Mitchell Ostrow

neurostrow.bsky.social

21 September 2026 18:22

Very excited to announce our most recent paper! TL;DR we made our prior work on comparing dynamics (DSA) much faster and generalizable to diverse data domains and problem settings (1/)

Greta Tuckute

gretatuckute.bsky.social

17 September 2026 17:30

Thanks so much for the fun conversation @wiair.bsky.social ! A pleasure to chat about language, LLMs, and memory--covering some work with @bkhmsi.bsky.social @mschrimpf.bsky.social @michael-lepori.bsky.social @klemenkotar.bsky.social @evfedorenko.bsky.social @thomashikaru.bsky.social, among others!

Deniz Bayazit

bayazitdeniz.bsky.social

08 September 2026 13:02

11/ Would love to hear your thoughts, especially on what other evidence could count as an internal language (e.g., formal languages!), and hope to see you at EMNLP 2026!

Deniz Bayazit

bayazitdeniz.bsky.social

08 September 2026 13:02

10/ Work done with the amazing Badr AlKhamissi @bkhmsi.bsky.social and Antoine Bosselut @abosselut.bsky.social ! 📄 Paper: arxiv.org/abs/2609.00155 💻 Code: github.com/bayazitdeniz...

Deniz Bayazit

bayazitdeniz.bsky.social

08 September 2026 13:02

9/ We conclude that LLID probes are not interchangeable windows onto one internal lingua franca. They expose complementary aspects of multilingual computation, activation-space geometry vs output-space decodability. The choice of probe is itself part of the claim.