{"id":"https://openalex.org/W7148553235","doi":"https://doi.org/10.1109/asru65441.2025.11434763","title":"Adaptive Audio-Visual Speech Recognition via Matryoshka-Based Multimodal LLMs","display_name":"Adaptive Audio-Visual Speech Recognition via Matryoshka-Based Multimodal LLMs","publication_year":2025,"publication_date":"2025-12-06","ids":{"openalex":"https://openalex.org/W7148553235","doi":"https://doi.org/10.1109/asru65441.2025.11434763"},"language":null,"primary_location":{"id":"doi:10.1109/asru65441.2025.11434763","is_oa":false,"landing_page_url":"https://doi.org/10.1109/asru65441.2025.11434763","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)","raw_type":"proceedings-article"},"type":"article","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/A5043797502","display_name":"Umberto Cappellazzo","orcid":"https://orcid.org/0009-0006-3443-5143"},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":true,"raw_author_name":"Umberto Cappellazzo","raw_affiliation_strings":["Imperial College,London"],"affiliations":[{"raw_affiliation_string":"Imperial College,London","institution_ids":["https://openalex.org/I47508984"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132807823","display_name":"Minsu Kim","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Minsu Kim","raw_affiliation_strings":["Meta AI"],"affiliations":[{"raw_affiliation_string":"Meta AI","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5009475700","display_name":"Stavros Petridis","orcid":"https://orcid.org/0000-0001-7478-9479"},"institutions":[{"id":"https://openalex.org/I47508984","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10","country_code":"GB","type":"education","lineage":["https://openalex.org/I47508984"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Stavros Petridis","raw_affiliation_strings":["Imperial College,London"],"affiliations":[{"raw_affiliation_string":"Imperial College,London","institution_ids":["https://openalex.org/I47508984"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5043797502"],"corresponding_institution_ids":["https://openalex.org/I47508984"],"apc_list":null,"apc_paid":null,"fwci":1.2783,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.8589783,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":96,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"8"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10860","display_name":"Speech and Audio Processing","score":0.91839998960495,"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"}},"topics":[{"id":"https://openalex.org/T10860","display_name":"Speech and Audio Processing","score":0.91839998960495,"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"}},{"id":"https://openalex.org/T10667","display_name":"Emotion and Mood Recognition","score":0.017100000753998756,"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"}},{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.009600000455975533,"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/feature","display_name":"Feature (linguistics)","score":0.3066999912261963},{"id":"https://openalex.org/keywords/field","display_name":"Field (mathematics)","score":0.28110000491142273},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.2554999887943268},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.2540999948978424}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5540000200271606},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.5141000151634216},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3862000107765198},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.32199999690055847},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3066999912261963},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.28110000491142273},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.2554999887943268},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2540999948978424},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.25360000133514404},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.2531000077724457},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.2531000077724457}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/asru65441.2025.11434763","is_oa":false,"landing_page_url":"https://doi.org/10.1109/asru65441.2025.11434763","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320320283","display_name":"Imperial College London","ror":"https://ror.org/041kmwe10"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":33,"referenced_works":["https://openalex.org/W1974387177","https://openalex.org/W2076462394","https://openalex.org/W2121486117","https://openalex.org/W2551572271","https://openalex.org/W2890952074","https://openalex.org/W2963785710","https://openalex.org/W3162293946","https://openalex.org/W3210177631","https://openalex.org/W4297841411","https://openalex.org/W4319300051","https://openalex.org/W4372346152","https://openalex.org/W4386071467","https://openalex.org/W4392903321","https://openalex.org/W4392903956","https://openalex.org/W4392931626","https://openalex.org/W4393178509","https://openalex.org/W4402112360","https://openalex.org/W4402112372","https://openalex.org/W4402671204","https://openalex.org/W4402727885","https://openalex.org/W4402753980","https://openalex.org/W4402754023","https://openalex.org/W4404792841","https://openalex.org/W4408354168","https://openalex.org/W4408354236","https://openalex.org/W4412944344","https://openalex.org/W4413144796","https://openalex.org/W4413147769","https://openalex.org/W4415433635","https://openalex.org/W7133193597","https://openalex.org/W7133197633","https://openalex.org/W7133214597","https://openalex.org/W7133227004"],"related_works":[],"abstract_inverted_index":{"Audio-Visual":[0],"Speech":[1],"Recognition":[2],"(AVSR)":[3],"leverages":[4],"audio":[5],"and":[6,112],"visual":[7],"modalities":[8],"to":[9,36,49],"improve":[10],"robustness":[11],"in":[12,17,25],"noisy":[13],"environments.":[14],"Recent":[15],"advances":[16],"Large":[18],"Language":[19],"Models":[20],"(LLMs)":[21],"show":[22,120],"strong":[23],"performance":[24],"speech":[26,33],"recognition,":[27],"including":[28],"AVSR.":[29],"However,":[30],"the":[31,63,97],"long":[32],"representations":[34,88],"lead":[35],"high":[37,52],"computational":[38],"costs":[39],"for":[40,68,99],"LLMs.":[41],"Prior":[42],"methods":[43],"compress":[44],"inputs":[45],"before":[46],"feeding":[47],"them":[48],"LLMs,":[50],"but":[51],"compression":[53,129],"often":[54],"harms":[55],"accuracy.":[56],"To":[57],"address":[58],"this,":[59],"we":[60,105],"propose":[61],"Llama-MTSK,":[62],"first":[64],"Matryoshka-based":[65],"Multimodal":[66],"LLM":[67],"AVSR,":[69],"which":[70],"flexibly":[71],"adapts":[72],"audio-visual":[73],"token":[74],"allocation":[75],"under":[76],"varying":[77],"compute":[78],"constraints.":[79],"Inspired":[80],"by":[81],"Matryoshka":[82],"Representation":[83],"Learning,":[84],"our":[85],"model":[86],"encodes":[87],"at":[89,127],"multiple":[90],"granularities":[91],"with":[92],"a":[93],"single":[94],"architecture,":[95],"avoiding":[96],"need":[98],"separate":[100],"models.":[101],"For":[102],"efficient":[103],"fine-tuning,":[104],"introduce":[106],"three":[107],"LoRA-based":[108],"strategies":[109],"using":[110],"global":[111],"scale-specific":[113],"modules.":[114],"Evaluations":[115],"on":[116],"major":[117],"AVSR":[118],"datasets":[119],"Llama-MTSK":[121],"matches":[122],"or":[123],"outperforms":[124],"models":[125],"trained":[126],"fixed":[128],"levels.":[130]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-04-23T09:07:50.710637","created_date":"2026-04-03T00:00:00"}