Deniz Bayazit

bayazitdeniz.bsky.social

25 September 2025 17:18

1/🚨 New preprint How do #LLMs’ inner features change as they train? Using #crosscoders + a new causal metric, we map when features appear, strengthen, or fade across checkpoints—opening a new lens on training dynamics beyond loss curves & benchmarks. #interpretability

pfeiffjo.bsky.social

26 March 2025 10:00

I am hiring a Student Researcher for our Modularity team at the Google DeepMind office in Zurich🇨🇭 Please fill out the interest form if you would like to work with us! The role would start mid/end 2025 and would be in-person in Zurich with 80-100% at GDM docs.google.com/forms/d/e/1F...

Antoine Bosselut

abosselut.bsky.social

25 February 2025 13:14

Lots of great news out of the EPFL NLP lab these last few weeks. We'll be at @iclr-conf.bsky.social and @naaclmeeting.bsky.social in April / May to present some of our work in training dynamics, model representations, reasoning, and AI democratization. Come chat with us during the conference!

Antoine Bosselut

abosselut.bsky.social

25 February 2025 10:43

(NAACL) @mismayil.bsky.social work shows that LLMs still aren’t as good as humans in low-level language creativity too! His work demonstrates that LLMs aren’t as morphologically productive and systematic as humans in morphologically-rich languages (Turkish and Finnish).

Defne Circi

defnecirci.bsky.social

20 February 2025 17:35

It was amazing to collaborate with researchers from 6 institutions on a multilingual project as a bilingual! This winter, I finally met @mismayil.bsky.social in NYC and also got to meet Abdullatif and Dr. Ataman at EMNLP in Miami. Incredible how collaboration can build deep connections!