01
The Geometry of In-Context Learning
A geometric account of how transformers construct and update implicit task representations from demonstrations.
THE AXIOM INDEX
148,291 papers across artificial intelligence, mathematics, and the ideas between them.
A geometric account of how transformers construct and update implicit task representations from demonstrations.
Empirical and theoretical scaling laws for learning maps between infinite-dimensional function spaces.
A search-and-verification framework that turns informal mathematical intuition into machine-checkable proofs.
Characterizing the frequency preferences of graph and manifold neural networks across non-Euclidean domains.