Tweet: <a href=\"https://x.com/HuggingPapers/status/2074384562952749254\" rel=\"nofollow\">https://x.com/HuggingPapers/status/2074384562952749254</a></p>\n","updatedAt":"2026-07-07T13:16:34.708Z","author":{"_id":"5f1158120c833276f61f1a84","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1608042047613-5f1158120c833276f61f1a84.jpeg","fullname":"Niels Rogge","name":"nielsr","type":"user","isPro":false,"isHf":true,"isHfAdmin":false,"isMod":false,"followerCount":1249,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.463441401720047},"editors":["nielsr"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1608042047613-5f1158120c833276f61f1a84.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.05196","authors":[{"_id":"6a4c700225849b193a8340d8","name":"Zhifeng Kong","hidden":false},{"_id":"6a4c700225849b193a8340d9","user":{"_id":"628dba7ea13c4b25ff786ec6","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1671164940704-628dba7ea13c4b25ff786ec6.jpeg","isPro":false,"fullname":"Sang-gil Lee","user":"L0SG","type":"user","name":"L0SG"},"name":"Sang-gil Lee","status":"claimed_verified","statusLastChangedAt":"2026-07-07T12:11:42.973Z","hidden":false},{"_id":"6a4c700225849b193a8340da","name":"Jaehyeon Kim","hidden":false},{"_id":"6a4c700225849b193a8340db","name":"Boxin Wang","hidden":false},{"_id":"6a4c700225849b193a8340dc","name":"Zihan Liu","hidden":false},{"_id":"6a4c700225849b193a8340dd","name":"Sungwon Kim","hidden":false},{"_id":"6a4c700225849b193a8340de","name":"Yang Chen","hidden":false},{"_id":"6a4c700225849b193a8340df","name":"Arushi Goel","hidden":false},{"_id":"6a4c700225849b193a8340e0","name":"Rajarshi Roy","hidden":false},{"_id":"6a4c700225849b193a8340e1","name":"Wenliang Dai","hidden":false},{"_id":"6a4c700225849b193a8340e2","name":"Zhuolin Yang","hidden":false},{"_id":"6a4c700225849b193a8340e3","name":"Yangyi Chen","hidden":false},{"_id":"6a4c700225849b193a8340e4","name":"Dongfu Jiang","hidden":false},{"_id":"6a4c700225849b193a8340e5","name":"Sreyan Ghosh","hidden":false},{"_id":"6a4c700225849b193a8340e6","name":"Tuomas Rintamaki","hidden":false},{"_id":"6a4c700225849b193a8340e7","name":"Andrew Tao","hidden":false},{"_id":"6a4c700225849b193a8340e8","name":"Jonathan Raiman","hidden":false},{"_id":"6a4c700225849b193a8340e9","name":"Mohammad Shoeybi","hidden":false},{"_id":"6a4c700225849b193a8340ea","name":"Bryan Catanzaro","hidden":false},{"_id":"6a4c700225849b193a8340eb","name":"Wei Ping","hidden":false}],"publishedAt":"2026-07-06T00:00:00.000Z","submittedOnDailyAt":"2026-07-07T00:00:00.000Z","title":"Unified Audio Intelligence Without Regressing on Text Intelligence","submittedOnDailyBy":{"_id":"5f1158120c833276f61f1a84","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1608042047613-5f1158120c833276f61f1a84.jpeg","isPro":false,"fullname":"Niels Rogge","user":"nielsr","type":"user","name":"nielsr"},"summary":"Audio intelligence involves understanding, reasoning about, and generating both audio and speech. In this work, we introduce Nemotron-Labs-Audex-30B-A3B (Audex), a unified audio-text LLM built on Nemotron-Cascade-2-30B-A3B, a strong text-only MoE LLM. Audex adopts a simple unified design with a single Transformer decoder: audio inputs are encoded and projected into the text embedding space, while text tokens and quantized audio output tokens are treated uniformly during generation. This architecture enables strong audio-text fusion, seamless multimodal generation, and compatibility with standard LLM training and inference infrastructure. For training, we meticulously curate audio-text datasets comprising 157.4B audio tokens and 320.5B text tokens. We apply multi-stage supervised training on these datasets, followed by text-only Cascade RL and multi-domain on-policy distillation. Audex delivers state-of-the-art audio understanding, speech recognition and translation, text-to-speech, audio generation, and speech-to-speech generation, while preserving very compelling reasoning, alignment, knowledge, long-context, and agentic capabilities of its text-only LLM backbone with marginal or no regression. We release the model checkpoints to facilitate open research.","upvotes":7,"discussionId":"6a4c700225849b193a8340ec","ai_summary":"A unified audio-text large language model is presented that integrates audio and text processing through a shared transformer decoder, achieving superior performance across multiple audio and speech tasks while maintaining strong text reasoning capabilities.","ai_keywords":["audio-text LLM","Transformer decoder","multimodal generation","supervised training","Cascade RL","on-policy distillation","text-to-speech","speech-to-speech generation","audio understanding","speech recognition","audio generation"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"60262b67268c201cdc8b7d43","name":"nvidia","fullname":"NVIDIA","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/65df9200dc3292a8983e5017/Vs5FPVCH-VZBipV3qKTuy.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"663ee43bfeeb49803537da98","avatarUrl":"/avatars/17c3e9c435cc36fb04b4589e6176a243.svg","isPro":false,"fullname":"Wei Ping","user":"wping","type":"user"},{"_id":"628dba7ea13c4b25ff786ec6","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1671164940704-628dba7ea13c4b25ff786ec6.jpeg","isPro":false,"fullname":"Sang-gil Lee","user":"L0SG","type":"user"},{"_id":"652a4dfc36f031c5e6f8b8a6","avatarUrl":"/avatars/9fb56b025dc25f91ca6c31136eaf74b2.svg","isPro":false,"fullname":"Zhifeng Kong","user":"ZhifengKong","type":"user"},{"_id":"69d08207e192e1734ff2a335","avatarUrl":"/avatars/6417436c8fcca47d67402a96511b5603.svg","isPro":false,"fullname":"kyosuke yanagisawa","user":"yyykyosuke","type":"user"},{"_id":"6a2da6c8ca070ee12c6e396c","avatarUrl":"/avatars/0355287dcabaa67dbc7f0b10b87451f9.svg","isPro":false,"fullname":"Joe Mama","user":"JoeMama123123123","type":"user"},{"_id":"6a159e2e31430965c653cd91","avatarUrl":"/avatars/3e2a5e959cae38e6ab0c90af59aadf37.svg","isPro":false,"fullname":"张千怡","user":"ISAACGRE","type":"user"},{"_id":"62bc9d90e81dfd65cced9316","avatarUrl":"/avatars/05df14cd1fdbc7d6a80d2960a05a94f0.svg","isPro":false,"fullname":"Yang Chen","user":"ychenNLP","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"60262b67268c201cdc8b7d43","name":"nvidia","fullname":"NVIDIA","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/65df9200dc3292a8983e5017/Vs5FPVCH-VZBipV3qKTuy.png"},"query":{}}">

Cite arxiv.org/abs/2607.05196 in a dataset README.md to link it from this page.