method
active
method:activation-dendrogram-analysis-pca-cosine-hierarchical-clusteringActivation dendrogram analysis (PCA + cosine + hierarchical clustering)
Preliminary interpretability of interventions.
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Policy rollout dendrograms cluster intervention, entity, and action in separate sub-branches; absent in base rolloutsassociated_withPreliminary mechanism evidence.
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.
- Used to test event alignment.
- Reveals low-dimensional counting manifolds.
- Method used to identify the periodic features and their periods in Llama-3.1-8B's MLP neurons
- Statistical method used to analyze neural activity data.
- Used to test linear separability of group position.
- Used to analyze emotion vector geometry.
- PCA on 171 emotion probe activations across all tokens to produce ordered linear combinations and test if lower PCs are more persistent
- Confirms three behaviors.