Exploring Task Performance with Interpretable Models via Sparse Auto-Encoders arxiv.org 2 points by PaulHoule 12 hours ago