{"id":"https://openalex.org/W7163511780","doi":"https://doi.org/10.1109/tip.2026.3698363","title":"Rethinking Token-Wise Feature Caching: Accelerating Diffusion Transformers With Dual Feature Caching","display_name":"Rethinking Token-Wise Feature Caching: Accelerating Diffusion Transformers With Dual Feature Caching","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7163511780","doi":"https://doi.org/10.1109/tip.2026.3698363","pmid":"https://pubmed.ncbi.nlm.nih.gov/42241263"},"language":"en","primary_location":{"id":"doi:10.1109/tip.2026.3698363","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tip.2026.3698363","pdf_url":null,"source":{"id":"https://openalex.org/S4210173141","display_name":"IEEE Transactions on Image Processing","issn_l":"1057-7149","issn":["1057-7149","1941-0042"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Image Processing","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","pubmed"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5129704192","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"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101323660","display_name":"Shikang Zheng","orcid":"https://orcid.org/0009-0003-9578-1937"},"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":"Shikang Zheng","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0009-0009-1999-4619","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129651924","display_name":"Evelyn Zhang","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":"Evelyn Zhang","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137853962","display_name":"Runlin Guo","orcid":null},"institutions":[{"id":"https://openalex.org/I82880672","display_name":"Beihang University","ror":"https://ror.org/00wk2mp56","country_code":"CN","type":"education","lineage":["https://openalex.org/I82880672"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Runlin Guo","raw_affiliation_strings":["Beihang University, Beijing, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Beihang University, Beijing, China","institution_ids":["https://openalex.org/I82880672"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013932458","display_name":"Haohang Xu","orcid":"https://orcid.org/0000-0002-4715-1338"},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Haohang Xu","raw_affiliation_strings":["Huawei Technologies Ltd., Shenzhen, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huawei Technologies Ltd., Shenzhen, China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137832461","display_name":"Zhengyi Shi","orcid":null},"institutions":[{"id":"https://openalex.org/I191208505","display_name":"Xiamen University","ror":"https://ror.org/00mcjh785","country_code":"CN","type":"education","lineage":["https://openalex.org/I191208505"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhengyi Shi","raw_affiliation_strings":["Xiamen University, Xiamen, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Xiamen University, Xiamen, China","institution_ids":["https://openalex.org/I191208505"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101615091","display_name":"Conghui He","orcid":"https://orcid.org/0000-0001-8697-695X"},"institutions":[{"id":"https://openalex.org/I4391012619","display_name":"Shanghai Artificial Intelligence Laboratory","ror":"https://ror.org/03wkvpx79","country_code":"CN","type":"facility","lineage":["https://openalex.org/I4391012619"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Conghui He","raw_affiliation_strings":["Shanghai AI Laboratory, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0001-8697-695X","affiliations":[{"raw_affiliation_string":"Shanghai AI Laboratory, Shanghai, China","institution_ids":["https://openalex.org/I4391012619"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137858368","display_name":"Xuming Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xuming Hu","raw_affiliation_strings":["The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100689117","display_name":"Linfeng Zhang","orcid":"https://orcid.org/0000-0002-8470-5846"},"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":1,"institutions_distinct_count":5,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.69161657,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"35","issue":null,"first_page":"6211","last_page":"6220"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11478","display_name":"Caching and Content Delivery","score":0.9419000148773193,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T11478","display_name":"Caching and Content Delivery","score":0.9419000148773193,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T11181","display_name":"Advanced Data Storage Technologies","score":0.012500000186264515,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T10772","display_name":"Distributed systems and fault tolerance","score":0.004699999932199717,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/feature","display_name":"Feature (linguistics)","score":0.5185999870300293},{"id":"https://openalex.org/keywords/feature-extraction","display_name":"Feature extraction","score":0.43369999527931213},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.4277999997138977},{"id":"https://openalex.org/keywords/image-processing","display_name":"Image processing","score":0.326200008392334},{"id":"https://openalex.org/keywords/feature-vector","display_name":"Feature vector","score":0.3249000012874603},{"id":"https://openalex.org/keywords/signal-processing","display_name":"Signal processing","score":0.26820001006126404}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6901000142097473},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5185999870300293},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5102999806404114},{"id":"https://openalex.org/C52622490","wikidata":"https://www.wikidata.org/wiki/Q1026626","display_name":"Feature extraction","level":2,"score":0.43369999527931213},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.4277999997138977},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.37630000710487366},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3262999951839447},{"id":"https://openalex.org/C9417928","wikidata":"https://www.wikidata.org/wiki/Q1070689","display_name":"Image processing","level":3,"score":0.326200008392334},{"id":"https://openalex.org/C83665646","wikidata":"https://www.wikidata.org/wiki/Q42139305","display_name":"Feature vector","level":2,"score":0.3249000012874603},{"id":"https://openalex.org/C104267543","wikidata":"https://www.wikidata.org/wiki/Q208163","display_name":"Signal processing","level":3,"score":0.26820001006126404},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.26589998602867126},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.2540999948978424},{"id":"https://openalex.org/C106516650","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm design","level":2,"score":0.25270000100135803}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1109/tip.2026.3698363","is_oa":false,"landing_page_url":"https://doi.org/10.1109/tip.2026.3698363","pdf_url":null,"source":{"id":"https://openalex.org/S4210173141","display_name":"IEEE Transactions on Image Processing","issn_l":"1057-7149","issn":["1057-7149","1941-0042"],"is_oa":false,"is_in_doaj":false,"is_core":true,"host_organization":"https://openalex.org/P4310319808","host_organization_name":"Institute of Electrical and Electronics Engineers","host_organization_lineage":["https://openalex.org/P4310319808"],"host_organization_lineage_names":["Institute of Electrical and Electronics Engineers"],"type":"journal"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE Transactions on Image Processing","raw_type":"journal-article"},{"id":"pmid:42241263","is_oa":false,"landing_page_url":"https://pubmed.ncbi.nlm.nih.gov/42241263","pdf_url":null,"source":{"id":"https://openalex.org/S4306525036","display_name":"PubMed","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I1299303238","host_organization_name":"National Institutes of Health","host_organization_lineage":["https://openalex.org/I1299303238"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"IEEE transactions on image processing : a publication of the IEEE Signal Processing Society","raw_type":null}],"best_oa_location":null,"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":[0],"Transformers":[1],"(DiT)":[2],"have":[3],"become":[4],"the":[5,33,44,51,54,78,86,97,104,115,140,153,195,204,219],"dominant":[6],"methods":[7,28],"in":[8,37,43,53,73,99,119,144,209],"image":[9],"and":[10,40,160,188,193,213],"video":[11],"generation":[12],"yet":[13],"still":[14,84],"suffer":[15],"substantial":[16],"computational":[17],"costs.":[18],"As":[19],"an":[20],"effective":[21],"approach":[22],"for":[23,70,80,197],"DiT":[24,36],"acceleration,":[25],"feature":[26,60,101,177,222],"caching":[27,61,68,102,178,186,190],"are":[29],"designed":[30],"to":[31,49,65,76,94,113,180],"cache":[32],"features":[34],"of":[35,152,206],"previous":[38,220],"timesteps":[39],"reuse":[41],"them":[42],"next":[45,55],"timesteps,":[46],"allowing":[47],"us":[48],"skip":[50,77],"computation":[52,79],"timesteps.":[56],"Among":[57],"them,":[58],"token-wise":[59,100,221],"has":[62],"been":[63],"introduced":[64],"perform":[66],"different":[67,71],"ratios":[69],"tokens":[72,82,118,126,196],"DiTs,":[74],"aiming":[75],"unimportant":[81],"while":[83],"computing":[85,139,198],"important":[87,125],"ones.":[88],"In":[89],"this":[90,130,173],"paper,":[91],"we":[92],"propose":[93],"carefully":[95],"check":[96],"effectiveness":[98,205],"with":[103],"following":[105],"two":[106],"questions:":[107],"1)":[108],"Is":[109],"it":[110],"really":[111,127],"necessary":[112],"compute":[114],"so-called":[116,124,154],"\"important\"":[117],"each":[120],"step?":[121],"2)":[122],"Are":[123],"important?":[128],"Surprisingly,":[129],"paper":[131,174],"gives":[132],"some":[133],"counter-intuition":[134],"answers,":[135],"demonstrating":[136,215],"that":[137],"consistently":[138],"selected":[141],"\"important":[142,155],"tokens\"":[143,156],"all":[145],"steps":[146],"is":[147,157],"not":[148],"necessary.":[149],"The":[150],"selection":[151],"often":[158],"ineffective,":[159],"even":[161],"sometimes":[162],"shows":[163],"inferior":[164],"performance":[165],"than":[166,218],"random":[167],"selection.":[168],"Based":[169],"on":[170],"these":[171],"observations,":[172],"introduces":[175],"dual":[176],"referred":[179],"as":[181],"DuCa,":[182],"which":[183],"performs":[184],"aggressive":[185],"strategy":[187,191],"conservative":[189],"iteratively":[192],"selects":[194],"randomly.":[199],"Extensive":[200],"experimental":[201],"results":[202],"demonstrate":[203],"our":[207],"method":[208],"DiT,":[210],"PixArt,":[211],"FLUX,":[212],"OpenSora,":[214],"significant":[216],"improvements":[217],"caching.":[223]},"counts_by_year":[],"updated_date":"2026-07-22T07:51:19.307946","created_date":"2026-06-05T00:00:00"}