{"id":"https://openalex.org/W4413146270","doi":"https://doi.org/10.1109/cvpr52734.2025.01464","title":"Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models","display_name":"Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models","publication_year":2025,"publication_date":"2025-06-10","ids":{"openalex":"https://openalex.org/W4413146270","doi":"https://doi.org/10.1109/cvpr52734.2025.01464"},"language":"en","primary_location":{"id":"doi:10.1109/cvpr52734.2025.01464","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cvpr52734.2025.01464","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"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/A5110981770","display_name":"Jingfeng Yao","orcid":null},"institutions":[{"id":"https://openalex.org/I47720641","display_name":"Huazhong University of Science and Technology","ror":"https://ror.org/00p991c53","country_code":"CN","type":"education","lineage":["https://openalex.org/I47720641"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jingfeng Yao","raw_affiliation_strings":["Huazhong University of Science and Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huazhong University of Science and Technology","institution_ids":["https://openalex.org/I47720641"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5025581143","display_name":"Bin Yang","orcid":"https://orcid.org/0000-0003-4486-7143"},"institutions":[{"id":"https://openalex.org/I47720641","display_name":"Huazhong University of Science and Technology","ror":"https://ror.org/00p991c53","country_code":"CN","type":"education","lineage":["https://openalex.org/I47720641"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Bin Yang","raw_affiliation_strings":["Huazhong University of Science and Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huazhong University of Science and Technology","institution_ids":["https://openalex.org/I47720641"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5115695302","display_name":"Xinggang Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I47720641","display_name":"Huazhong University of Science and Technology","ror":"https://ror.org/00p991c53","country_code":"CN","type":"education","lineage":["https://openalex.org/I47720641"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xinggang Wang","raw_affiliation_strings":["Huazhong University of Science and Technology"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Huazhong University of Science and Technology","institution_ids":["https://openalex.org/I47720641"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I47720641"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":23,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"15703","last_page":"15712"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11986","display_name":"Scientific Computing and Data Management","score":0.7289999723434448,"subfield":{"id":"https://openalex.org/subfields/1802","display_name":"Information Systems and Management"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T11986","display_name":"Scientific Computing and Data Management","score":0.7289999723434448,"subfield":{"id":"https://openalex.org/subfields/1802","display_name":"Information Systems and Management"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10317","display_name":"Advanced Database Systems and Queries","score":0.6291999816894531,"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/dilemma","display_name":"Dilemma","score":0.688725471496582},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.5984506011009216},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.33127903938293457},{"id":"https://openalex.org/keywords/mathematics","display_name":"Mathematics","score":0.13353323936462402}],"concepts":[{"id":"https://openalex.org/C2778496695","wikidata":"https://www.wikidata.org/wiki/Q254128","display_name":"Dilemma","level":2,"score":0.688725471496582},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5984506011009216},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.33127903938293457},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.13353323936462402},{"id":"https://openalex.org/C2524010","wikidata":"https://www.wikidata.org/wiki/Q8087","display_name":"Geometry","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/cvpr52734.2025.01464","is_oa":false,"landing_page_url":"https://doi.org/10.1109/cvpr52734.2025.01464","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2350209916","https://openalex.org/W2475524763","https://openalex.org/W2363851383","https://openalex.org/W2354517252","https://openalex.org/W2387460998","https://openalex.org/W2356631479","https://openalex.org/W2386467653"],"abstract_inverted_index":{"Latent":[0],"diffusion":[1,38,128],"models":[2,39,106],"with":[3,102,154,174],"Transformer":[4],"architectures":[5],"excel":[6],"at":[7,213],"generating":[8],"high-fidelity":[9],"images.":[10],"However,":[11],"recent":[12],"studies":[13],"reveal":[14],"an":[15,150,175,187,198],"optimization":[16],"dilemma":[17,81],"in":[18,28,87,137,192],"this":[19,80],"two-stage":[20],"design:":[21],"while":[22,180],"increasing":[23],"the":[24,84,99,109,123,143,205],"per-token":[25],"feature":[26],"dimension":[27],"visual":[29,59,110],"tokenizers":[30,66],"improves":[31],"reconstruction":[32],"quality,":[33],"it":[34],"requires":[35],"substantially":[36],"larger":[37],"and":[40,158,209],"more":[41],"training":[42,108,156,183],"iterations":[43],"to":[44,62,69,73,204],"achieve":[45],"comparable":[46],"generation":[47,173],"performance.":[48],"Consequently,":[49],"existing":[50],"systems":[51],"often":[52],"settle":[53],"for":[54],"suboptimal":[55],"solutions,":[56],"either":[57],"producing":[58],"artifacts":[60],"due":[61,72],"information":[63],"loss":[64],"within":[65],"or":[67],"failing":[68],"converge":[70],"fully":[71],"expensive":[74],"computation":[75],"costs.":[76],"We":[77],"argue":[78],"that":[79],"stems":[82],"from":[83],"inherent":[85],"difficulty":[86],"learning":[88],"unconstrained":[89],"high-dimensional":[90,138],"latent":[91,100,127,139],"spaces.":[92,140],"To":[93,141],"address":[94],"this,":[95],"we":[96,148],"propose":[97],"aligning":[98],"space":[101],"pretrained":[103],"vision":[104],"foundation":[105,116],"when":[107],"tokenizers.":[111],"Our":[112],"proposed":[113],"VA-VAE":[114],"(Vision":[115],"model":[117],"Aligned":[118],"Variational":[119],"AutoEncoder)":[120],"significantly":[121],"expands":[122],"reconstruction-generation":[124],"frontier":[125],"of":[126,133,146,178,190],"models,":[129],"enabling":[130],"faster":[131],"convergence":[132,201],"Diffusion":[134],"Transformers":[135],"(DiT)":[136],"exploit":[142],"full":[144],"potential":[145],"VA-VAE,":[147],"build":[149],"enhanced":[151],"DiT":[152],"baseline":[153],"improved":[155],"strategies":[157],"architecture":[159],"designs,":[160],"termed":[161],"LightningDiT.":[162],"The":[163],"integrated":[164],"system":[165],"achieves":[166],"state-of-the-art":[167],"(SOTA)":[168],"performance":[169],"on":[170],"ImageNet":[171],"256\u00d7256":[172],"FID":[176,188],"score":[177,189],"1.35":[179],"demonstrating":[181],"remarkable":[182],"efficiency":[184],"by":[185],"reaching":[186],"2.11":[191],"just":[193],"64":[194],"epochs":[195],"\u2013":[196],"representing":[197],"over":[199],"21\u00d7":[200],"speedup":[202],"compared":[203],"original":[206],"DiT.":[207],"Models":[208],"codes":[210],"are":[211],"available":[212],"https://github.com/hustvl/LightningDiT.":[214]},"counts_by_year":[{"year":2026,"cited_by_count":8},{"year":2025,"cited_by_count":15}],"updated_date":"2026-07-17T09:13:05.818461","created_date":"2025-10-10T00:00:00"}