Melanie Weber

Priors in time: Missing inductive biases for language model interpretability

Achieving Approximate Symmetry Is Exponentially Easier than Exact Symmetry

Latent Structure of Affective Representations in Large Language Models

Neural Feature Geometry Evolves as Discrete Ricci Flow

Unitary Convolutions for Message-passing and Positional Encodings on Directed Graphs

Shared Global and Local Geometry of Language Model Embeddings

Lie Algebra Canonicalization: Equivariant Neural Operators under arbitrary Lie Groups

Higher-Order Learning with Graph Neural Networks via Hypergraph Encodings

Performance Heterogeneity in Message-Passing and Transformer-based Graph Neural Networks