Cem Subakan (@CemSubakan) profile picture

Cem Subakan

@CemSubakan

Assistant Professor of Computer Science, Laval University / Associate Academic Member at Mila. Machine Learning Researcher, Speech and Audio, Explainable ML

Joined October 2020
  • Tweets115
  • Following12
  • Followers276
10Media
Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)Photo by Cem Subakan (@CemSubakan)
We are happy to share our paper Audio-R1!
Jihoon Jeong (@jhjeong98) profile pictureJihoon Jeong@jhjeong98Oct 28
Audio-R1 🎧 SSL is still in the game. Now with RL rewards. We turn self-supervised audio tasks into verifiable rewards for RL, improving audio-language models on downstream tasks beyond those used in training. 🌐 ijihoon98.github.io/Audio-R1…
5
127
Cem Subakan retweeted
Jihoon Jeong (@jhjeong98) profile picture
Jihoon Jeong@jhjeong98Oct 28
Audio-R1 🎧 SSL is still in the game. Now with RL rewards. We turn self-supervised audio tasks into verifiable rewards for RL, improving audio-language models on downstream tasks beyond those used in training. 🌐 ijihoon98.github.io/Audio-R1…
1
11
2
104
5,010
Cem Subakan retweeted
francesco paissan (@fpaissan_) profile picture
francesco paissan@fpaissan_Oct 8
📣 Our ICASSP 2027 workshop welcomes work on explainability, robustness, and controllability in speech and audio broadly. No need to focus on audio language models (ALMs) only. Submissions are open! 🗓️ Deadline: Nov 11, 2026 (AoE) cmsworkshops.com/ICASSP2027/…
4
5
298
Cem Subakan retweeted
francesco paissan (@fpaissan_) profile picture
francesco paissan@fpaissan_Sep 1
📢 I'm happy to share that our workshop on Explainability, Robustness, and Controllability of Language Modeling for Speech and Audio was accepted at @ieeeICASSP ! If you have ongoing relevant to these topics, consider submitting your work.
Image from francesco paissan (@fpaissan_)
1
5
3
13
1,509
If work with audio language models, you can consider submitting a paper to our ICASSP 2027 workshop!
francesco paissan (@fpaissan_) profile picturefrancesco paissan@fpaissan_Sep 1
📢 I'm happy to share that our workshop on Explainability, Robustness, and Controllability of Language Modeling for Speech and Audio was accepted at @ieeeICASSP ! If you have ongoing relevant to these topics, consider submitting your work.
Image from francesco paissan (@fpaissan_)
2
120
Cem Subakan retweeted
Shubham Gupta (@Gupta30Shubham) profile picture
Shubham Gupta@Gupta30ShubhamJul 3
Excited to share ReTreever, our #ICML2026 paper on hierarchical retrieval at scale! ReTreever learns an inspectable tree over frozen encoder embeddings for fast, scalable, coarse-to-fine retrieval. w/ Zichao Li, Tianyi Chen, @CemSubakan, @sivareddyg, @PerouzT, @vzantedesc 🧵👇
🤖 Made with AI
Image from Shubham Gupta (@Gupta30Shubham)
8
4
5
558
Totally deserved!
francesco paissan (@fpaissan_) profile picturefrancesco paissan@fpaissan_May 26
I’m grateful to have been awarded a Doctoral Research Scholarship from the FRQ @ to support my PhD studies at @universitelaval and @Mila_Quebec. Many thanks to my advisors, @CemSubakan and @mirco_ravanelli and to all my collaborators!
82
Cem Subakan retweeted
francesco paissan (@fpaissan_) profile picture
francesco paissan@fpaissan_May 26
I’m grateful to have been awarded a Doctoral Research Scholarship from the FRQ @ to support my PhD studies at @universitelaval and @Mila_Quebec. Many thanks to my advisors, @CemSubakan and @mirco_ravanelli and to all my collaborators!
1
1
12
333
Conversational AI reading group is back!
Convai_rg (@convAI2024) profile pictureConvai_rg@convAI2024Jan 27
🎉 ConvAI Reading Group is back for 2026! This week we welcome Daniel P. W. Ellis (Google DeepMind) presenting his ICML paper: 🎧 Recomposer: Event-roll-guided Audio Editing 📅 Thu, Jan 29 | 11 AM EST 🔗 poonehmousavi.github.io/rg.h… ▶️ youtube.com/@CONVAI_RG
100
Cem Subakan retweeted
Convai_rg (@convAI2024) profile picture
Convai_rg@convAI2024Jan 27
🎉 ConvAI Reading Group is back for 2026! This week we welcome Daniel P. W. Ellis (Google DeepMind) presenting his ICML paper: 🎧 Recomposer: Event-roll-guided Audio Editing 📅 Thu, Jan 29 | 11 AM EST 🔗 poonehmousavi.github.io/rg.h… ▶️ youtube.com/@CONVAI_RG
3
1
6
727
If you are interspeech stop by our poster!
Pooneh Mousavi (@MousaviPooneh) profile picturePooneh Mousavi@MousaviPooneh19 Aug 2025
📢 Presenting our paper “LiSTEN: Learning Soft Token Embeddings for Neural Audio LLMs” — an interpretable fine-tuning method for spoken language understanding. 🗓 Wed, Aug 20 | 08:30–10:30 📍 A11-P2B-03 Hope to see you there! 📄 arxiv.org/pdf/2505.18517 @ISCAInterspeech
267
Cem Subakan retweeted
MT Group at FBK (@fbk_mt) profile picture
MT Group at FBK@fbk_mt11 Jun 2025
Our pick of the week by @beomseok_lee_: "ALAS: Measuring Latent Speech-Text Alignment For Spoken Language Understanding In Multimodal LLMs" by Pooneh Mousavi, @yingzhi_wang, @mirco_ravanelli, and @CemSubakan (2025) arxiv.org/abs/2505.19937 #SLU #speech #multimodal #LLM
ALAS: An Automatic Latent Alignment Score for Audio Language Models

ALAS: An Automatic Latent Alignment Score for Audio Language Models

Large Language Models (LLMs) are extended into Speech-LLMs, and the quality of the audio--text alignment they learn affects most downstream Spoken Language Understanding (SLU) behavior. Yet...

arxiv.org
Beomseok LEE (@beomseok_lee_) profile pictureBeomseok LEE@beomseok_lee_11 Jun 2025
Speech-language models show promise in multimodal tasks—but how well are speech & text actually aligned? 🤔 This paper arxiv.org/abs/2505.19937 proposes a new metric to measure layer-wise correlation between the two, with a focus on SLU tasks. 🔍🗣️📄
ALAS: An Automatic Latent Alignment Score for Audio Language Models

ALAS: An Automatic Latent Alignment Score for Audio Language Models

Large Language Models (LLMs) are extended into Speech-LLMs, and the quality of the audio--text alignment they learn affects most downstream Spoken Language Understanding (SLU) behavior. Yet...

arxiv.org
7
11
1,020
Last chance to submit a paper to MLSP 2025!
Mirco Ravanelli (@mirco_ravanelli) profile pictureMirco Ravanelli@mirco_ravanelli30 Apr 2025
IEEE MLSP 2025: Deadline extended! The submission deadline is now May 6. - New papers welcome - Revisions accepted 2025.ieeemlsp.org/en/ Don't miss this chance to contribute! #MLSP2025 #MachineLearning #DeepLearning #SignalProcessing #IEEE #CallForPapers #Research
Image from Mirco Ravanelli (@mirco_ravanelli)
2
212
IEEE MLSP 2025: Deadline extended! The submission deadline is now May 6. - New papers welcome - Revisions accepted 2025.ieeemlsp.org/en/ Don't miss this chance to contribute! #MLSP2025 #MachineLearning #DeepLearning #SignalProcessing #IEEE #CallForPapers #Research
Image from Mirco Ravanelli (@mirco_ravanelli)
2
1
8
892
#IEEE MLSP 2025 – Call for Papers The OpenReview portal is now open! 📅 Deadline: April 22, 2025 🔗 Website: 2025.ieeemlsp.org/ We welcome #papers on #deeplearning, #speech/#audio, #LLMs, #CV, #NLP, #EEG, #optimization, #fintech, #healthcare, #wireless comms, & more!
3
10
1,266
This work is a great step towards explainable document retrieval!
Perouz Taslakian (@PerouzT) profile picturePerouz Taslakian@PerouzT13 Feb 2025
🚀 We have released our paper on ReTreever! 🌳🔍 ReTreever organizes and represents documents in a binary tree across various granular levels, balancing cost & utility while enhancing retrieval transparency. 📜 Read it here: arxiv.org/pdf/2502.07971 #AI @ServiceNowRSRCH 🧵👇
Image from Perouz Taslakian (@PerouzT)
1
145
🚀 We have released our paper on ReTreever! 🌳🔍 ReTreever organizes and represents documents in a binary tree across various granular levels, balancing cost & utility while enhancing retrieval transparency. 📜 Read it here: arxiv.org/pdf/2502.07971 #AI @ServiceNowRSRCH 🧵👇
Image from Perouz Taslakian (@PerouzT)
1
21
2
31
4,329
The Special Session proposal deadline is February 1, please consider organizing one! 2025.ieeemlsp.org/en/CALL-FO… The data competitions proposal deadline is February 15: 2025.ieeemlsp.org/en/CALL-FO…
Mirco Ravanelli (@mirco_ravanelli) profile pictureMirco Ravanelli@mirco_ravanelli9 Jan 2025
We are working hard to organize the 35th #IEEE #MachineLearning for #SignalProcessing Workshop (MLSP) in #Istanbul (Aug 31–Sept 3, 2025)! It will be an unforgettable event with inspiring #research, talks, and keynotes. Paper submissions: Apr 22, 2025 2025.ieeemlsp.org/en/
https://2025.ieeemlsp.org/en/

ALT https://2025.ieeemlsp.org/en/

1
91
Cem Subakan retweeted
Convai_rg (@convAI2024) profile picture
Convai_rg@convAI202417 Dec 2024
📢Join our last meeting of 2024 Conversational AI Reading Group on Thursday 19/12/2024, 11 AM-12 PM EST! @mirco_ravanelli presents "Discrete Audio Tokens for Multimodal LLMs" Details: poonehmousavi.github.io/rg You could also follow us on BlueSky : bsky.app/profile/convai-rg.b…
4
1
10
976
I will attend #NeurIPS2024 in Vancouver this week. If you'd like to connect, feel free to reach out privately! We’ll present our paper, "Listenable Maps for Zero-Shot Audio Classifiers" (Main Track, Fri 13 Dec 11 a.m. PST - 2 p.m.) arxiv.org/pdf/2405.17615" #AI #DeepLearning
https://arxiv.org/pdf/2405.17615

ALT https://arxiv.org/pdf/2405.17615

4
7
31
1,949
Load more