Goodfire(@GoodfireAI):A simple example: days of the week, which lie on a circular path in models’ activations. Steering linearly from Monday to Friday gets you incoherent outputs in between. Steering along the circular manifold means you cleanly shift from Mon → Tues → Wed → Thurs → Fri. (5/8)

Goodfire

@GoodfireAI

Using interpretability to understand, learn from, and design AI.

가입 August 2024

31 팔로잉 중 26.3K 팬

Goodfire@GoodfireAI

2026.05.07 16:09

A simple example: days of the week, which lie on a circular path in models’ activations. Steering linearly from Monday to Friday gets you incoherent outputs in between. Steering along the circular manifold means you cleanly shift from Mon → Tues → Wed → Thurs → Fri. (5/8)