{"id":"b697793e9087ffd8c2e13165dc16f2897cd228e4897fca228c32c25344224943","title":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","argument":"The authors of Confucius4-R2T2 present a streaming ASR model designed not to revise words it has already output. They argue this matters because translators and voice agents may act as each token appears: a later revision can waste work or trigger a wrong action. The model is made by post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio, while freezing its pre-trained audio encoder and projector. The authors introduce Longest Stable Prefix (LSP), a way to label output prefixes that remain stable as more audio arrives. They say learning from LSP labels makes a full-context model into a stable-prefix streaming model without changing its architecture. Their abstract reports streaming recognition close to the base model's full-context quality, with stable text and low emission lag, and says gains extend beyond the post-training languages. This suggests a route to speech recognition that machines can use as speech arrives, rather than waiting for the full audio; the paper's abstract does not give the measurements behind those reported results. The authors say they open-source the model weights and code.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the arXiv abstract, which reports evaluation outcomes but gives no quantitative results or evaluation details to assess those outcomes.","speaker":{"name":"Huang, Daniel, He, Qi, Duan, Shaoxiong, Li, Yingjie, Cheng, Qiao, Duan, Yitao","roleAtSource":"Article author"},"source":{"title":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.06043","publishedAt":"2026-11-04","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors present Confucius4-R2T2 as a streaming ASR model that does not revise earlier output while aiming for high accuracy and low latency.","excerpt":"a stable-prefix streaming ASR model that never modifies its earlier output while maintaining high accuracy and low latency.","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"},{"text":"They post-train Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio, freezing its pre-trained audio encoder and projector.","excerpt":"post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio while freezing its pre-trained audio encoder and projector.","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"},{"text":"The authors say learning from Longest Stable Prefix labels enables streaming without an architectural change.","excerpt":"Learning from LSP labels converts a full-context model into a stable-prefix streaming one without any architectural change.","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"},{"text":"The abstract reports recognition close to the base model's full-context quality, low emission lag, and gains beyond the post-training languages.","excerpt":"R2T2 brings streaming recognition close to the base model's full-context quality while delivering stable text with low emission lag. Its gains extend beyond the post-training languages","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"},{"text":"Because machines may act on each token as it appears, revisions can waste work or lead to a wrong action; stable output is intended to address that problem.","excerpt":"A token that is later revised costs wasted work or a wrong action.","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.06043","checkedAt":"2026-11-10T02:44:08.170Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What evaluation results quantify how close R2T2 comes to the base model's full-context quality and how low its emission lag is?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","summary":"The authors of Confucius4-R2T2 present a streaming ASR model designed not to revise words it has already output. They argue this matters because translators and voice agents may act as each token appears: a later revision can waste work or trigger a wrong action. The model is made by post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio, while freezing its pre-trained audio encoder and projector. The authors introduce Longest Stable Prefix (LSP), a way to label output prefixes that remain stable as more audio arrives. They say learning from LSP labels makes a full-context model into a stable-prefix streaming model without changing its architecture. Their abstract reports streaming recognition close to the base model's full-context quality, with stable text and low emission lag, and says gains extend beyond the post-training languages. This suggests a route to speech recognition that machines can use as speech arrives, rather than waiting for the full audio; the paper's abstract does not give the measurements behind those reported results. The authors say they open-source the model weights and code.","question":"What evaluation results quantify how close R2T2 comes to the base model's full-context quality and how low its emission lag is?","claims":[{"text":"The authors present Confucius4-R2T2 as a streaming ASR model that does not revise earlier output while aiming for high accuracy and low latency.","excerpt":"a stable-prefix streaming ASR model that never modifies its earlier output while maintaining high accuracy and low latency.","support":"source-stated"},{"text":"They post-train Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio, freezing its pre-trained audio encoder and projector.","excerpt":"post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio while freezing its pre-trained audio encoder and projector.","support":"source-stated"},{"text":"The authors say learning from Longest Stable Prefix labels enables streaming without an architectural change.","excerpt":"Learning from LSP labels converts a full-context model into a stable-prefix streaming one without any architectural change.","support":"source-stated"},{"text":"The abstract reports recognition close to the base model's full-context quality, low emission lag, and gains beyond the post-training languages.","excerpt":"R2T2 brings streaming recognition close to the base model's full-context quality while delivering stable text with low emission lag. Its gains extend beyond the post-training languages","support":"source-stated"},{"text":"Because machines may act on each token as it appears, revisions can waste work or lead to a wrong action; stable output is intended to address that problem.","excerpt":"A token that is later revised costs wasted work or a wrong action.","support":"source-stated"}],"limitations":["The supplied evidence is the arXiv abstract, which reports evaluation outcomes but gives no quantitative results or evaluation details to assess those outcomes."],"mediaId":"e5dd910470b46d8ecd42b7bb736b0481ad7fec0441f6614bff60133a8a614b5b","correctionReason":null,"id":"b697793e9087ffd8c2e13165dc16f2897cd228e4897fca228c32c25344224943","publicationId":"18508383-94a7-4829-8147-e9d4854da441","version":1,"sourceUrl":"https://arxiv.org/abs/2611.06043","sourceTitle":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-04","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-05T05:06:41.791Z","createdAt":"2026-11-10T01:40:24.576Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"e5dd910470b46d8ecd42b7bb736b0481ad7fec0441f6614bff60133a8a614b5b","kind":"article","url":"https://arxiv.org/abs/2611.06043","sourceUrl":"https://arxiv.org/abs/2611.06043","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-10T01:39:45.320Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.06043","checkedAt":"2026-11-10T02:44:08.170Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"e5dd910470b46d8ecd42b7bb736b0481ad7fec0441f6614bff60133a8a614b5b","confidence":0.95,"provider":"typesafe","model":"jev-1.13.0","inputHash":"5990c67a963089b989a7e819d81dc860656dcc8ea6664a5e85f03908d922aece","decidedAt":"2026-11-10T01:39:47.193Z"},"provenance":{"textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observationId":"c6a1759febe66614adb8a4fa491866ab5bd12c2bf5210b032f29941602fa22e9","retrievedAt":"2026-11-10T01:39:45.025Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"7ace5af9-8842-4445-860f-55595e06b5b4"},"reviewer":{"model":"gpt-6-luna","runId":"596eb43e-b5ab-4d6e-86c2-22dc54e5333d"}},"audit":{"status":"passed","reviewedAt":"2026-11-10T01:40:24.576Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Huang, Daniel","url":null,"kind":"unknown"},{"name":"He, Qi","url":null,"kind":"unknown"},{"name":"Duan, Shaoxiong","url":null,"kind":"unknown"},{"name":"Li, Yingjie","url":null,"kind":"unknown"},{"name":"Cheng, Qiao","url":null,"kind":"unknown"},{"name":"Duan, Yitao","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Huang, Daniel","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Huang, Daniel","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"He, Qi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"He, Qi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"Duan, Shaoxiong","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Duan, Shaoxiong","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"Li, Yingjie","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Li, Yingjie","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"Cheng, Qiao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Cheng, Qiao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"Duan, Yitao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Duan, Yitao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.06043","snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z","basis":"publisher-page-metadata","sourceLinks":48,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/b697793e9087ffd8c2e13165dc16f2897cd228e4897fca228c32c25344224943/source"}},"canonicalUrl":"https://escapemeridian.com/readings/confucius4-r2t2-a-stable-prefix-streaming-asr-model-for-real-b697793e","formats":{"html":"/readings/confucius4-r2t2-a-stable-prefix-streaming-asr-model-for-real-b697793e","markdown":"/readings/confucius4-r2t2-a-stable-prefix-streaming-asr-model-for-real-b697793e.md","json":"/readings/confucius4-r2t2-a-stable-prefix-streaming-asr-model-for-real-b697793e.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-pre-training","name":"Pre-training","explanation":"A model's initial training on a large dataset, before further adaptation for particular uses.","referenceUrl":"https://developers.google.com/machine-learning/glossary#pre-training"},{"id":"term-model-weights","name":"Model weights","explanation":"The learned numbers that make up a trained model; publishing them lets others run it.","referenceUrl":"https://developers.google.com/machine-learning/glossary#weight"}],"entities":[{"id":"time","name":"TIME","kind":"organization","context":"TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.","referenceUrls":["https://time.com/"],"path":"/entities/time","json":"/entities/time.json","markdown":"/entities/time.md"},{"id":"arxiv","name":"arXiv","kind":"organization","context":"arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.","referenceUrls":["https://arxiv.org/","https://info.arxiv.org/help/api/index.html"],"path":"/entities/arxiv","json":"/entities/arxiv.json","markdown":"/entities/arxiv.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"b697793e9087ffd8c2e13165dc16f2897cd228e4897fca228c32c25344224943","revision":"18508383-94a7-4829-8147-e9d4854da441","revisionKind":"publication-id","title":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","url":"https://escapemeridian.com/readings/confucius4-r2t2-a-stable-prefix-streaming-asr-model-for-real-b697793e"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.06043","revision":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","revisionKind":"sha256-source-text","title":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","url":"https://arxiv.org/abs/2611.06043"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"b697793e9087ffd8c2e13165dc16f2897cd228e4897fca228c32c25344224943","title":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","argument":"The authors of Confucius4-R2T2 present a streaming ASR model designed not to revise words it has already output. They argue this matters because translators and voice agents may act as each token appears: a later revision can waste work or trigger a wrong action. The model is made by post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio, while freezing its pre-trained audio encoder and projector. The authors introduce Longest Stable Prefix (LSP), a way to label output prefixes that remain stable as more audio arrives. They say learning from LSP labels makes a full-context model into a stable-prefix streaming model without changing its architecture. Their abstract reports streaming recognition close to the base model's full-context quality, with stable text and low emission lag, and says gains extend beyond the post-training languages. This suggests a route to speech recognition that machines can use as speech arrives, rather than waiting for the full audio; the paper's abstract does not give the measurements behind those reported results. The authors say they open-source the model weights and code.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the arXiv abstract, which reports evaluation outcomes but gives no quantitative results or evaluation details to assess those outcomes.","speaker":{"name":"Huang, Daniel, He, Qi, Duan, Shaoxiong, Li, Yingjie, Cheng, Qiao, Duan, Yitao","roleAtSource":"Article author"},"source":{"title":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.06043","publishedAt":"2026-11-04","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors present Confucius4-R2T2 as a streaming ASR model that does not revise earlier output while aiming for high accuracy and low latency.","excerpt":"a stable-prefix streaming ASR model that never modifies its earlier output while maintaining high accuracy and low latency.","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"},{"text":"They post-train Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio, freezing its pre-trained audio encoder and projector.","excerpt":"post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio while freezing its pre-trained audio encoder and projector.","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"},{"text":"The authors say learning from Longest Stable Prefix labels enables streaming without an architectural change.","excerpt":"Learning from LSP labels converts a full-context model into a stable-prefix streaming one without any architectural change.","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"},{"text":"The abstract reports recognition close to the base model's full-context quality, low emission lag, and gains beyond the post-training languages.","excerpt":"R2T2 brings streaming recognition close to the base model's full-context quality while delivering stable text with low emission lag. Its gains extend beyond the post-training languages","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"},{"text":"Because machines may act on each token as it appears, revisions can waste work or lead to a wrong action; stable output is intended to address that problem.","excerpt":"A token that is later revised costs wasted work or a wrong action.","url":"https://arxiv.org/abs/2611.06043","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.06043","checkedAt":"2026-11-10T02:44:08.170Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What evaluation results quantify how close R2T2 comes to the base model's full-context quality and how low its emission lag is?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","summary":"The authors of Confucius4-R2T2 present a streaming ASR model designed not to revise words it has already output. They argue this matters because translators and voice agents may act as each token appears: a later revision can waste work or trigger a wrong action. The model is made by post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio, while freezing its pre-trained audio encoder and projector. The authors introduce Longest Stable Prefix (LSP), a way to label output prefixes that remain stable as more audio arrives. They say learning from LSP labels makes a full-context model into a stable-prefix streaming model without changing its architecture. Their abstract reports streaming recognition close to the base model's full-context quality, with stable text and low emission lag, and says gains extend beyond the post-training languages. This suggests a route to speech recognition that machines can use as speech arrives, rather than waiting for the full audio; the paper's abstract does not give the measurements behind those reported results. The authors say they open-source the model weights and code.","question":"What evaluation results quantify how close R2T2 comes to the base model's full-context quality and how low its emission lag is?","claims":[{"text":"The authors present Confucius4-R2T2 as a streaming ASR model that does not revise earlier output while aiming for high accuracy and low latency.","excerpt":"a stable-prefix streaming ASR model that never modifies its earlier output while maintaining high accuracy and low latency.","support":"source-stated"},{"text":"They post-train Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio, freezing its pre-trained audio encoder and projector.","excerpt":"post-training Qwen3-ASR-1.7B on 1,500 hours of Chinese and English audio while freezing its pre-trained audio encoder and projector.","support":"source-stated"},{"text":"The authors say learning from Longest Stable Prefix labels enables streaming without an architectural change.","excerpt":"Learning from LSP labels converts a full-context model into a stable-prefix streaming one without any architectural change.","support":"source-stated"},{"text":"The abstract reports recognition close to the base model's full-context quality, low emission lag, and gains beyond the post-training languages.","excerpt":"R2T2 brings streaming recognition close to the base model's full-context quality while delivering stable text with low emission lag. Its gains extend beyond the post-training languages","support":"source-stated"},{"text":"Because machines may act on each token as it appears, revisions can waste work or lead to a wrong action; stable output is intended to address that problem.","excerpt":"A token that is later revised costs wasted work or a wrong action.","support":"source-stated"}],"limitations":["The supplied evidence is the arXiv abstract, which reports evaluation outcomes but gives no quantitative results or evaluation details to assess those outcomes."],"mediaId":"e5dd910470b46d8ecd42b7bb736b0481ad7fec0441f6614bff60133a8a614b5b","correctionReason":null,"id":"b697793e9087ffd8c2e13165dc16f2897cd228e4897fca228c32c25344224943","publicationId":"18508383-94a7-4829-8147-e9d4854da441","version":1,"sourceUrl":"https://arxiv.org/abs/2611.06043","sourceTitle":"Confucius4-R2T2: A Stable-Prefix Streaming ASR Model for Real Real-Time Transcription","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-04","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-05T05:06:41.791Z","createdAt":"2026-11-10T01:40:24.576Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"e5dd910470b46d8ecd42b7bb736b0481ad7fec0441f6614bff60133a8a614b5b","kind":"article","url":"https://arxiv.org/abs/2611.06043","sourceUrl":"https://arxiv.org/abs/2611.06043","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-10T01:39:45.320Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.06043","checkedAt":"2026-11-10T02:44:08.170Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"e5dd910470b46d8ecd42b7bb736b0481ad7fec0441f6614bff60133a8a614b5b","confidence":0.95,"provider":"typesafe","model":"jev-1.13.0","inputHash":"5990c67a963089b989a7e819d81dc860656dcc8ea6664a5e85f03908d922aece","decidedAt":"2026-11-10T01:39:47.193Z"},"provenance":{"textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observationId":"c6a1759febe66614adb8a4fa491866ab5bd12c2bf5210b032f29941602fa22e9","retrievedAt":"2026-11-10T01:39:45.025Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"7ace5af9-8842-4445-860f-55595e06b5b4"},"reviewer":{"model":"gpt-6-luna","runId":"596eb43e-b5ab-4d6e-86c2-22dc54e5333d"}},"audit":{"status":"passed","reviewedAt":"2026-11-10T01:40:24.576Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Huang, Daniel","url":null,"kind":"unknown"},{"name":"He, Qi","url":null,"kind":"unknown"},{"name":"Duan, Shaoxiong","url":null,"kind":"unknown"},{"name":"Li, Yingjie","url":null,"kind":"unknown"},{"name":"Cheng, Qiao","url":null,"kind":"unknown"},{"name":"Duan, Yitao","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Huang, Daniel","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Huang, Daniel","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"He, Qi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"He, Qi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"Duan, Shaoxiong","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Duan, Shaoxiong","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"Li, Yingjie","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Li, Yingjie","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"Cheng, Qiao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Cheng, Qiao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}},{"actor":{"name":"Duan, Yitao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Duan, Yitao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.06043","locator":"head/meta[name=author|citation_author]"},"assertionId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-10T02:44:08.170Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","sourceUrl":"https://arxiv.org/abs/2611.06043","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.06043","snapshotId":"b831bdf5fd69fc14c4a378c3516bbcc1a138944263a3822119a1fcfa967e0a14","contentSha256":"eaf676bc0758fa73e3f3af8f0d1e13a36a2eaeb6956750f234aec310ec5056ba","textHash":"051d8ac507da1ae033195ab696194734420b223714bd5c32d41fc2fdb1bee884","observedAt":"2026-11-10T02:44:08.170Z","basis":"publisher-page-metadata","sourceLinks":48,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/b697793e9087ffd8c2e13165dc16f2897cd228e4897fca228c32c25344224943/source"}}}],"audits":[{"publicationId":"18508383-94a7-4829-8147-e9d4854da441","verdict":"passed","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the model, the LSP concept and the reported results to the paper's authors. Evaluation outcomes are framed as what the abstract reports, not as established facts. This matches the arXiv listing by Daniel Huang et al.","excerpt":"Authors:Daniel Huang, Qi He, Shaoxiong Duan, Yingjie Li, Qiao Cheng, Yitao Duan"},{"dimension":"summary","status":"pass","explanation":"The summary leads with the development: a streaming speech recognition model that does not revise output it has already produced. It gives the authors' motivation in their own terms (wasted work or a wrong action), the training method, the LSP idea, the reported results and the open release. Plain-language glosses (not revising already-output words; usable as speech arrives instead of after the full audio) make the core idea accessible. Remaining terms such as token and emission lag can be handled by tooltips. The question addresses a real gap in the abstract.","excerpt":"A token that is later revised costs wasted work or a wrong action."},{"dimension":"date","status":"pass","explanation":"The originalPublished date, 2026-11-04, matches the v1 submission date on the arXiv page. It is not the later feed date of 2026-11-05.","excerpt":"[Submitted on 4 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim is supported by abstract text. The closing significance sentence tracks the authors' own 'practical path' claim and is hedged with 'suggests'. It also notes that the abstract gives no measurements. No invented benefits, costs or numbers appear. The limitation statement accurately describes the evidence.","excerpt":"demonstrating a practical path from full-context ASR to stable streaming"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct arXiv abstract page for this paper, which is the original source. No better source media was found.","excerpt":null}],"createdAt":"2026-11-10T01:40:24.576Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}