{"id":"https://openalex.org/W7208764013","doi":"https://doi.org/10.48550/arxiv.2609.03586","title":"The Attention Triangle in Audio-Video Models","display_name":"The Attention Triangle in Audio-Video Models","publication_year":2026,"publication_date":"2026-09-03","ids":{"openalex":"https://openalex.org/W7208764013","doi":"https://doi.org/10.48550/arxiv.2609.03586"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.03586","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.03586","pdf_url":"https://arxiv.org/pdf/2609.03586","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"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.03586","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5069631276","display_name":"Sagi Polaczek","orcid":"https://orcid.org/0009-0004-4145-1235","observed_orcids":["https://orcid.org/0009-0004-4145-1235"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Polaczek, Sagi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5092994564","display_name":"Noa Kraicer","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kraicer, Noa","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5085386004","display_name":"Gal Metzer","orcid":"https://orcid.org/0009-0007-6541-6129","observed_orcids":["https://orcid.org/0009-0007-6541-6129"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Metzer, Gal","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5150042988","display_name":"Zhuo Ning","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ning, Zhuo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021862056","display_name":"Ali Mahdavi Amiri","orcid":"https://orcid.org/0000-0002-4693-3565","observed_orcids":["https://orcid.org/0000-0002-4693-3565"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mahdavi-Amiri, Ali","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036688260","display_name":"Daniel Cohen\u2010Or","orcid":"https://orcid.org/0000-0001-6777-7445","observed_orcids":["https://orcid.org/0000-0001-6777-7445"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cohen-Or, Daniel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5072571599","display_name":"Raja Giryes","orcid":"https://orcid.org/0000-0002-2830-0297","observed_orcids":["https://orcid.org/0000-0002-2830-0297"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Giryes, Raja","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":0.8115000128746033,"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":0.8115000128746033,"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":0.18199999630451202,"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/T12032","display_name":"Multisensory perception and integration","score":0.0024999999441206455,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/cross-modal-attention","display_name":"cross-modal attention","score":0.9800000190734863},{"id":"https://openalex.org/keywords/learned-priors","display_name":"learned priors","score":0.5799999833106995},{"id":"https://openalex.org/keywords/cross-modal-alignment","display_name":"cross-modal alignment","score":0.6970000267028809}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7276999950408936},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.6610999703407288},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.5645999908447266},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.48339998722076416},{"id":"https://openalex.org/C2779903281","wikidata":"https://www.wikidata.org/wiki/Q6888026","display_name":"Modalities","level":2,"score":0.39820000529289246},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.3880000114440918},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.38530001044273376},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.34360000491142273},{"id":"https://openalex.org/C130318100","wikidata":"https://www.wikidata.org/wiki/Q2268914","display_name":"Semantic similarity","level":2,"score":0.29919999837875366},{"id":"https://openalex.org/C90312973","wikidata":"https://www.wikidata.org/wiki/Q7449052","display_name":"Semantic data model","level":2,"score":0.2962000072002411},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.29440000653266907},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.2922999858856201},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2833000123500824},{"id":"https://openalex.org/C2779631151","wikidata":"https://www.wikidata.org/wiki/Q7883767","display_name":"Underspecification","level":2,"score":0.27390000224113464},{"id":"https://openalex.org/C2778180026","wikidata":"https://www.wikidata.org/wiki/Q18378163","display_name":"Semantic heterogeneity","level":4,"score":0.2671000063419342},{"id":"https://openalex.org/C74172769","wikidata":"https://www.wikidata.org/wiki/Q1446839","display_name":"Routing (electronic design automation)","level":2,"score":0.2547999918460846},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.25220000743865967}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.03586","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.03586","pdf_url":"https://arxiv.org/pdf/2609.03586","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.2609.03586","doi":"https://doi.org/10.48550/arxiv.2609.03586","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.03586","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:2609.03586","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.03586","pdf_url":"https://arxiv.org/pdf/2609.03586","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":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4386075767","https://openalex.org/W7139922230","https://openalex.org/W7162406746","https://openalex.org/W6929528238","https://openalex.org/W7214405946","https://openalex.org/W7160902017","https://openalex.org/W4414663893","https://openalex.org/W7160929242","https://openalex.org/W4416019625","https://openalex.org/W3160431667"],"abstract_inverted_index":{"Audio-video":[0],"diffusion":[1],"models":[2,28],"rely":[3],"on":[4,149],"cross-modal":[5,108,192,214],"attention":[6,134],"to":[7,98,173,186,206],"coordinate":[8],"text,":[9,43],"sound,":[10],"and":[11,21,31,45,48,92,115,162,166,176,194,221],"visual":[12],"content,":[13],"yet":[14],"this":[15,150],"same":[16],"mechanism":[17],"can":[18,71,77],"introduce":[19],"subtle":[20],"systematic":[22],"semantic":[23,51,128,224],"leakage.":[24],"We":[25,201],"study":[26],"these":[27,204],"by":[29,85],"probing":[30],"analyzing":[32],"the":[33,37,42,65,89,112,188,196],"``attention":[34],"triangle,''":[35],"comprising":[36],"three":[38],"cross-attention":[39],"edges":[40],"connecting":[41],"audio,":[44],"video":[46,73,76],"streams,":[47],"examine":[49],"how":[50,158],"information":[52],"is":[53,68,83],"routed":[54],"across":[55,164],"modalities":[56],"during":[57],"generation.":[58,80],"Our":[59],"analysis":[60,220],"reveals":[61],"that":[62,127,156,210],"routing":[63,193],"along":[64,145],"audio-video":[66],"edge":[67,82],"bidirectional:":[69],"audio":[70,79],"influence":[72,78],"generation,":[74],"while":[75,226],"This":[81,183],"shaped":[84],"biases":[86],"encoded":[87],"in":[88,103],"model's":[90],"parameters":[91],"emerges":[93],"as":[94,169],"a":[95,170],"major":[96],"contributor":[97],"leakage:":[99],"when":[100],"prompts":[101],"are":[102,160],"tension":[104],"with":[105],"learned":[106],"priors,":[107],"interactions":[109,144],"may":[110],"override":[111],"intended":[113,138],"conditioning":[114],"reroute":[116],"semantics":[117,159],"toward":[118],"visually":[119],"canonical":[120],"but":[121,140],"incorrect":[122],"outcomes.":[123],"These":[124],"effects":[125],"suggest":[126],"artifacts":[129],"arise":[130],"not":[131],"merely":[132],"from":[133,141],"spreading":[135],"beyond":[136],"its":[137],"target,":[139],"structured,":[142],"bias-driven":[143],"specific":[146],"pathways.":[147],"Building":[148],"perspective,":[151],"we":[152],"extract":[153],"attention-derived":[154],"signals":[155,205],"expose":[157],"distributed":[161],"grounded":[163],"modalities,":[165],"use":[167],"them":[168],"diagnostic":[171],"tool":[172],"both":[174],"analyze":[175],"deliberately":[177],"incur":[178],"leakage":[179],"under":[180],"controlled":[181],"conditions.":[182],"enables":[184],"us":[185],"probe":[187],"internal":[189],"dynamics":[190],"of":[191,198],"isolate":[195],"role":[197],"individual":[199],"interactions.":[200],"further":[202],"leverage":[203],"guide":[207],"inference-time":[208],"interventions":[209],"encourage":[211],"more":[212],"consistent":[213],"alignment.":[215],"Extensive":[216],"experiments":[217],"support":[218],"our":[219],"demonstrate":[222],"improved":[223],"grounding":[225],"preserving":[227],"generation":[228],"quality.":[229]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-05T00:00:00"}