{"id":"https://openalex.org/W7166872196","doi":"https://doi.org/10.18653/v1/2026.findings-acl.2040","title":"Optimizing Conversational Quality in Spoken Dialogue Systems with Reinforcement Learning from AI Feedback","display_name":"Optimizing Conversational Quality in Spoken Dialogue Systems with Reinforcement Learning from AI Feedback","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166872196","doi":"https://doi.org/10.18653/v1/2026.findings-acl.2040"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.2040","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.2040","pdf_url":"https://aclanthology.org/2026.findings-acl.2040.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 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.2040.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5047892839","display_name":"Siddhant Arora","orcid":"https://orcid.org/0000-0003-0375-496X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Siddhant Arora","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139813046","display_name":"Jinchuan Tian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jinchuan Tian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139834805","display_name":"Jiatong Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiatong Shi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5034216052","display_name":"Hayato Futami","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hayato Futami","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109644502","display_name":"Yosuke Kashiwagi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yosuke Kashiwagi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5011937407","display_name":"Emiru Tsunoo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Emiru Tsunoo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139833411","display_name":"Shinji Watanabe","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shinji Watanabe","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.86109116,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"41049","last_page":"41066"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12031","display_name":"Speech and dialogue systems","score":0.945900022983551,"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/T12031","display_name":"Speech and dialogue systems","score":0.945900022983551,"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/T10028","display_name":"Topic Modeling","score":0.011300000362098217,"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/T10667","display_name":"Emotion and Mood Recognition","score":0.006800000090152025,"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/quality","display_name":"Quality (philosophy)","score":0.5896000266075134},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.5478000044822693},{"id":"https://openalex.org/keywords/control","display_name":"Control (management)","score":0.3716000020503998},{"id":"https://openalex.org/keywords/reinforcement","display_name":"Reinforcement","score":0.3628999888896942},{"id":"https://openalex.org/keywords/action","display_name":"Action (physics)","score":0.27219998836517334}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6060000061988831},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.5896000266075134},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5478000044822693},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.46970000863075256},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.39820000529289246},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.3716000020503998},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.3628999888896942},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.2847999930381775},{"id":"https://openalex.org/C2780791683","wikidata":"https://www.wikidata.org/wiki/Q846785","display_name":"Action (physics)","level":2,"score":0.27219998836517334},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.2687000036239624},{"id":"https://openalex.org/C2777200299","wikidata":"https://www.wikidata.org/wiki/Q52943","display_name":"Conversation","level":2,"score":0.2655999958515167},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2639000117778778},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2639000117778778},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.2567000091075897}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.2040","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.2040","pdf_url":"https://aclanthology.org/2026.findings-acl.2040.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 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.2040","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.2040","pdf_url":"https://aclanthology.org/2026.findings-acl.2040.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 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.6263943314552307}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166872196.pdf","grobid_xml":"https://content.openalex.org/works/W7166872196.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1,122],"from":[2],"human":[3],"or":[4],"AI":[5],"feedback":[6],"(RLHF/RLAIF)":[7],"for":[8,83,123,175],"speech-in/speechout":[9],"dialogue":[10,59],"systems":[11,60],"(SDS)":[12],"remains":[13],"underexplored,":[14],"with":[15,56,77,94],"prior":[16],"work":[17],"largely":[18],"limited":[19],"to":[20,141],"single":[21,112],"semantic":[22,41,162],"rewards":[23],"applied":[24],"at":[25],"the":[26,31,78,116,169],"utterance":[27],"level.Such":[28],"setups":[29],"overlook":[30],"multidimensional":[32],"and":[33,49,88,106,131,135,164],"multi-modal":[34],"nature":[35],"of":[36,120,171],"conversational":[37,177],"quality,":[38],"which":[39],"encompasses":[40],"coherence,":[42],"audio":[43,165],"naturalness,":[44],"speaker":[45],"consistency,":[46],"emotion":[47],"alignment,":[48],"turn-taking":[50],"behavior.Moreover,":[51],"they":[52],"are":[53],"fundamentally":[54],"mismatched":[55],"duplex":[57,99,133],"spoken":[58],"that":[61,146],"generate":[62],"responses":[63],"incrementally,":[64],"where":[65],"agents":[66],"must":[67],"make":[68],"decisions":[69],"based":[70],"on":[71],"partial":[72],"utterances.We":[73],"address":[74],"these":[75],"limitations":[76],"first":[79,117],"multi-reward":[80,138,156,173],"RLAIF":[81,148],"framework":[82],"SDS,":[84],"combining":[85],"semantic,":[86],"audio-quality,":[87],"emotion-consistency":[89],"rewards.To":[90],"align":[91],"utterance-level":[92],"preferences":[93],"incremental,":[95],"blockwise":[96,132],"decoding":[97],"in":[98,127],"models,":[100,134],"we":[101],"apply":[102],"turn-level":[103],"preference":[104,121],"sampling":[105],"aggregate":[107],"per-block":[108],"log-probabilities":[109],"within":[110],"a":[111,137],"DPO":[113,139],"objective.We":[114],"present":[115],"systematic":[118],"study":[119],"improving":[124],"SDS":[125],"quality":[126,163],"both":[128],"multi-turn":[129],"Chain-of-Thought":[130],"release":[136],"dataset":[140],"support":[142],"reproducible":[143],"research.Experiments":[144],"show":[145],"singlereward":[147],"selectively":[149],"improves":[150],"its":[151],"targeted":[152],"metric,":[153],"while":[154],"joint":[155],"training":[157],"yields":[158],"consistent":[159],"gains":[160],"across":[161],"naturalness.These":[166],"results":[167],"highlight":[168],"importance":[170],"holistic,":[172],"alignment":[174],"practical":[176],"SDS.":[178]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}