method
active
method:intrinsic-dictionary-health-auditIntrinsic Dictionary Health Audit
A hyperparameter selection procedure driven by intrinsic measures of SAE dictionary quality that transfers across architectures
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Dictionary Health Auditrelated_toIntrinsic hyperparameter selection procedure based on dictionary quality metrics; introduced in this paper to transfer across architectures.
Claims (1)
claim
- Key methodological contribution claim about architecture-agnostic SAE tuning
Related by similarity (6)
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.
- Demonstrates architecture-agnostic applicability of the SAE tuning method
- Value derived from information gain; comprises salience and novelty.
- Reflection level where a model spontaneously revises reasoning without explicit trigger instructions.
- The drive to explore arising from epistemic value, independent of extrinsic reward, naturally emerging in active inference.
- Ability to maintain function despite perturbations.