{"id":"https://openalex.org/W7138270481","doi":"https://doi.org/10.48550/arxiv.2603.13952","title":"LLM-Guided Reinforcement Learning for Audio-Visual Speech Enhancement","display_name":"LLM-Guided Reinforcement Learning for Audio-Visual Speech Enhancement","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138270481","doi":"https://doi.org/10.48550/arxiv.2603.13952"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.13952","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13952","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2603.13952","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5013577711","display_name":"Chih\u2010Ning Chen","orcid":"https://orcid.org/0000-0002-3274-3488"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Chen, Chih-Ning","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5058567113","display_name":"Jen-Cheng Hou","orcid":"https://orcid.org/0000-0002-4010-0220"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hou, Jen-Cheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129680777","display_name":"Hsin-Min Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Hsin-Min","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009305504","display_name":"Shao\u2010Yi Chien","orcid":"https://orcid.org/0000-0002-0634-6294"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chien, Shao-Yi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5129725982","display_name":"Yu Tsao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tsao, Yu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5129644098","display_name":"Fan-Gang Zeng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Fan-Gang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5013577711"],"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/T10860","display_name":"Speech and Audio Processing","score":0.977400004863739,"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":0.977400004863739,"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/T10283","display_name":"Hearing Loss and Rehabilitation","score":0.006200000178068876,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T11233","display_name":"Advanced Adaptive Filtering Techniques","score":0.006200000178068876,"subfield":{"id":"https://openalex.org/subfields/2206","display_name":"Computational Mechanics"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/interpretability","display_name":"Interpretability","score":0.8424000144004822},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.6093000173568726},{"id":"https://openalex.org/keywords/active-listening","display_name":"Active listening","score":0.5716000199317932},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.5024999976158142},{"id":"https://openalex.org/keywords/speech-enhancement","display_name":"Speech enhancement","score":0.4575999975204468},{"id":"https://openalex.org/keywords/mean-squared-error","display_name":"Mean squared error","score":0.4361000061035156},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.36469998955726624},{"id":"https://openalex.org/keywords/supervised-learning","display_name":"Supervised learning","score":0.3544999957084656}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.8424000144004822},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7088000178337097},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6093000173568726},{"id":"https://openalex.org/C177291462","wikidata":"https://www.wikidata.org/wiki/Q423038","display_name":"Active listening","level":2,"score":0.5716000199317932},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.545799970626831},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5450000166893005},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.5024999976158142},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4643000066280365},{"id":"https://openalex.org/C2776182073","wikidata":"https://www.wikidata.org/wiki/Q7575395","display_name":"Speech enhancement","level":3,"score":0.4575999975204468},{"id":"https://openalex.org/C139945424","wikidata":"https://www.wikidata.org/wiki/Q1940696","display_name":"Mean squared error","level":2,"score":0.4361000061035156},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.36469998955726624},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.36169999837875366},{"id":"https://openalex.org/C136389625","wikidata":"https://www.wikidata.org/wiki/Q334384","display_name":"Supervised learning","level":3,"score":0.3544999957084656},{"id":"https://openalex.org/C26760741","wikidata":"https://www.wikidata.org/wiki/Q160402","display_name":"Perception","level":2,"score":0.329800009727478},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.328900009393692},{"id":"https://openalex.org/C2777601683","wikidata":"https://www.wikidata.org/wiki/Q6499736","display_name":"Vocabulary","level":2,"score":0.3221000134944916},{"id":"https://openalex.org/C167085575","wikidata":"https://www.wikidata.org/wiki/Q6803654","display_name":"Mean squared prediction error","level":2,"score":0.3163999915122986},{"id":"https://openalex.org/C167310288","wikidata":"https://www.wikidata.org/wiki/Q7564808","display_name":"Sound quality","level":2,"score":0.3147999942302704},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.303600013256073},{"id":"https://openalex.org/C74672266","wikidata":"https://www.wikidata.org/wiki/Q815859","display_name":"Language acquisition","level":2,"score":0.2973000109195709},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.29010000824928284},{"id":"https://openalex.org/C207609745","wikidata":"https://www.wikidata.org/wiki/Q4944086","display_name":"Bootstrapping (finance)","level":2,"score":0.2847999930381775},{"id":"https://openalex.org/C61328038","wikidata":"https://www.wikidata.org/wiki/Q3358061","display_name":"Speech processing","level":2,"score":0.27300000190734863}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.13952","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13952","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2603.13952","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.13952","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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":"article"},"sustainable_development_goals":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.5433067083358765}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"In":[0],"existing":[1],"Audio-Visual":[2],"Speech":[3],"Enhancement":[4],"(AVSE)":[5],"methods,":[6],"objectives":[7],"such":[8],"as":[9,77],"Scale-Invariant":[10],"Signal-to-Noise":[11],"Ratio":[12],"(SI-SNR)":[13],"and":[14,30,96,121,132],"Mean":[15],"Squared":[16],"Error":[17],"(MSE)":[18],"are":[19,64],"widely":[20],"used;":[21],"however,":[22],"they":[23],"often":[24],"correlate":[25],"poorly":[26],"with":[27,44,88],"perceptual":[28],"quality":[29,130],"provide":[31],"limited":[32],"interpretability":[33],"for":[34,81],"optimization.":[35],"This":[36],"work":[37],"proposes":[38],"a":[39,45,67,72,83,118,122],"reinforcement":[40],"learning-based":[41],"AVSE":[42,85,108],"framework":[43],"Large":[46],"Language":[47],"Model":[48],"(LLM)-based":[49],"interpretable":[50],"reward":[51,80],"model.":[52,86],"An":[53],"audio":[54],"LLM":[55],"generates":[56],"natural":[57],"language":[58],"descriptions":[59],"of":[60],"enhanced":[61],"speech,":[62],"which":[63],"converted":[65],"by":[66],"sentiment":[68],"analysis":[69],"model":[70],"into":[71],"1-5":[73],"rating":[74],"score":[75],"serving":[76],"the":[78,105,114],"PPO":[79],"fine-tuning":[82],"pretrained":[84],"Compared":[87],"scalar":[89],"metrics,":[90,131],"LLM-generated":[91],"feedback":[92],"is":[93],"semantically":[94],"rich":[95],"explicitly":[97],"describes":[98],"improvements":[99],"in":[100,126],"speech":[101],"quality.":[102],"Experiments":[103],"on":[104],"4th":[106],"COG-MHEAR":[107],"Challenge":[109],"(AVSEC-4)":[110],"dataset":[111],"show":[112],"that":[113],"proposed":[115],"method":[116],"outperforms":[117],"supervised":[119],"baseline":[120,125],"DNSMOS-based":[123],"RL":[124],"PESQ,":[127],"STOI,":[128],"neural":[129],"subjective":[133],"listening":[134],"tests.":[135]},"counts_by_year":[],"updated_date":"2026-05-05T08:41:31.759640","created_date":"2026-03-18T00:00:00"}