mingjie_chen

work for @VoxNexusAI4ALL

[mingjie_chen @_chenmj]

@VoxNexusAI4All Nice work!

[mingjie_chen @_chenmj]

@iclr_conf The information has been downloaded as an .jsonl file and spreaded all over the places

[Alejandro Piad Morffis @alepiad]

LLMs cannot reason.

Despite their impressive capabilities, all LLMs, including OpenAI o1, are still fundamentally limited by design constraints that make them incapable of true, open-ended reasoning.

Let's break it down. 🧵 (1/5)

[Haiping Lu @haipinglu]

🚀Excited to speak at Open Data Science Conference Europe 2024 in London tomorrow 6th Sep at 10:30 on "Exploring Multimodal AI beyond Vision and Language."

Join me to explore multimodal AI's potential across six disciplines!
📍 Novotel London West
👉 https://t.co/vtM671z1e7

[Pascale Fung @pascalefung]

#ACL2024 was great. But very few papers were asking hard scientific questions. A lot of “LLM engineering”. In particular, university labs should really defocus on this type of engineering as industry labs, though not publishing every minute detail, are doing way more advanced “LLM engineering” work. Academic reviewers reviewing academic work of this type contribute very little to scientific progress. “LLM engineering” is very seductive - but it becomes obsolete very quickly if you are not looking at the core problem. A few percentage increase on a benchmark can be completely washed out by a bigger model with more training data or another round of RLHF. So ask the hard questions - ask the WHY not just the how.

[Hung-yi Lee (李宏毅) @HungyiLee2]

Discover the cutting-edge world of spoken LLMs with this comprehensive survey!

link: https://t.co/vikqrRT04i

[Xi Wang @wangxieric]

This afternoon @ecir2024, we are going to present our paper “Simulated Task Oriented Dialogues for Developing Versatile Conversational Agents” in the QA & Conversational session. We will share with you the insights we obtained in generating synthetic dialogues. #ecir2024

[CHiME Challenge @chimechallenge]

It gives us great pleasure to pre-announce the 8th CHiME Speech Separation and Recognition Challenge (CHiME-8) that will launch in February 2024.

CHiME-8 TASKS includes:
Task 1 - DASR
Task 2 - NOTSOFAR-1
Task 3 - MMCSG

Please check https://t.co/rhQiJtrdWe

[Sasha Rush @srush_nlp]

Introducing COLM (https://t.co/7T42bAAQa4) the Conference on Language Modeling. A new research venue dedicated to the theory, practice, and applications of language models.

Submissions: March 15 (it's pronounced "collum" 🕊️)

[Hung-yi Lee (李宏毅) @HungyiLee2]

Join us for ASRU's satellite event - the Workshop on Speech Foundation Models & Performance Benchmarks (SPARKS), on Dec 16th, 2023, in Taiwan.

📌 Paper Submission: Oct 19th
🔗 Webpage: https://t.co/ctPGLbJprp
Tip: When registering for ASRU, tick the SPARKS option. #ASRU

[arXiv Sound @ArxivSound]

``One model to rule them all ? Towards End-to-End Joint Speaker Diarization and Speech Recognition. (arXiv:2310.01688v1 [https://t.co/3pcQCkeyAA]),'' Samuele Cornell, Jee-weon Jung, Shinji Watanabe, Stefano Squartini, https://t.co/oxNpZwUTx2

[Wen-Chin Huang @unilightwf]

Check out the top system for track 3 (noisy/enhanced/clean speech quality assessment) for VoiceMOS 2023! https://t.co/tiCl5NUkp1

[Yifan Peng @pengyf21]

@shinjiw_at_cmu @WavLab has released an Open Whisper-style Speech Model (OWSM), perhaps the first attempt in academia to reproduce OpenAI's Whisper from scratch using public data and open-source toolkits. OWSM even supports more translation directions and can be more efficient.

[Guillaume Lample @ NeurIPS 2024 @GuillaumeLample]

Mistral 7B is out. It outperforms Llama 2 13B on every benchmark we tried. It is also superior to LLaMA 1 34B in code, math, and reasoning, and is released under the Apache 2.0 licence.
https://t.co/krGs0xwbLH

[Chenzi Xu @ChenziAmy]

If you ever try to install Kaldi on a Mac with Apple silicon, feel free to check out this new section 1.1.1 in my ASR blog: https://t.co/11FnKgcADr

#Kaldiinstallation #macM1

[Wen-Chin Huang @unilightwf]

💡New preprint on non-autoregressive sequence-to-sequence voice conversion (non-AR seq2seq VC) ‼️
We made seq2seq VC training fast and simple, and it can work on a 5 min parallel dataset!

Demo: https://t.co/chNvaMUfwu
Code: https://t.co/0c0WjX2Ivp
Paper: https://t.co/hkuM0dG2Ct

[Wen-Chin Huang @unilightwf]

Check out our new paper on foreign accent conversion (FAC)! Accepted to APSIPA ASC 2023 🇹🇼
Demo: https://t.co/q4kLFpXxEy
Code: https://t.co/0c0WjX2Ivp
Paper: https://t.co/bi5qEcgVwM

We found none of the three most recent FAC methods is superior to the other 🤔

[Prof Heidi Christensen @christensendkuk]

A further opportunity to join my team. This time as a Gr8 >>Senior Research Fellow<< 3-year post with possibilities for extension.

https://t.co/o551sbPnkf

Pls, RT

[@[email protected] @TomKenter]

Interesting #INTERSPEECH2023 paper (I think ;-)) on using MOS scores vs side-by-side preference tests when comparing TTS systems.

How to choose? Is one more robust/sensitive than the other?

Ever wondered about that? Here is the paper:

https://t.co/K9NS3qe8XM

#NLProc #GoogleAI

[Prof Heidi Christensen @christensendkuk]

+++ PostDoc position+++

Come work with me and my team @shefcompsci on the detection of cognitive decline on the @CognoSpeak project: https://t.co/TsokDwHNou (please RT)