{"id":"https://openalex.org/W4400527815","doi":"https://doi.org/10.1109/fg59268.2024.10581912","title":"Discovering Interpretable Directions in the Semantic Latent Space of Diffusion Models","display_name":"Discovering Interpretable Directions in the Semantic Latent Space of Diffusion Models","publication_year":2024,"publication_date":"2024-05-27","ids":{"openalex":"https://openalex.org/W4400527815","doi":"https://doi.org/10.1109/fg59268.2024.10581912"},"language":"en","primary_location":{"id":"doi:10.1109/fg59268.2024.10581912","is_oa":false,"landing_page_url":"https://doi.org/10.1109/fg59268.2024.10581912","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2024 IEEE 18th International Conference on Automatic Face and Gesture Recognition (FG)","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/A5088345591","display_name":"Ren\u00e9 Haas","orcid":"https://orcid.org/0000-0002-9417-8787"},"institutions":[{"id":"https://openalex.org/I83467386","display_name":"IT University of Copenhagen","ror":"https://ror.org/02309jg23","country_code":"DK","type":"education","lineage":["https://openalex.org/I83467386"]}],"countries":["DK"],"is_corresponding":false,"raw_author_name":"Ren\u00e9 Haas","raw_affiliation_strings":["IT University of Copenhagen,Computer Science,Denmark"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IT University of Copenhagen,Computer Science,Denmark","institution_ids":["https://openalex.org/I83467386"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052278471","display_name":"Inbar Huberman-Spiegelglas","orcid":null},"institutions":[{"id":"https://openalex.org/I174306211","display_name":"Technion \u2013 Israel Institute of Technology","ror":"https://ror.org/03qryx823","country_code":"IL","type":"education","lineage":["https://openalex.org/I174306211"]}],"countries":["IL"],"is_corresponding":false,"raw_author_name":"Inbar Huberman-Spiegelglas","raw_affiliation_strings":["Technion,Computer Science,Israel"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technion,Computer Science,Israel","institution_ids":["https://openalex.org/I174306211"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076334995","display_name":"Rotem Mulayoff","orcid":"https://orcid.org/0000-0002-5991-0602"},"institutions":[{"id":"https://openalex.org/I174306211","display_name":"Technion \u2013 Israel Institute of Technology","ror":"https://ror.org/03qryx823","country_code":"IL","type":"education","lineage":["https://openalex.org/I174306211"]}],"countries":["IL"],"is_corresponding":false,"raw_author_name":"Rotem Mulayoff","raw_affiliation_strings":["Technion,Computer Science,Israel"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technion,Computer Science,Israel","institution_ids":["https://openalex.org/I174306211"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5074208572","display_name":"Stella Gra\u00dfhof","orcid":"https://orcid.org/0000-0002-6791-7425"},"institutions":[{"id":"https://openalex.org/I83467386","display_name":"IT University of Copenhagen","ror":"https://ror.org/02309jg23","country_code":"DK","type":"education","lineage":["https://openalex.org/I83467386"]}],"countries":["DK"],"is_corresponding":false,"raw_author_name":"Stella Gra\u00dfhof","raw_affiliation_strings":["IT University of Copenhagen,Computer Science,Denmark"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IT University of Copenhagen,Computer Science,Denmark","institution_ids":["https://openalex.org/I83467386"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5083670972","display_name":"Sami S. Brandt","orcid":"https://orcid.org/0000-0003-2141-9815"},"institutions":[{"id":"https://openalex.org/I83467386","display_name":"IT University of Copenhagen","ror":"https://ror.org/02309jg23","country_code":"DK","type":"education","lineage":["https://openalex.org/I83467386"]}],"countries":["DK"],"is_corresponding":false,"raw_author_name":"Sami S. Brandt","raw_affiliation_strings":["IT University of Copenhagen,Computer Science,Denmark"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IT University of Copenhagen,Computer Science,Denmark","institution_ids":["https://openalex.org/I83467386"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5058178036","display_name":"Tomer Michaeli","orcid":"https://orcid.org/0000-0003-0525-8054"},"institutions":[{"id":"https://openalex.org/I174306211","display_name":"Technion \u2013 Israel Institute of Technology","ror":"https://ror.org/03qryx823","country_code":"IL","type":"education","lineage":["https://openalex.org/I174306211"]}],"countries":["IL"],"is_corresponding":false,"raw_author_name":"Tomer Michaeli","raw_affiliation_strings":["Technion,Computer Science,Israel"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Technion,Computer Science,Israel","institution_ids":["https://openalex.org/I174306211"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":22,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"9"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9847999811172485,"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"}},"topics":[{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9847999811172485,"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"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9793000221252441,"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"}},{"id":"https://openalex.org/T12205","display_name":"Time Series Analysis and Forecasting","score":0.9775000214576721,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/computer-science","display_name":"Computer science","score":0.6748580932617188},{"id":"https://openalex.org/keywords/space","display_name":"Space (punctuation)","score":0.5922638773918152},{"id":"https://openalex.org/keywords/diffusion","display_name":"Diffusion","score":0.589654266834259},{"id":"https://openalex.org/keywords/semantic-space","display_name":"Semantic space","score":0.5184237957000732},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4185093641281128},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.329987108707428},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.06048974394798279}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6748580932617188},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.5922638773918152},{"id":"https://openalex.org/C69357855","wikidata":"https://www.wikidata.org/wiki/Q163214","display_name":"Diffusion","level":2,"score":0.589654266834259},{"id":"https://openalex.org/C2986420190","wikidata":"https://www.wikidata.org/wiki/Q39045939","display_name":"Semantic space","level":2,"score":0.5184237957000732},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4185093641281128},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.329987108707428},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.06048974394798279},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0},{"id":"https://openalex.org/C97355855","wikidata":"https://www.wikidata.org/wiki/Q11473","display_name":"Thermodynamics","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/fg59268.2024.10581912","is_oa":false,"landing_page_url":"https://doi.org/10.1109/fg59268.2024.10581912","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2024 IEEE 18th International Conference on Automatic Face and Gesture Recognition (FG)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[{"score":0.7300000190734863,"display_name":"Reduced inequalities","id":"https://metadata.un.org/sdg/10"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":61,"referenced_works":["https://openalex.org/W1834627138","https://openalex.org/W1901129140","https://openalex.org/W2137306662","https://openalex.org/W2139047213","https://openalex.org/W2480854438","https://openalex.org/W2607662938","https://openalex.org/W2747329762","https://openalex.org/W2963644257","https://openalex.org/W2990138404","https://openalex.org/W3014852036","https://openalex.org/W3014859719","https://openalex.org/W3026885507","https://openalex.org/W3034431451","https://openalex.org/W3035355202","https://openalex.org/W3094954720","https://openalex.org/W3121370741","https://openalex.org/W3129576130","https://openalex.org/W3173241699","https://openalex.org/W3173562028","https://openalex.org/W3177221875","https://openalex.org/W3217030260","https://openalex.org/W4214926101","https://openalex.org/W4224035735","https://openalex.org/W4280557136","https://openalex.org/W4287817254","https://openalex.org/W4289785045","https://openalex.org/W4289785095","https://openalex.org/W4300979859","https://openalex.org/W4308170885","https://openalex.org/W4312740349","https://openalex.org/W4312933868","https://openalex.org/W4318203092","https://openalex.org/W4320900831","https://openalex.org/W4322616363","https://openalex.org/W4385271134","https://openalex.org/W4386057725","https://openalex.org/W4386072096","https://openalex.org/W4386076532","https://openalex.org/W4386113271","https://openalex.org/W4387171440","https://openalex.org/W6679045638","https://openalex.org/W6685352114","https://openalex.org/W6779823529","https://openalex.org/W6783713337","https://openalex.org/W6789730206","https://openalex.org/W6791353385","https://openalex.org/W6795288823","https://openalex.org/W6796788208","https://openalex.org/W6809885388","https://openalex.org/W6810165568","https://openalex.org/W6810940779","https://openalex.org/W6838639034","https://openalex.org/W6838713958","https://openalex.org/W6840155194","https://openalex.org/W6840815571","https://openalex.org/W6841366371","https://openalex.org/W6845433216","https://openalex.org/W6846344700","https://openalex.org/W6847090916","https://openalex.org/W6848774630","https://openalex.org/W6850025991"],"related_works":["https://openalex.org/W3188962172","https://openalex.org/W2772917594","https://openalex.org/W4312825515","https://openalex.org/W4306742369","https://openalex.org/W4303457083","https://openalex.org/W2386796262","https://openalex.org/W2131146434","https://openalex.org/W2951359407","https://openalex.org/W4376623224","https://openalex.org/W3204019825"],"abstract_inverted_index":{"Denoising":[0],"Diffusion":[1],"Models":[2],"(DDMs)":[3],"have":[4],"emerged":[5],"as":[6,32,128],"a":[7,36,56,139,169,188],"strong":[8],"competitor":[9],"to":[10,50,196],"Generative":[11],"Adversarial":[12],"Networks":[13],"(GANs).":[14],"However,":[15],"despite":[16],"their":[17,26],"widespread":[18],"use":[19],"in":[20,55,72,118,132,168,172],"image":[21,53],"synthesis":[22],"and":[23,95],"editing":[24,54],"applications,":[25],"latent":[27,38,134,158],"space":[28,39],"is":[29,66],"still":[30],"not":[31],"well":[33],"understood.":[34],"Recently,":[35],"semantic":[37,52,103,116,145],"for":[40,100,113,142],"DDMs,":[41],"coined":[42],"\u2018":[43],"<tex":[44,62,91],"xmlns:mml=\"http://www.w3.org/1998/Math/MathML\"":[45,63,92],"xmlns:xlink=\"http://www.w3.org/1999/xlink\">$h$</tex>":[46,64,93],"-space\u2019,":[47],"was":[48],"shown":[49],"facilitate":[51],"way":[57],"reminiscent":[58],"of":[59,68,79,90,150,153],"GANs.":[60],"The":[61],"-space":[65,94],"comprised":[67],"the":[69,73,80,88,129,133,151,154,157,163,199],"bottleneck":[70],"activations":[71],"DDM's":[74],"denoiser":[75,155],"across":[76],"all":[77],"timesteps":[78],"diffusion":[81],"process.":[82],"In":[83],"this":[84],"paper,":[85],"we":[86,122,137,161],"explore":[87],"properties":[89],"propose":[96],"several":[97],"novel":[98,140],"methods":[99,112],"finding":[101,166],"meaningful":[102],"directions":[104,117,126,146,167,179,201],"within":[105],"it.":[106],"We":[107,175,192],"start":[108],"by":[109,147,165,183,202],"studying":[110],"unsupervised":[111],"revealing":[114],"interpretable":[115,125],"pretrained":[119],"DDMs.":[120,174],"Specifically,":[121],"show":[123,194],"that":[124],"emerge":[127],"principal":[130],"components":[131],"space.":[135],"Additionally,":[136],"provide":[138],"method":[141],"discovering":[143],"image-specific":[144],"spectral":[148],"analysis":[149,164],"Jacobian":[152],"w.r.t.":[156],"code.":[159],"Next,":[160],"extend":[162],"supervised":[170],"fashion":[171],"unconditional":[173],"demonstrate":[176],"how":[177,195],"such":[178],"can":[180],"be":[181],"found":[182,200],"annotating":[184],"generated":[185],"samples":[186],"with":[187],"domain-specific":[189],"attribute":[190],"classifier.":[191],"further":[193],"semantically":[197],"disentangle":[198],"simple":[203],"linear":[204],"projection.":[205],"Our":[206],"approaches":[207],"are":[208],"applicable":[209],"without":[210],"requiring":[211],"any":[212],"architectural":[213],"modifications,":[214],"text-based":[215],"guidance,":[216],"CLIP-based":[217],"optimization,":[218],"or":[219],"model":[220],"fine-tuning.":[221]},"counts_by_year":[{"year":2026,"cited_by_count":6},{"year":2025,"cited_by_count":8},{"year":2024,"cited_by_count":7},{"year":2023,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}