activation-patching

Causal intervention via activation patching to identify important model components. Use when determining which layers, heads, or positions are causally responsible for model behavior.

by ndif-team· Repository·other
Also installable via skills CLI
npx skills add ndif-team/skills/plugins/nnsight/skills/activation-patching

Source

Path:plugins/nnsight/skills/activation-patching(main)

Related in other

activation-patching | AgentArea Skills