A negative result: approximations of the Fisher information don’t yield better clusters of parameter-decomposition components than coactivation does. Playing around with Silico from Goodfire.
Blog
Belief manifolds, and how to steer along them
May 24, 2026A reproduction of Sarfati et al.’s “The Shape of Beliefs”
How LLMs encode in-context beliefs as curved manifolds, and how manifold-aware steering changes them with fewer side effects than linear steering.
BlueDot Technical AI Safety Project
Decomposed concept-injection introspection (Gemma-3 12B, Qwen-2.5 32B) into separable components: representation (what the model encodes about an injection) and report (the prompt-dependent late-layer circuitry that surfaces it), explaining apparent conflicts across prior protocols.