method
active
method:mean-max-cosine-similarity-mmcsMean Max Cosine Similarity (MMCS)
Measures alignment of learned components with target columns.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Stochastic Parameter Decompositioncitesmentions
Findings (1)
finding
- Ground-truth columns recovered with negligible shrinkage.
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 measure alignment between DIM direction and cone basis vectors to assess overlap
- Cosine similarity between feature activations restricted to tokens where one of the features fires; used to identify feature splitting relationships
- Measures QK geometry between country queries and city keys.
- High cosine similarity for Gemma3 steering vectors suggests strong linear reflection structure.
- Classifier using cosine similarity between activation vectors and steering vectors to detect deception with 89% accuracy
- Used to quantify the semantic clustering of adjective-set embeddings across model families and conditions
- Detects tiling in real SAEs.
- Geometric evaluation of truth direction alignment across layers and prompt templates.