2 ms·Exploring Task Performance with Interpretable Models via Sparse Auto-Encoders3 points by PaulHoule 1y ago