Decomposing MLP Activations into Interpretable Features via Semi-Nonnegative Matrix Factorization

Decomposing MLP Activations into Interpretable Features via Semi-Nonnegative Matrix Factorization

12 June 2025

ArXiv (abs)PDF HTML

Papers citing "Decomposing MLP Activations into Interpretable Features via Semi-Nonnegative Matrix Factorization"

Title
No papers