(transformer-circuits.pub)

168 points 1wheel | 2 comments | 21 May 24 15:15 UTC | HN request time: 0.555s | source

1. gdiamos ◴[22 May 24 08:07 UTC] No.40438502[source]▶

It looks like Anthropic is now leading the charge on safety

replies(1): >>40456745 #

2. maherbeg ◴[23 May 24 16:30 UTC] No.40456745[source]▶

They always were given that is a part of their mission.

Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet