char counts : each represented on 1-dimensional feature manifolds embedded with high curvature in low-dimensional subspaces of the residual stream

representing char count

ringing in manifolds == interference in superposition linear probees work well

1. SAE feature showing like tuning curve

hand picked SAE features showing tuning features

2.

detecting boundary

QK matrix does it such that inner product is large for certain region

final decision - word or \n predicting new line

orthogonal subspaces?