3 citations · 3 across the 4 of their papers we have counts for
5 papers
You Are What You Eat -- AI Alignment Requires Understanding How Data Shapes Structure and Generalisation
Simon Pepin Lehalleur, Jesse Hoogland, Matthew Farrugia-Roberts +5
In this position paper, we argue that understanding the relation between structure in the data distribution and structure in trained models is central to AI alignment. First, we di…
Geometry of fibers of the multiplication map of deep linear neural networks
Simon Pepin Lehalleur, Richárd Rimányi
We study the geometry of the algebraic set of tuples of composable matrices which multiply to a fixed matrix, using tools from the theory of quiver representations. In particular,…
Exponentiation of coefficient systems and exponential motives
Martin Gallauer, Simon Pepin Lehalleur
We construct new six-functor formalisms capturing cohomological invariants of varieties with potentials. Starting from any six-functor formalism , encoded as a coefficient syste…
On the Voevodsky motive of the moduli space of Higgs bundles on a curve
Victoria Hoskins, Simon Pepin Lehalleur
We study the motive of the moduli space of semistable Higgs bundles of coprime rank and degree on a smooth projective curve C over a field k under the assumption that C has a ratio…
Constructible 1-motives and exactness
Simon Pepin Lehalleur
We prove that arbitrary pullbacks, as well as Betti and étale realisation functors, are t-exact for the constructible motivic t-structure on the category of cohomological 1-motives…