method
active
method:whitening-and-z-scoring-protocolwhitening and z-scoring protocol
Standardization of ρd, dr, and log k on dev set for computing S.
Neighborhood — ranked by edge-count
Methods (2)
method
- whitening and z-scoring procedurerelated_toCalibration protocol: whiten embeddings on dev pool, z-score ρd and dr per layer.
- Dev-Set Whitening and Z-Scoring Protocolrelated_toPreprocessing pipeline for standardizing ρd, dr, and S across layers/models using dev-set covariance
Artifacts (1)
artifact
- Main paper presenting UCCT and semantic anchoring framework.
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.
- Preprocessing step using dev-set covariance to standardize span embeddings before computing S
- Per-(emotion, token) z-score computed as injected emotion activation minus mean of 170 other probes, contrasted against no-steering baseline
- Alternative interpretation of IID mass-mean probing as projection onto θ_mm after Mahalanobis whitening
- Adaptive gradient clipping method used during training to mitigate spikes
- Preprocessing step that uses dev-set covariance to standardize embedding scales before computing ρd and dr.
- Standardization of ρd and dr components using development-set mean and standard deviation.
- Score = (sum of completed quartet values) × (number of quartets), making portfolio composition consequential.
- Explains Alexander finding that Haiku outranks Opus despite Opus being more capable