Invited Talks

Reinforcement Learning for Theorem Proving
Alex Best
Harmonic, Palo Alto, USA
I'll discuss recent advances training language models to prove theorems at a very advanced level via reinforcement learning on formal languages. I will cover the progression of these systems from proving standalone statements, to building up whole new theories. I'll give a demo of the system we have been developing with these techniques, called Aristotle, and survey interesting use cases so far. Additionally I will discuss the potential ways that a mathematicians work will change when incorporating these tools.

Please consider subscribing to the SCML Mailing List in order to stay informed about the SCML-2026 conference.