dataset
active
dataset:development-set-dev-pool

development set (dev pool)

Set of 1,000 prompts with k=3 examples used for whitening and z-scoring calibration.

Neighborhood — ranked by edge-count

Methods (2)

method
  • per-dev z-scaling
    associated_with
    Standardizing ρd and dr using dev-set means and stds to form dimensionless components of S.
  • Whitening of span embeddings
    associated_with
    Preprocessing step that uses dev-set covariance to standardize embedding scales before computing ρd and dr.