{"id":"https://openalex.org/W7138085134","doi":"https://doi.org/10.1609/aaai.v40i36.40323","title":"Boosting ASR Robustness via Test-Time Reinforcement Learning with Audio-Text Semantic Rewards","display_name":"Boosting ASR Robustness via Test-Time Reinforcement Learning with Audio-Text Semantic Rewards","publication_year":2026,"publication_date":"2026-03-14","ids":{"openalex":"https://openalex.org/W7138085134","doi":"https://doi.org/10.1609/aaai.v40i36.40323"},"language":null,"primary_location":{"id":"doi:10.1609/aaai.v40i36.40323","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i36.40323","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://doi.org/10.1609/aaai.v40i36.40323","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Linghan Fang","orcid":null},"institutions":[{"id":"https://openalex.org/I62916508","display_name":"Technical University of Munich","ror":"https://ror.org/02kkvpp62","country_code":"DE","type":"education","lineage":["https://openalex.org/I62916508"]},{"id":"https://openalex.org/I889458895","display_name":"University of Hong Kong","ror":"https://ror.org/02zhqgq86","country_code":"HK","type":"education","lineage":["https://openalex.org/I889458895"]}],"countries":["DE","HK"],"is_corresponding":true,"raw_author_name":"Linghan Fang","raw_affiliation_strings":["The Hong Kong University of Science and Technology (Guangzhou)\nTechnische Universit\u00e4t M\u00fcnchen"],"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology (Guangzhou)\nTechnische Universit\u00e4t M\u00fcnchen","institution_ids":["https://openalex.org/I62916508","https://openalex.org/I889458895"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Tianxin Xie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tianxin Xie","raw_affiliation_strings":["The Hong Kong University of Science and Technology (Guangzhou)"],"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology (Guangzhou)","institution_ids":[]}]},{"author_position":"last","author":{"id":null,"display_name":"Li Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li Liu","raw_affiliation_strings":["The Hong Kong University of Science and Technology (Guangzhou)"],"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology (Guangzhou)","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I62916508","https://openalex.org/I889458895"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.30460385,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"40","issue":"36","first_page":"30673","last_page":"30681"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.8352000117301941,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.8352000117301941,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.021299999207258224,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.019200000911951065,"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/reinforcement-learning","display_name":"Reinforcement learning","score":0.7128999829292297},{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.5730999708175659},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5573999881744385},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.539900004863739},{"id":"https://openalex.org/keywords/adaptability","display_name":"Adaptability","score":0.5285000205039978},{"id":"https://openalex.org/keywords/decoding-methods","display_name":"Decoding methods","score":0.4925000071525574},{"id":"https://openalex.org/keywords/entropy","display_name":"Entropy (arrow of time)","score":0.3801000118255615},{"id":"https://openalex.org/keywords/principle-of-maximum-entropy","display_name":"Principle of maximum entropy","score":0.32710000872612}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7735000252723694},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.7128999829292297},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.5730999708175659},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5573999881744385},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5559999942779541},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.539900004863739},{"id":"https://openalex.org/C177606310","wikidata":"https://www.wikidata.org/wiki/Q5674297","display_name":"Adaptability","level":2,"score":0.5285000205039978},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5016999840736389},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.4925000071525574},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.3801000118255615},{"id":"https://openalex.org/C9679016","wikidata":"https://www.wikidata.org/wiki/Q1417473","display_name":"Principle of maximum entropy","level":2,"score":0.32710000872612},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.2987000048160553},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.2976999878883362},{"id":"https://openalex.org/C160920958","wikidata":"https://www.wikidata.org/wiki/Q7662746","display_name":"Synthetic data","level":2,"score":0.2865999937057495},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.2847999930381775},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.2815000116825104},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.271699994802475},{"id":"https://openalex.org/C23224414","wikidata":"https://www.wikidata.org/wiki/Q176769","display_name":"Hidden Markov model","level":2,"score":0.2621000111103058},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.2574000060558319},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2533000111579895},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.25290000438690186},{"id":"https://openalex.org/C2776321320","wikidata":"https://www.wikidata.org/wiki/Q857525","display_name":"Annotation","level":2,"score":0.2515000104904175}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1609/aaai.v40i36.40323","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i36.40323","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1609/aaai.v40i36.40323","is_oa":true,"landing_page_url":"https://doi.org/10.1609/aaai.v40i36.40323","pdf_url":null,"source":{"id":"https://openalex.org/S4210191458","display_name":"Proceedings of the AAAI Conference on Artificial Intelligence","issn_l":"2159-5399","issn":["2159-5399","2374-3468"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/P4310320058","host_organization_name":"Association for the Advancement of Artificial Intelligence","host_organization_lineage":["https://openalex.org/P4310320058"],"host_organization_lineage_names":["Association for the Advancement of Artificial Intelligence"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the AAAI Conference on Artificial Intelligence","raw_type":"journal-article"},"sustainable_development_goals":[{"score":0.40919700264930725,"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recently,":[0],"Automatic":[1],"Speech":[2],"Recognition":[3],"(ASR)":[4],"systems":[5,214],"(e.g.,":[6],"Whisper)":[7],"have":[8],"achieved":[9],"remarkable":[10],"accuracy":[11,179],"improvements":[12],"but":[13],"remain":[14],"highly":[15],"sensitive":[16],"to":[17,81,118,141],"real-world":[18,217],"unseen":[19],"data":[20],"(data":[21],"with":[22,155],"large":[23],"distribution":[24],"shifts),":[25],"including":[26,173],"noisy":[27],"environments":[28],"and":[29,54,113,135,145,158,175,180,192,200,208],"diverse":[30,120],"accents.":[31],"To":[32,87],"address":[33],"this":[34],"issue,":[35],"test-time":[36],"adaptation":[37],"(TTA)":[38],"has":[39],"shown":[40],"great":[41],"potential":[42],"in":[43,177,215],"improving":[44],"the":[45,136,187],"model":[46,68,129,144],"adaptability":[47],"at":[48],"inference":[49,181],"time":[50],"without":[51],"ground-truth":[52],"labels,":[53],"existing":[55,170],"TTA":[56],"methods":[57,75],"often":[58],"rely":[59],"on":[60,153],"pseudo-labeling":[61],"or":[62],"entropy":[63],"minimization.":[64],"However,":[65],"by":[66,101,126],"treating":[67],"confidence":[69],"as":[70],"a":[71,94,109,127,206],"learning":[72],"signal,":[73],"these":[74,89],"may":[76],"reinforce":[77],"high-confidence":[78],"errors,":[79],"leading":[80],"confirmation":[82],"bias":[83],"that":[84,130,165],"undermines":[85],"adaptation.":[86],"overcome":[88],"limitations,":[90],"we":[91],"present":[92],"ASR-TRA,":[93],"novel":[95],"Test-time":[96],"Reinforcement":[97],"Adaptation":[98],"framework":[99],"inspired":[100],"causal":[102],"intervention.":[103],"More":[104],"precisely,":[105],"our":[106,166,196,203],"method":[107,167],"introduces":[108],"learnable":[110],"decoder":[111],"prompt":[112,146],"utilizes":[114],"temperature-controlled":[115],"stochastic":[116],"decoding":[117],"generate":[119],"transcription":[121],"candidates.":[122],"These":[123],"are":[124],"scored":[125],"reward":[128],"measures":[131],"audio-text":[132],"semantic":[133],"alignment,":[134],"resulting":[137],"feedback":[138],"is":[139],"used":[140],"update":[142],"both":[143,178],"parameters":[147],"via":[148],"reinforcement":[149],"learning.":[150],"Comprehensive":[151],"experiments":[152],"LibriSpeech":[154],"synthetic":[156],"noise":[157],"L2":[159],"Arctic":[160],"accented":[161],"English":[162],"datasets":[163],"demonstrate":[164],"significantly":[168],"outperforms":[169],"state-of-the-art":[171],"(SOTA),":[172],"SUTA":[174],"SGEM,":[176],"speed.":[182],"Ablation":[183],"studies":[184],"further":[185],"confirm":[186],"effectiveness":[188],"of":[189],"combining":[190],"audio":[191],"language-based":[193],"rewards,":[194],"highlighting":[195],"method's":[197],"enhanced":[198],"stability":[199],"interpretability.":[201],"Overall,":[202],"approach":[204],"provides":[205],"practical":[207],"robust":[209],"solution":[210],"for":[211],"deploying":[212],"ASR":[213],"challenging":[216],"conditions.":[218]},"counts_by_year":[],"updated_date":"2026-03-20T20:47:17.329874","created_date":"2026-03-10T00:00:00"}