AbsoluteZero: Reinforced Self-play Reasoning with Zero Data

A seminar for discussing the recent AbsoluteZero paper.

Abstract

In this talk I will present the motivation for learning without data. Then, the AbsoluteZero method will be discussed with detailed equations and insights. Finally, I will summarize the results, highlight the key findings, discuss the advantages, and outline opportunities for future improvements.

Please join the meeting using the link above.