{"id":"https://openalex.org/W7162145380","doi":"https://doi.org/10.48550/arxiv.2605.21573","title":"Lens: Rethinking Training Efficiency for Foundational Text-to-Image Models","display_name":"Lens: Rethinking Training Efficiency for Foundational Text-to-Image Models","publication_year":2026,"publication_date":"2026-05-20","ids":{"openalex":"https://openalex.org/W7162145380","doi":"https://doi.org/10.48550/arxiv.2605.21573"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.21573","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21573","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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.2605.21573","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136812891","display_name":"Dong Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Dong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090973869","display_name":"Fangyun Wei","orcid":"https://orcid.org/0000-0001-8784-4916"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Fangyun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136805980","display_name":"Ziyu Wan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wan, Ziyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136739117","display_name":"Dongdong Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Dongdong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136770380","display_name":"Jiawei Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Jiawei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136732895","display_name":"Jinjing Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Jinjing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136806612","display_name":"Sirui Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Sirui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136801775","display_name":"Yang Yue","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yue, Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132720841","display_name":"Zhiyang Liang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Zhiyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136743344","display_name":"Baining Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Baining","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136736817","display_name":"Chong Luo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Luo, Chong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136756162","display_name":"Jianmin Bao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bao, Jianmin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136749225","display_name":"Ji Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Ji","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136806108","display_name":"Lei Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Lei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129750622","display_name":"Qinhong Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Qinhong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090680444","display_name":"Xiuyu Wu","orcid":"https://orcid.org/0000-0003-0201-0379"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Xiuyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136777946","display_name":"Xuelu Feng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Feng, Xuelu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136729684","display_name":"Yan Lu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu, Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039574432","display_name":"Yanchen Dong","orcid":"https://orcid.org/0000-0001-8011-7193"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dong, Yanchen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136761598","display_name":"Yitong Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yitong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136769495","display_name":"Yunuo Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Yunuo","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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.5885000228881836,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.5885000228881836,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.2840999960899353,"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":0.03590000048279762,"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/key","display_name":"Key (lock)","score":0.6365000009536743},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.5354999899864197},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.42570000886917114},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.39910000562667847},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.37070000171661377},{"id":"https://openalex.org/keywords/lens","display_name":"Lens (geology)","score":0.3686999976634979},{"id":"https://openalex.org/keywords/image","display_name":"Image (mathematics)","score":0.35760000348091125},{"id":"https://openalex.org/keywords/representation","display_name":"Representation (politics)","score":0.35350000858306885}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7720000147819519},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.6365000009536743},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.5354999899864197},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49889999628067017},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.42570000886917114},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.41819998621940613},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.39910000562667847},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.37070000171661377},{"id":"https://openalex.org/C15336307","wikidata":"https://www.wikidata.org/wiki/Q1766051","display_name":"Lens (geology)","level":2,"score":0.3686999976634979},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.35760000348091125},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.35350000858306885},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.35199999809265137},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.3476000130176544},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.3425000011920929},{"id":"https://openalex.org/C111640148","wikidata":"https://www.wikidata.org/wiki/Q847349","display_name":"Rubric","level":2,"score":0.33899998664855957},{"id":"https://openalex.org/C9616225","wikidata":"https://www.wikidata.org/wiki/Q3929429","display_name":"Semantic reasoner","level":2,"score":0.3262999951839447},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3253999948501587},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.3167000114917755},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3005000054836273},{"id":"https://openalex.org/C113843644","wikidata":"https://www.wikidata.org/wiki/Q901882","display_name":"Interface (matter)","level":4,"score":0.2962999939918518},{"id":"https://openalex.org/C90312973","wikidata":"https://www.wikidata.org/wiki/Q7449052","display_name":"Semantic data model","level":2,"score":0.295199990272522},{"id":"https://openalex.org/C43091099","wikidata":"https://www.wikidata.org/wiki/Q1067788","display_name":"Through-the-lens metering","level":3,"score":0.28940001130104065},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.28780001401901245},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2703000009059906},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.25999999046325684},{"id":"https://openalex.org/C2778915421","wikidata":"https://www.wikidata.org/wiki/Q3643177","display_name":"Performance improvement","level":2,"score":0.25220000743865967}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.21573","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21573","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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.2605.21573","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21573","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":[{"display_name":"Quality Education","score":0.591084361076355,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"We":[0],"introduce":[1],"Lens,":[2],"a":[3,76,140,150,185,243,250],"3.8B-parameter":[4],"T2I":[5],"model":[6,60],"that":[7,143,154],"achieves":[8],"performance":[9],"competitive":[10],"with,":[11],"and":[12,90,105,115,148,174,181,201,210,223,228],"in":[13,231,246,263],"several":[14,232],"cases":[15],"surpassing,":[16],"state-of-the-art":[17],"models":[18],"with":[19,112,170,188,198],"more":[20],"than":[21,101],"6B":[22],"parameters":[23],"across":[24],"various":[25],"benchmarks,":[26],"while":[27,157,255],"requiring":[28],"significantly":[29],"less":[30],"training":[31,42,48,69,73,163,209],"compute.":[32],"For":[33],"example,":[34],"Lens":[35,51,213,241],"requires":[36],"only":[37],"about":[38],"19.3%":[39],"of":[40,50,78,125],"the":[41,121,199],"compute":[43],"used":[44,234],"by":[45,71,88],"Z-Image.":[46],"The":[47],"efficiency":[49],"stems":[52],"from":[53,110,161,219],"two":[54],"key":[55],"strategies":[56],"beyond":[57],"its":[58,238,256],"compact":[59,239],"size.":[61],"First,":[62],"we":[63,130,167],"maximize":[64],"data":[65],"information":[66],"density":[67],"per":[68],"batch":[70,109],"(i)":[72],"on":[74,95,249],"Lens-800M,":[75],"dataset":[77],"800M":[79],"densely":[80],"captioned":[81],"image-text":[82],"pairs":[83],"whose":[84],"captions":[85],"are":[86],"generated":[87],"GPT-4.1":[89],"contain":[91],"approximately":[92],"109":[93],"words":[94],"average,":[96],"providing":[97],"richer":[98],"semantic":[99,141],"supervision":[100],"conventional":[102],"short":[103],"captions,":[104],"(ii)":[106],"constructing":[107],"each":[108,126],"images":[111],"multiple":[113],"resolutions":[114,224],"diverse":[116],"aspect":[117,217],"ratios,":[118],"thereby":[119],"enlarging":[120],"effective":[122],"visual":[123,183],"coverage":[124],"optimization":[127,156],"step.":[128],"Second,":[129],"improve":[131,182],"convergence":[132],"speed":[133],"through":[134],"careful":[135],"architectural":[136],"choices,":[137],"including":[138],"adopting":[139],"VAE":[142],"provides":[144],"better":[145,194],"latent":[146],"representations":[147],"employing":[149],"strong":[151],"language":[152],"encoder":[153],"accelerates":[155],"enabling":[158],"multilingual":[159],"generalization":[160],"English-only":[162],"data.":[164],"After":[165],"pre-training,":[166],"apply":[168],"RL":[169],"taxonomy-driven":[171],"prompts":[172,230],"(Lens-RL-8K)":[173],"structured":[175],"reward":[176],"rubrics":[177],"to":[178,193,215,221,226,237],"suppress":[179],"artifacts":[180],"quality,":[184],"reasoner":[186],"module":[187],"training-free":[189],"system":[190],"prompt":[191],"search":[192],"align":[195],"user":[196],"requests":[197],"model,":[200],"distillation-based":[202],"acceleration":[203],"for":[204],"4-step":[205,261],"inference.":[206],"Through":[207],"efficient":[208],"systematic":[211],"optimization,":[212],"generalizes":[214],"arbitrary":[216],"ratios":[218],"1:2":[220],"2:1":[222],"up":[225],"1440^2,":[227],"supports":[229],"commonly":[233],"languages.":[235],"Thanks":[236],"size,":[240],"generates":[242],"1024^2":[244],"image":[245],"3.15":[247],"seconds":[248],"single":[251],"NVIDIA":[252],"H100":[253],"GPU,":[254],"distilled":[257],"turbo":[258],"version":[259],"performs":[260],"generation":[262],"0.84":[264],"seconds.":[265]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-23T00:00:00"}