method
active
method:average-l0-metricAverage L0 metric
Mean number of nonzero features as sparsity metric, tracked as EMA.
Neighborhood — ranked by edge-count
Papers (1)
paper
Artifacts (1)
artifact
- ReLU SAE trained on residual stream layer 19 with MSE+L1 loss.
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.
- Caveat on metric.
- Quantitative measure of morphogenetic success: Euclidean distance between evolving embryo phenotype and target smiling-face pattern.
- Task-dependent optimal sparsity.
- Sparsity trade-off.
- Average number of nonzero feature entries per input; primary measure of activation sparsity in the autoencoder
- Task signal in representation normalized to random baseline.
- Measure of features activated by clean vs adversarial inputs.
- Sparsity penalty weighting by sum of per-layer decoder norms, enabling comparability with per-layer SAEs.