Sign in

Klemen Kotar

@klemenkotar.bsky.social
28 followers 51 following 0 posts

CS PhD Candidate at Stanford NeuroAI Lab

PostsRepliesMedia
Reposted by Klemen Kotar
Greta Tuckute @gretatuckute.bsky.social · 19/08/2025
Humans largely learn language through speech. In contrast, most LLMs learn from pre-tokenized text. In our #Interspeech2025 paper, we introduce AuriStream: a simple, causal model that learns phoneme, word & semantic information from speech. Poster P6, tomorrow (Aug 19) at 1:30 pm, Foyer 2.2!
15110
Reposted by Klemen Kotar
Greta Tuckute @gretatuckute.bsky.social · 19/08/2025
Joint with @klemenkotar.bsky.social, and with @evfedorenko.bsky.social @dyamins.bsky.social Paper: www.isca-archive.org/interspeech_... Website: tukoresearch.github.io/auristream-s... (with audio examples) HuggingFace: huggingface.co/TuKoResearch...
isca-archive.org
ISCA Archive - Representing Speech Through Autoregressive Prediction of Cochlear Tokens
121
Reposted by Klemen Kotar
Greta Tuckute @gretatuckute.bsky.social · 23/05/2025
What are the organizing dimensions of language processing? We show that voxel responses during comprehension are organized along 2 main axes: processing difficulty & meaning abstractness—revealing an interpretable, topographic representational basis for language processing shared across individuals
37129
Reposted by Klemen Kotar
Greta Tuckute @gretatuckute.bsky.social · 28/04/2025
Sadly couldn’t make it to ICLR Re-Align, but check out @klemenkotar.bsky.social and my prelim work on ‘model connectomes’—sparse initializations derived across LLM generations to enable efficient learning in low-data regimes, loosely inspired by evolution and lifetime learning. shorturl.at/PNXXW
tinyurl.com
Model Connectomes: A Generational Approach to Data-Efficient...
Biological neural networks are shaped both by evolution across generations and by individual learning within an organism’s lifetime, whereas standard artificial neural networks undergo a single...
071