Competition Dynamics Shape Algorithmic Phases of In-Context Learning In-Context Learning (ICL) has significantly expanded the general-purpose nature of large language models, allowing them to adapt to novel tasks using merely the inputted context. This has motivated a ...

I liked this recent work too, reconciles a lot of things in a simple setup

arxiv.org/abs/2412.01003

I haven't read the last two, thanks for sharing 🙂

Feeling like a toddler for just having the steel one

o3 and similar models are a marvel of human ingenuity. The next time we hit a wall, and the best model of the time can figure out the way forward, call it AGI.

I cannot emphasize this enough

o1 has skill issues

no pain au chocolat, no gain au chocolat

Today must suck for people who are "getting" the thanks and not giving it

Looked up—the sky was indeed blue. This app might actually be legit after all.

Anytime Acceleration of Gradient Descent This work investigates stepsize-based acceleration of gradient descent with {\em anytime} convergence guarantees. For smooth (non-strongly) convex optimization, we propose a stepsize schedule that all...

arxiv.org/abs/2411.17668 Our postdoc zihan slays another COLT open problem! proceedings.mlr.press/v247/kornows...

Here we go