{"id":"https://openalex.org/W4412887726","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1065","title":"MMS-LLaMA: Efficient LLM-based Audio-Visual Speech Recognition with Minimal Multimodal Speech Tokens","display_name":"MMS-LLaMA: Efficient LLM-based Audio-Visual Speech Recognition with Minimal Multimodal Speech Tokens","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412887726","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1065"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.1065","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1065","pdf_url":"https://aclanthology.org/2025.findings-acl.1065.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-acl.1065.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5089479205","display_name":"Jeong Hun Yeo","orcid":"https://orcid.org/0009-0002-8135-6625"},"institutions":[{"id":"https://openalex.org/I4210107233","display_name":"Language Science (South Korea)","ror":"https://ror.org/01h9v1373","country_code":"KR","type":"company","lineage":["https://openalex.org/I4210107233"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jeong Hun Yeo","raw_affiliation_strings":["Integrated Vision and Language Lab , KAIST , South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Integrated Vision and Language Lab , KAIST , South Korea","institution_ids":["https://openalex.org/I4210107233"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5036329402","display_name":"Hyeongseop Rha","orcid":null},"institutions":[{"id":"https://openalex.org/I4210107233","display_name":"Language Science (South Korea)","ror":"https://ror.org/01h9v1373","country_code":"KR","type":"company","lineage":["https://openalex.org/I4210107233"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Hyeongseop Rha","raw_affiliation_strings":["Integrated Vision and Language Lab , KAIST , South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Integrated Vision and Language Lab , KAIST , South Korea","institution_ids":["https://openalex.org/I4210107233"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100376689","display_name":"Se Jin Park","orcid":"https://orcid.org/0000-0002-1618-5906"},"institutions":[{"id":"https://openalex.org/I4210107233","display_name":"Language Science (South Korea)","ror":"https://ror.org/01h9v1373","country_code":"KR","type":"company","lineage":["https://openalex.org/I4210107233"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Se Jin Park","raw_affiliation_strings":["Integrated Vision and Language Lab , KAIST , South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Integrated Vision and Language Lab , KAIST , South Korea","institution_ids":["https://openalex.org/I4210107233"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5038798134","display_name":"Yong Man Ro","orcid":"https://orcid.org/0000-0001-5306-6853"},"institutions":[{"id":"https://openalex.org/I4210107233","display_name":"Language Science (South Korea)","ror":"https://ror.org/01h9v1373","country_code":"KR","type":"company","lineage":["https://openalex.org/I4210107233"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Yong Man Ro","raw_affiliation_strings":["Integrated Vision and Language Lab , KAIST , South Korea"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Integrated Vision and Language Lab , KAIST , South Korea","institution_ids":["https://openalex.org/I4210107233"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":4,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":3.1921,"has_fulltext":true,"cited_by_count":3,"citation_normalized_percentile":{"value":0.9219421,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":98,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"20724","last_page":"20735"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10860","display_name":"Speech and Audio Processing","score":0.9988999962806702,"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":0.9988999962806702,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.989300012588501,"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/T11309","display_name":"Music and Audio Processing","score":0.958299994468689,"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/speech-recognition","display_name":"Speech recognition","score":0.7947120666503906},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7611701488494873},{"id":"https://openalex.org/keywords/audio-mining","display_name":"Audio mining","score":0.554130494594574},{"id":"https://openalex.org/keywords/voice-activity-detection","display_name":"Voice activity detection","score":0.46513575315475464},{"id":"https://openalex.org/keywords/speech-synthesis","display_name":"Speech synthesis","score":0.4293527603149414},{"id":"https://openalex.org/keywords/audio-visual","display_name":"Audio visual","score":0.42584189772605896},{"id":"https://openalex.org/keywords/speech-coding","display_name":"Speech coding","score":0.42249369621276855},{"id":"https://openalex.org/keywords/speech-processing","display_name":"Speech processing","score":0.4220195412635803},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3219904899597168},{"id":"https://openalex.org/keywords/multimedia","display_name":"Multimedia","score":0.13728514313697815}],"concepts":[{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.7947120666503906},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7611701488494873},{"id":"https://openalex.org/C157968479","wikidata":"https://www.wikidata.org/wiki/Q3079876","display_name":"Audio mining","level":4,"score":0.554130494594574},{"id":"https://openalex.org/C204201278","wikidata":"https://www.wikidata.org/wiki/Q1332614","display_name":"Voice activity detection","level":3,"score":0.46513575315475464},{"id":"https://openalex.org/C14999030","wikidata":"https://www.wikidata.org/wiki/Q16346","display_name":"Speech synthesis","level":2,"score":0.4293527603149414},{"id":"https://openalex.org/C3017588708","wikidata":"https://www.wikidata.org/wiki/Q758901","display_name":"Audio visual","level":2,"score":0.42584189772605896},{"id":"https://openalex.org/C13895895","wikidata":"https://www.wikidata.org/wiki/Q3270773","display_name":"Speech coding","level":2,"score":0.42249369621276855},{"id":"https://openalex.org/C61328038","wikidata":"https://www.wikidata.org/wiki/Q3358061","display_name":"Speech processing","level":2,"score":0.4220195412635803},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3219904899597168},{"id":"https://openalex.org/C49774154","wikidata":"https://www.wikidata.org/wiki/Q131765","display_name":"Multimedia","level":1,"score":0.13728514313697815}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.1065","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1065","pdf_url":"https://aclanthology.org/2025.findings-acl.1065.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.1065","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1065","pdf_url":"https://aclanthology.org/2025.findings-acl.1065.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: ACL 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G300623544","display_name":null,"funder_award_id":"2022-0-00124","funder_id":"https://openalex.org/F4320328359","funder_display_name":"Ministry of Science and ICT, South Korea"},{"id":"https://openalex.org/G4700831490","display_name":null,"funder_award_id":"2022-","funder_id":"https://openalex.org/F4320335489","funder_display_name":"Institute for Information and Communications Technology Promotion"},{"id":"https://openalex.org/G5178579001","display_name":null,"funder_award_id":"NRF-2022R1A2C2005529","funder_id":"https://openalex.org/F4320322120","funder_display_name":"National Research Foundation of Korea"}],"funders":[{"id":"https://openalex.org/F4320320671","display_name":"National Research Foundation","ror":"https://ror.org/05s0g1g46"},{"id":"https://openalex.org/F4320322120","display_name":"National Research Foundation of Korea","ror":"https://ror.org/013aysd81"},{"id":"https://openalex.org/F4320328359","display_name":"Ministry of Science and ICT, South Korea","ror":"https://ror.org/01wpjm123"},{"id":"https://openalex.org/F4320335489","display_name":"Institute for Information and Communications Technology Promotion","ror":"https://ror.org/01g0hqq23"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412887726.pdf","grobid_xml":"https://content.openalex.org/works/W4412887726.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2496295964","https://openalex.org/W1976952689","https://openalex.org/W2336887028","https://openalex.org/W2547431347","https://openalex.org/W1984921740","https://openalex.org/W2070982348","https://openalex.org/W2365485488","https://openalex.org/W2910493550","https://openalex.org/W1911859126","https://openalex.org/W642007152"],"abstract_inverted_index":{"Audio-Visual":[0],"Speech":[1],"Recognition":[2],"(AVSR)":[3],"achieves":[4,114],"robust":[5],"speech":[6,37,48,72,90,143],"recognition":[7],"in":[8],"noisy":[9],"environments":[10],"by":[11,39,136,151,154],"combining":[12],"auditory":[13],"and":[14,82,157],"visual":[15],"information.However,":[16],"recent":[17],"Large":[18],"Language":[19],"Model":[20],"(LLM)":[21],"based":[22,78],"AVSR":[23],"systems":[24],"incur":[25],"high":[26,32],"computational":[27,149],"costs":[28],"due":[29],"to":[30,93,98,139],"the":[31,107,140],"temporal":[33],"resolution":[34],"of":[35,101,120],"audio-visual":[36,71],"processed":[38],"LLMs.In":[40],"this":[41],"work,":[42],"we":[43],"introduce":[44],"an":[45,62,70],"efficient":[46],"multimodal":[47,142],"LLM":[49,144],"framework":[50],"that":[51,74,111],"minimizes":[52],"token":[53,95,134],"length":[54],"while":[55,122],"preserving":[56],"essential":[57],"linguistic":[58],"content.Our":[59],"approach":[60,130],"employs":[61],"early":[63],"AV-fusion":[64],"module":[65],"for":[66],"streamlined":[67],"feature":[68],"integration,":[69],"Q-Former":[73],"dynamically":[75],"allocates":[76],"tokens":[77,126],"on":[79,106],"input":[80],"duration,":[81],"a":[83,89,118],"refined":[84],"query":[85],"allocation":[86,96],"strategy":[87],"with":[88,117],"rate":[91],"predictor":[92],"adjust":[94],"according":[97],"speaking":[99],"speed":[100],"each":[102],"audio":[103],"sample.Extensive":[104],"experiments":[105],"LRS3":[108],"dataset":[109],"show":[110],"our":[112,129],"method":[113],"state-of-the-art":[115],"performance":[116],"WER":[119],"0.72%":[121],"using":[123],"only":[124,132],"3.5":[125],"per":[127],"second.Moreover,":[128],"not":[131],"reduces":[133],"usage":[135],"86%":[137],"compared":[138],"previous":[141],"framework,":[145],"but":[146],"also":[147],"improves":[148],"efficiency":[150],"reducing":[152],"FLOPs":[153],"35.7%.The":[155],"code":[156],"models":[158],"are":[159],"available":[160],"https:":[161],"//github.com/JeongHun0716/MMS-LLaMA.":[162]},"counts_by_year":[{"year":2026,"cited_by_count":3}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-10-10T00:00:00"}