claim
active
claim:causal-importance-is-related-but-not-identical-to-causal-dependence-because-of-self-repairCausal importance is related but not identical to causal dependence because of self-repair
Related work distinction.
Source paper
extracted_from(2025) · Lucius Bushnaq · Dan Braun · Lee Sharkey
Neighborhood — ranked by edge-count
Papers (1)
paper
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- A measure of whether a subcomponent is necessary to reproduce model behavior on a specific prompt, predicted by the causal importance network.
- Small per-subcomponent MLP with hard sigmoid predicting causal importance.
- Auxiliary model trained alongside VPD to predict which subcomponents are causally important for each prompt, enabling mechanistic isolation of components.
- Attribution methods are necessary for causal explanations but are often difficult to interpretclaim0.754Open problem in attribution.
- Promised for Book 4, chapter 4 (Note 15).
- Self-Normalizing Importance Sampling performed worse than Importance Sampling in this settingfinding0.740Appendix baseline comparison.
- Probability that effect would not occur without cause; counterfactual primitive.