT":[152],"fidelity,":[156],"79.84%":[158],"VBench":[159],"score":[160],"6.1x":[162],"HunyuanVideo.":[165],"The":[166],"incurs":[169],"overhead":[171],"(1.67%-3.5%":[172],"of":[173],"full":[174],"inference":[175,185],"costs),":[176],"establishing":[177],"new":[179],"paradigm":[180],"efficient":[182],"model":[184],"maintaining":[187],"even":[190],"aggressive":[192],"ratios.":[194],"codes":[196],"been":[198],"released":[199],"Github:":[201],"\\textbf{https://github.com/Shenyi-Z/Cache4Diffusion}":[202]},"counts_by_year":[],"updated_date":"2026-04-29T09:16:38.111599","created_date":"2025-10-12T00:00:00"}