method
active
method:per-dev-z-scaling

per-dev z-scaling

Standardizing ρd and dr using dev-set means and stds to form dimensionless components of S.

Neighborhood — ranked by edge-count

Concepts (1)

concept
  • anchoring strength S
    associated_with
    Composite score S = ρd − dr − log k predicting anchoring success.

Methods (1)

method
  • Standardization of ρd and dr components using development-set mean and standard deviation.

Datasets (1)

dataset

Artifacts (1)

artifact

Related by similarity (8)

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.

  • Preprocessing pipeline for standardizing ρd, dr, and S across layers/models using dev-set covariance
  • Scaling Supervisionframework0.726
    Techniques that leverage AI to help humans more efficiently supervise AI.
  • Mechanisms by which smaller competent subunits bind into a higher-level Self with larger goals; key example via gap junction connections.
  • Used in the color cooccurrence experiment to embed colors into 3D space preserving dissimilarity matrix distances
  • Implementation detail weighting softmax by log(n_memories) to prevent down-weighting of attention values as memory set grows.
  • Interpretive process for transforming many-valued contexts into formal contexts via scale attributes.
  • Power law scalingconcept0.721
    Observation that SAE loss decreases as a power law with compute budget.
  • Multitask Scalingconcept0.715
    The pressure on models trained on more tasks to find representations that generalize across all tasks, reducing the solution space