Author
Vijay Erramilli
0 works0 citations
Recent research
- Physics & SpaceOpen access
Mechanistic Tomography: Designed Measurement for Control-Oriented Interpretability
Mechanistic interpretability seeks quantities that a model does not expose directly: represented states, component effects, interactions, and responses to interventions. Patching, gradients, Hessian-vector products, and subset interventions obtain different measurements under dif...