method
active
method:normalized-signal-metricNormalized Signal metric
Task signal in representation normalized to random baseline.
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.
- Core abstraction in Fruit model: a function from continuous time to a value; foundation for reactive programming in Fruit.
- Reconstruction quality metric.
- Average of S(ℓ) across layers.
- Training-free technique normalizing all task gradients to the maximum gradient norm magnitude
- Normalized cross-entropy metric used as language model performance measure on OpenWebText
- EI and normalized EI could serve as a unified metric for out-of-distribution generalization.claim0.725Conjecture that maximizing EI yields causal representations invariant to distribution shifts.
- Divides score by 1 − q/((n−2)p), clipped at 1.
- Patterns introduced by conditioning paradigms, learned through perceptual/value learning.