Causal Mediation

Whether an internal direction causally controls a target behavior, verified by intervention success

Neighborhood — ranked by edge-count

method

Answer Switching Rate (ASR)
implements
Key evaluation metric: proportion of inputs for which an intervention successfully flips model output

concept

Causal Mediation Analysis
related_to
Framework for evaluating whether probe directions are causally implicated in model outputs via activation patching
Causal Mediation of Model Outputs
related_to
The extent to which a probe direction, when intervened upon, actually changes model outputs — contrasted with mere classification accuracy

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Causal Mechanismconcept0.842
Function determining the value of a variable based on its causal parents in an acyclic causal model.
causal maskingconcept0.829
Attention restricted to previous tokens only, as in decoder-only models; leads to AR(ω)-like behaviour and no ordered phase
Causal Invarianceconcept0.809
Property that causal mechanisms remain stable across environments; desirable for OOD.
Causal abstractionconcept0.809
A framework the paper uses alongside feature geometry to deepen mechanistic understanding of LMs
Deterministic Causal Modelconcept0.805
Formal representation of algorithms as directed acyclic graphs computing functions f_A
Causal Emergenceconcept0.805
Core concept: degree to which an agent exerts unique predictive power on its future; key to cognition at all scales.
Causal Geometryframework0.800
Chvykov and Hoel's geometric extension of causal emergence to continuous systems using Fisher information.
Causal Tracingconcept0.800
Mechanistic interpretability technique for locating factual associations, mentioned as future work direction.