Thomas Schaaf

[Andrew Perrault @PerraultAndrew]

Interested in long-context audio LLMs and hallucinations? We released ~1,140 hrs of synthetic doctor-patient conversations with reference SOAP notes. BeTraC Challenge: build the best open end-to-end SOAP-note system. Two tracks: ≤6B and ≤36B params. https://t.co/UzTx4AHw9I

[David Stutz @davidstutz92]

Part of our Med-Gemini work was a full relabelling of MedQA, revealing that at least 7.4% of examples are unfit for evaluation. Today, we open sourced these annotations alongside our evaluation script as a new standard evaluation on MedQA. A thread 🧵:

https://t.co/7TE40sWUBU

[Paola Garcia @leibnyPaola]

Reminder!!🌟@jhuclsp Save the date! JSALT 2024 Opening day presentations (June 24)
Agenda: https://t.co/qsYpU37V1o
YouTube: https://t.co/Mq8NJbFm4W

[Paola Garcia @leibnyPaola]

📢📢 @jhuclsp Starting in 50 min
The 2024 NAACL and JHU Summer School Topic: Automatic Speech Recognition, June 11, 9AM (ET). Presenters: Sanjeev Khudanpur and Matthew Wiesner
YouTube: https://t.co/Mq8NJbFTUu
Agenda: https://t.co/pdhwpHZsOb

[Paola Garcia @leibnyPaola]

YouTube link: https://t.co/Zj0ezyfdgu

[Paola Garcia @leibnyPaola]

⭐⭐⭐ @jhuclsp JSALT Summer School is on! Come and join us! https://t.co/yOiaSMwTWE

[INTERSPEECH 2025 @ISCAInterspeech]

Last week in Paris, area chairs and Technical Program Chairs finalized an exceptional scientific program for #INTERSPEECH2024. This program reflects major breakthroughs, pushing forward the frontiers of speech technology. Excited for the unveiling! https://t.co/EVu6xBRGvo

[Thomas Schaaf @thomasschaaf20]

@rdesh26 @JohnsHopkins You look good in that gown!

[Leo Boytsov @srchvrs]

In machine learning there are three key factors that affect your model performance: data, data, data.

[Thomas Schaaf @thomasschaaf20]

@sleepinyourhat Not sure if a phrase of „recognize themselves“ is too anthropomorphizing LLM. It sounds reasonable that an model gives a preference to output it generated because this is a more likely sequence for the model.

On the other hand, this sounds also true for a human generated text.

[Sam Bowman @sleepinyourhat]

🤖🥇🤖

[Thomas Schaaf @thomasschaaf20]

@mdredze Seems to be a reasonable rule of thumb from your friend. Sudo is often not restricted enough for some roles people have at after that age. Very funny picture.

[Yifan Peng @pengyf21]

We are hosting a special session at INTERSPEECH 2024 (@ISCAInterspeech) titled "Spoken Language Models for Universal Speech Processing". It would be a great opportunity for researchers interested in speech + LLM to communicate with each other.

Website: https://t.co/CUCzG3B4oa

[Nick Cummins @nickcummins41]

I'm excited to announce our special session on "Speech and Language in Health: from Remote Monitoring to Medical Conversations" is going to be held for a 3rd time @ISCAInterspeech this year. Details at: https://t.co/eUgsNFCApZ.

[Desh Raj @rdesh26]

📢📢 **Defending my PhD in a week**

Date & Time: January 26, 2024, 9 to 11 AM EST
Committee: Sanjeev Khudanpur, Dan Povey, Jinyu Li
Dissertation Title: "Listening to multi-talker conversations: Modular and end-to-end perspectives"

DM me for a Zoom link if interested 😀

[Thomas Schaaf @thomasschaaf20]

@rdesh26 That is great news Desh. All the best!

[Thomas Schaaf @thomasschaaf20]

@zacharylipton Well, generative AI started getting hip a long time ago with Weizenbaums Eliza ;-)

AI is quite a fuzzy term.

[David Mortensen @dmort27]

The happiest years of my life have been spent at @LTIatCMU and I'd like to share. We are hiring. In particular, we would like to hire a mature (junior or senior-rank) Speech Processing researcher. However, we are open to hiring an exceptional candidate in another area. 1/2

[Desh Raj @rdesh26]

Presenting at the NSF CIRC meeting today

[Desh Raj @rdesh26]

My friend Vipul invited me on his podcast to to talk about speech, PhD, and more!