finding
active
finding:gemma-scope-decoder-nearest-neighbor-cosine-similarities-skewed-above-random-but-few-extremely-highGemma Scope decoder nearest-neighbor cosine similarities skewed above random but few extremely high
Fig 7.
Source paper
extracted_from(2025) · Michaud, Eric J. · Gorton, Liv · McGrath, Tom
Neighborhood — ranked by edge-count
Papers (1)
paper
Hypotheses (1)
hypothesis
- Many latents tiling a low-dimensional manifold would show decoder cosine similarities near 1associated_withsupportsMotivates Appendix E.
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.
- Detects tiling in real SAEs.
- Training artifact.
- Identifying related features by cosine distance in SAE decoder space.
- Experiment 4 result showing DIM captures only one facet of the multi-dimensional truth subspace
- Larger rotations between shared features skew decoder cosine similarities downward in the toy modelfinding0.791Supports 'same features, used differently' interpretation.
- Geometric evaluation of truth direction alignment across layers and prompt templates.
- Detection mechanism computing cosine similarity between activation vectors and steering vectors to classify deception
- High cosine similarity for Gemma3 steering vectors suggests strong linear reflection structure.