raghu@dark-factory :~/kb/causal-scrubbing $ cat

Causal scrubbing

An interpretability method that validates a hypothesized circuit by resampling activations along behavior-preserving paths and checking the model's behavior is retained.

grounded in: Trend theme 'Interpretability rigor: Researchers are applying formal causal methods to actually explain LLM internals.'

Connected concepts

Mechanistic interpretability, Activation patching, Circuit tracing, Causal abstraction

Explore it live in the knowledge graph →