{"id":"https://openalex.org/W4387964026","doi":"https://doi.org/10.48550/arxiv.2310.16656","title":"A Picture is Worth a Thousand Words: Principled Recaptioning Improves Image Generation","display_name":"A Picture is Worth a Thousand Words: Principled Recaptioning Improves Image Generation","alternate_titles":[],"publication_year":2023,"publication_date":"2023-10-25","ids":{"openalex":"https://openalex.org/W4387964026","doi":"https://doi.org/10.48550/arxiv.2310.16656"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2310.16656","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.16656","pdf_url":"https://arxiv.org/pdf/2310.16656","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"parent_works":[],"lineage":["https://openalex.org/W4387964026"],"child_works_count":0,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2310.16656","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5007976734","display_name":"Eyal Segalis","orcid":"https://orcid.org/0009-0004-6529-5361","observed_orcids":["https://orcid.org/0009-0004-6529-5361"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Segalis, Eyal","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5038116890","display_name":"Dani Valevski","orcid":"https://orcid.org/0000-0002-4211-2866","observed_orcids":["https://orcid.org/0000-0002-4211-2866"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Valevski, Dani","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5093129970","display_name":"Danny Lumen","orcid":"https://orcid.org/0009-0003-2875-3629","observed_orcids":["https://orcid.org/0009-0003-2875-3629"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lumen, Danny","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5065128060","display_name":"Yossi Matias","orcid":"https://orcid.org/0000-0003-3960-6002","observed_orcids":["https://orcid.org/0000-0003-3960-6002"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Matias, Yossi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5003826475","display_name":"Yaniv Leviathan","orcid":"https://orcid.org/0009-0000-4080-4845","observed_orcids":["https://orcid.org/0009-0000-4080-4845"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leviathan, Yaniv","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":true,"cited_by_count":61,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_pulled":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.9970999956130981,"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":0.9970999956130981,"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":0.00279999990016222,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.0,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/text-to-image-diffusion-models","display_name":"text-to-image diffusion models","score":0.8550000190734863},{"id":"https://openalex.org/keywords/caption-quality","display_name":"caption quality","score":0.7400000095367432},{"id":"https://openalex.org/keywords/automatic-captioning","display_name":"automatic captioning","score":0.6190000176429749},{"id":"https://openalex.org/keywords/training-data-quality","display_name":"training data quality","score":0.40400001406669617},{"id":"https://openalex.org/keywords/semantic-alignment","display_name":"semantic alignment","score":0.8050000071525574},{"id":"https://openalex.org/keywords/sample-efficiency","display_name":"sample efficiency","score":0.7889999747276306},{"id":"https://openalex.org/keywords/image-captioning","display_name":"image captioning","score":0.49799999594688416}],"concepts":[{"id":"https://openalex.org/C157657479","wikidata":"https://www.wikidata.org/wiki/Q2367247","display_name":"Closed captioning","level":3,"score":0.8555773496627808},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7711158394813538},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6107704639434814},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.5528094172477722},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5498396158218384},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5473275184631348},{"id":"https://openalex.org/C115961682","wikidata":"https://www.wikidata.org/wiki/Q860623","display_name":"Image (mathematics)","level":2,"score":0.5364680290222168},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.49715355038642883},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.47261786460876465},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.4553963243961334},{"id":"https://openalex.org/C55020928","wikidata":"https://www.wikidata.org/wiki/Q3813865","display_name":"Image quality","level":3,"score":0.4548937678337097},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.4496609568595886},{"id":"https://openalex.org/C2781238097","wikidata":"https://www.wikidata.org/wiki/Q175026","display_name":"Object (grammar)","level":2,"score":0.4445739686489105},{"id":"https://openalex.org/C185592680","wikidata":"https://www.wikidata.org/wiki/Q2329","display_name":"Chemistry","level":0,"score":0.0},{"id":"https://openalex.org/C111368507","wikidata":"https://www.wikidata.org/wiki/Q43518","display_name":"Oceanography","level":1,"score":0.0},{"id":"https://openalex.org/C43617362","wikidata":"https://www.wikidata.org/wiki/Q170050","display_name":"Chromatography","level":1,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0},{"id":"https://openalex.org/C111472728","wikidata":"https://www.wikidata.org/wiki/Q9471","display_name":"Epistemology","level":1,"score":0.0},{"id":"https://openalex.org/C127313418","wikidata":"https://www.wikidata.org/wiki/Q1069","display_name":"Geology","level":0,"score":0.0},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2310.16656","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.16656","pdf_url":"https://arxiv.org/pdf/2310.16656","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2310.16656","doi":"https://doi.org/10.48550/arxiv.2310.16656","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2310.16656","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2310.16656","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2310.16656","pdf_url":"https://arxiv.org/pdf/2310.16656","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,"listed_in":[],"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4387964026.pdf","grobid_xml":"https://content.openalex.org/works/W4387964026.grobid-xml"},"referenced_works_count":19,"referenced_works":["https://openalex.org/W2963966654","https://openalex.org/W2981852735","https://openalex.org/W2982450728","https://openalex.org/W3129576130","https://openalex.org/W3135367836","https://openalex.org/W3193402170","https://openalex.org/W3203737321","https://openalex.org/W4224035735","https://openalex.org/W4225323055","https://openalex.org/W4229042118","https://openalex.org/W4283388932","https://openalex.org/W4296406182","https://openalex.org/W4298289240","https://openalex.org/W4312564121","https://openalex.org/W4312933868","https://openalex.org/W4313484371","https://openalex.org/W4385270985","https://openalex.org/W4390873281","https://openalex.org/W7133243546"],"related_works":["https://openalex.org/W4404782000","https://openalex.org/W4416059700","https://openalex.org/W4304731239","https://openalex.org/W4411635418","https://openalex.org/W7164916451","https://openalex.org/W4416158451","https://openalex.org/W4404782101","https://openalex.org/W4404782023","https://openalex.org/W4416048728","https://openalex.org/W4401043304"],"abstract_inverted_index":{"Text-to-image":[0],"diffusion":[1],"models":[2,30,48],"achieved":[3],"a":[4,22,126,133],"remarkable":[5],"leap":[6],"in":[7,40,114,148,163,172],"capabilities":[8],"over":[9],"the":[10,27,38,59,64,67,78,107,115,123,137,140,145,156,200,214,219,231,236],"last":[11],"few":[12],"years,":[13],"enabling":[14],"high-quality":[15],"and":[16,66,85,101,131,160,188,202,217,229,240],"diverse":[17],"synthesis":[18],"of":[19,37,46,54,98,158],"images":[20,60],"from":[21,63],"textual":[23,116],"prompt.":[24],"However,":[25],"even":[26],"most":[28],"advanced":[29],"often":[31,61,97],"struggle":[32],"to":[33,110,168,198,233],"precisely":[34],"follow":[35],"all":[36],"directions":[39],"their":[41,70],"prompts.":[42,117],"The":[43],"vast":[44],"majority":[45],"these":[47,94],"are":[49,69,96],"trained":[50],"on":[51,136],"datasets":[52],"consisting":[53],"(image,":[55],"caption)":[56],"pairs":[57],"where":[58],"come":[62],"web,":[65],"captions":[68,95,239],"HTML":[71],"alternate":[72],"text.":[73],"A":[74],"notable":[75],"example":[76],"is":[77],"LAION":[79],"dataset,":[80,139],"used":[81],"by":[82,121],"Stable":[83],"Diffusion":[84],"other":[86],"models.":[87],"In":[88],"this":[89,104,206],"work":[90],"we":[91,209],"observe":[92],"that":[93,103,120,205],"low":[99],"quality,":[100],"argue":[102],"significantly":[105],"affects":[106],"model's":[108],"capability":[109],"understand":[111,235],"nuanced":[112],"semantics":[113],"We":[118,194],"show":[119],"relabeling":[122],"corpus":[124,201],"with":[125,221],"specialized":[127],"automatic":[128],"captioning":[129],"model":[130,135,141,220,232],"training":[132],"text-to-image":[134],"recaptioned":[138],"benefits":[142],"substantially":[143],"across":[144],"board.":[146],"First,":[147],"overall":[149],"image":[150,165],"quality:":[151],"e.g.":[152,175],"FID":[153],"14.84":[154],"vs.":[155,180,186,192],"baseline":[157],"17.87,":[159],"64.3%":[161],"improvement":[162],"faithful":[164],"generation":[166],"according":[167],"human":[169],"evaluation.":[170],"Second,":[171],"semantic":[173,176],"alignment,":[174],"object":[177],"accuracy":[178],"84.34":[179],"78.90,":[181],"counting":[182],"alignment":[183,190],"errors":[184],"1.32":[185],"1.44":[187],"positional":[189],"62.42":[191],"57.60.":[193],"analyze":[195],"various":[196],"ways":[197],"relabel":[199],"provide":[203],"evidence":[204],"technique,":[207],"which":[208],"call":[210],"RECAP,":[211],"both":[212],"reduces":[213],"train-inference":[215],"discrepancy":[216],"provides":[218],"more":[222],"information":[223],"per":[224],"example,":[225],"increasing":[226],"sample":[227],"efficiency":[228],"allowing":[230],"better":[234],"relations":[237],"between":[238],"images.":[241]},"cited_by_api_url":"https://api.openalex.org/works?filter=cites:W4387964026","counts_by_year":[{"year":2026,"cited_by_count":13},{"year":2025,"cited_by_count":24},{"year":2024,"cited_by_count":24}],"updated_date":"2026-11-16T12:05:36.265796","created_date":"2023-10-27T00:00:00"}
