←back to thread

Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet

(transformer-circuits.pub)
168 points 1wheel | 1 comments | 21 May 24 15:15 UTC | HN request time: 0.375s | source
1. feverzsj ◴[21 May 24 15:23 UTC] No.40429632[source]▶
So they made a system by trying out thousands of combinations to find the one gives best result, but they don't understand what's actually going on inside.