4 papers
Sky sphere representation in language models
Aleksandr Berdnikov, Yevgeny Liokumovich
We analyze whether language models of size ~100B have a representation of the night sky map that is decodable from their residual stream. We find that most of the considered open-s…
Quantifying stability of non-power-seeking in artificial agents
Evan Ryan Gunter, Yevgeny Liokumovich, Victoria Krakovna
We investigate the question: if an AI agent is known to be safe in one setting, is it also safe in a new setting similar to the first? This is a core question of AI alignment--we t…
Area of convex disks
Gregory R. Chambers, Christopher Croke, Yevgeny Liokumovich +1
This paper considers metric balls in two dimensional Riemannian manifolds when is less than half the convexity radius. We prove that .…
Lengths of three simple periodic geodesics on a Riemannian -sphere
Yevgeny Liokumovich, Alexander Nabutovsky, Regina Rotman
Let be a Riemannian -sphere. A classical theorem of Lyusternik and Shnirelman asserts the existence of three distinct simple non-trivial periodic geodesics on . In this p…