method
active
method:fisher-effectivenessFisher effectiveness
Second-order KL estimate of a weight's effect on outputs using Fisher metric.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Characterizing interference weights in a tiny language modelintroducesmentions
Concepts (1)
concept
- Interference WeightsaboutLogit weight contributions from a feature that arise due to superposition with other features, not from the feature's own causal role
Methods (3)
method
- Prior coactivation-based proxy for effectiveness.
- Computes feature attribution passing ablated pre-activation through JumpReLU to handle inhibition.
- Second-order saliency pruning; ancestor of Fisher effectiveness.
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.
- Curvature proxy approximated by K-FAC.
- Normalized EI bounded 0-1, decomposed into determinism minus degeneracy.
- Core measure of causal effect in Hoel's theory; mutual information between uniform input and output distributions.
- Diversity measured over a set of m responses for a single conversation, as opposed to test set diversity
- The mechanism by which each step's effect is evaluated against the life of the whole, guiding the unfolding.
- Phenomenological fit P(success)=σ(αS+β) used to summarize sharpness and midpoints.
- The role played by the price system: enabling members to form mutually compatible plans.
- The speed and reliability with which a collective reaches the target morphology.