Oliver Clive-Griffin comments on How to use and interpret activation patching

[Oliver Clive-Griffin]

This implies the components make up a cross-section of the circuit

What exactly is mean by “cross-section” here? My intuition would be that if patching a set of components is sufficient to restore behaviour then that means it’s a non-strict superset of components that make up the circuit?

[StefanHex]

Imagine a circuit

A --> B --> C
D --> E --> F
C + F --> output

Denoising (restoring) B + E is sufficient to restore behaviour, but B + E are just a “cross-section” (cutting through the circuit at the 2nd layer), not a superset of all components

[Neel Nanda]

If you imagine the circuit as a computational subgraph, a crosssection is a set of nodes that, if removed, separate the graph into two pieces. I think this is equivalent to your definition

[Oliver Clive-Griffin]

Ah that’s not actually what I meant but that’s clarifies it fully, thanks