Clare Lyle
Publishes 1 feed
Facilis descensus ad minimum
A place to discuss the gory details of the current state of machine learning research: what works, what doesn’t, and why we don’t have AGI yet.
10 posts · theirs
Lately
Sigmoids all the way down
Moravec's paradox applies to genetics, too
The state of plasticity in 2025
A visual guide to proximal methods
Does parameter norm growth cause networks to lose plasticity?
Paper review: Auditing language models for hidden objectives
Can you train a neural network forever?
Neural networks live life on the edge
Do we know why deep learning generalizes yet?
When do adaptive optimizers fail to generalize?
Everything on this page was read from markup Clare Lyle published — a rel="me" link, an h-card, or the feed’s own author element. Nothing was inferred from anywhere else. To correct or remove it, get in touch. Machine-readable: JSON