Lately we have been studying how Speech-to-Text Translation (S2TT) systems scale with more data. 🚀

We find that Direct S2TT scales more reliably than Chain of Thought (CoT) prompting when trained on pseudo-labeled data.

[1/8]

During the last months we have been curious whether Chain-of-Thought (CoT) Speech-to-Text Translation (S2TT) systems truly use the spoken input during translation.

We find that they rely more on the transcript and behave closer to cascade systems than expected.

[1/9]

Roger Moore reminds Interspeech audience that speech is not audible text. Text is a technology.

A quick (and slightly late) career update: I joined the Barcelona Supercomputing Center (BSC) in January 2025! I'm now in a full-time role, back to Speech Translation after a few years of internships and detours. 🧵

As we welcome 2025, we're excited to share that our paper, "Single-stage TTS with Masked Audio Token Modeling and Semantic Knowledge Distillation", has been accepted to #ICASSP2025!
This work advances single-stage Non-Autoregressive TTS based on audio token modeling.
🧵

We outperform Llama 70B with Llama 3B on hard math by scaling test-time compute 🔥

How? By combining step-wise reward models with tree search algorithms :)

We're open sourcing the full recipe and sharing a detailed blog post 👇

congratulations, @ian-goodfellow.bsky.social, for the test-of-time award at @neuripsconf.bsky.social!

this award reminds me of how GAN started with this one email ian sent to the Mila (then Lisa) lab mailing list in May 2014. super insightful and amazing execution!

Arxiv sharing reminder

pdf ❌
abs ✅