Congratulations to Li-Wei @liweiche77 on successfully defending his PhD today! 🎉 Wishing him all the best in his next chapter!
English
WAVLab | @CarnegieMellon
310 posts
WAVLab | @CarnegieMellon
@WavLab
Shinji Watanabe's Audio and Voice Lab | WAVLab @LTIatCMU @SCSatCMU | Speech Recognition, Speech Enhancement, Spoken Language Understanding, and more.
شامل ہوئے Ağustos 2021
146 فالونگ2.4K فالوورز
Congratulations to Siddhant @Sid_Arora_18 on a successful PhD defense today! It was wonderful to celebrate this big milestone together. Wishing him all the best for the exciting journey ahead.
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
PRiSM: Benchmarking Phone Realization in Speech Models
Shikhar Bharadwaj, Chin-Jou Li, Yoonjae Kim, Kwanghee Choi, Eunjung Yeo, Ryan Soh-Eun Shim, Hanyu Zhou, Brendon Boldt, Karen Rosero Jacome, Kalvin Chang, Darsh Agrawal, …
arxiv.org/abs/2601.14046 [𝚌𝚜.𝙲𝙻 𝚌𝚜.𝚂𝙳]
Indonesia
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Chenda Li, Wei Wang, Marvin Sach, Wangyou Zhang, Kohei Saijo, Samuele Cornell, Yihui Fu, Zhaoheng Ni, Tim Fingscheidt, Shinji Watanabe, Yanmin Qian, "ICASSP 2026 URGENT Speech Enhancement Challenge," arxiv.org/abs/2601.13531
Deutsch
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Pu Wang, Shinji Watanabe, Hugo Van hamme, "SSVD-O: Parameter-Efficient Fine-Tuning with Structured SVD for Speech Recognition," arxiv.org/abs/2601.12600
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Shih-Heng Wang, Jiatong Shi, Jinchuan Tian, Haibin Wu, Shinji Watanabe, "Do Neural Codecs Generalize? A Controlled Study Across Unseen Languages and Non-Speech Tasks," arxiv.org/abs/2601.12205
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Heading to NeurIPS 2025 in San Diego!
I’ll present our spotlight poster, ARECHO, focusing on speech multi-metric estimation.
📍 Exhibit Hall C,D,E #2000
🗓️ Thu Dec 4, 11 a.m.–2 p.m. PST
If you’re around, let’s say hi or grab a coffee!
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
This is exactly the reason we worked for ESPnet-Codec, but being really hard to keep tracking as people are fast nowadays.
The similar issue happens at most speech tasks from ASR, TTS, to general speech LLM. It's a bit sad time for driving scientific findings 🥲
🐿️🐒🗻📚🐹🦈@SythonUK
ﾇﾗｰﾙｵﾁﾞｵｰｺﾃﾞｸの論文、全く違うデータで学習されたモデルを比較して「ワイらのモデル最強や！！😤😤😤」と主張しているものばかりで😩😩😩😩😩😩😩😩😩😩😩に関するMOS値が1000000になった
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Speech isn’t just sound -> it’s how we turn thought into expression.
Our new work, Speech-DRAME, measures how well speech AI can act, aligning evaluation with human perception.
Paper: arxiv.org/abs/2511.01261
Code: github.com/Anuttacon/spee…
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
🚀 I’m open to new opportunities in industry!
Ph.D. candidate @CMU (advisor: @shinjiw_at_cmu ).
Research: speech/audio AI, speech LLMs, evaluation frameworks.
Ex-Meta AI, Tencent, IBM Research.
DMs open — let’s connect!
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Sharing our initial leaderboard and open data release for 🎶Music Arena⚔️!
Music is subjective and multi-dimensional. A key goal of Music Arena is to provide insights beyond binary preferences! 🧵
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
ARECHO has been accepted by #neurips25 as spotlight! Many thanks to all the co-authors for their great effort and support!
jiatongshi@jiatongshi
🔊 New release: #ARECHO -> Autoregressive Evaluation via Chain-based Hypothesis Optimization. • 87-metric coverage in one model 🧮 • Dynamic classifier chain 🤝 • Unified tokenization 🧩 • Confidence-aware decoding 🛡️ Built on #UniVERSA, heading to #VERSA. More ↓
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
espnet v.202509 released 🚀 github.com/espnet/espnet/…
Includes many updates + fixes for NumPy 2.0 & Python 3.12 (thanks Nelson!).
This is the last major update before we shift to the next-gen framework, ESPnet3
Interested in collaborating? Let us know!
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Thanks, @HungyiLee2, for visiting CMU! Great discussions, inspiring research exchanges, and exciting seeds for collaboration ahead.
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
I’m happy to share that our paper, "Discrete Audio Tokens: More Than a Survey!", has been accepted at TMLR. 🎉
📄 Read: arxiv.org/pdf/2506.10274
🔎 Explore our tokenizer database & submit yours: poonehmousavi.github.io/dates-website/…
Gallil Maimon@GallilMaimon
🎉🥳 I am thrilled to share that our work on audio tokenisers has been accepted to #TMLR The tokeniser DB is ever updating so submit your new tokenisers 💪 poonehmousavi.github.io/dates-website/
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
For my CMU course, we built the OWSM v4 demo😀
Check it out here: huggingface.co/spaces/espnet/…
Shinji Watanabe@shinjiw_at_cmu
Our work on OWSM v4 received the Best Student Paper Award at #Interspeech2025! 🏆🎉 Huge congratulations to the team! 🚀👏 I’m especially happy to see our open science efforts for speech foundation models recognized by the community. 🙌 🔗 isca-archive.org/interspeech_20…
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
OWSM-V4 is now available as a demo!
It includes both the OWSM-V4 Medium model and OWSM-V4 CTC model, each with about 1B parameters.
👉 Try it out here: huggingface.co/spaces/espnet/…
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Our work on OWSM v4 received the Best Student Paper Award at #Interspeech2025! 🏆🎉
Huge congratulations to the team! 🚀👏
I’m especially happy to see our open science efforts for speech foundation models recognized by the community. 🙌
🔗 isca-archive.org/interspeech_20…
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Excited to be presenting 3 papers at #Interspeech2025 in Rotterdam this week!
1. Chain-of-Thought Reasoning for E2E Spoken Dialogue Systems
Adds reasoning to real-time dialogue models, with open-source toolkit.
Oral — Thu, 10:10 | Dock 10B arxiv.org/abs/2506.00722
English
WAVLab | @CarnegieMellon ری ٹویٹ کیا
Meet Masao next week in #Interspeech2025
We use sound event, language and speaker context to prune large speech models for ASR and Speech Translation.
Masao@mmiagshatoy
🚀 Happy to share our #INTERSPEECH2025 paper: Using speaker & acoustic context, we dynamically adjust model paths, resulting in a 25.7% relative BLEU improvement in speech translation. We also analyze how context influences model behavior. 📜 Paper: arxiv.org/abs/2505.18860
English