method
active
method:svd-orthogonalization-of-emotion-probesSVD Orthogonalization of Emotion Probes
Orthogonalizes the 171 emotion probes via SVD to create an orthonormal basis for computing SAE feature subspace overlap
Neighborhood — ranked by edge-count
Methods (1)
method
- Fraction of an SAE feature's length lying inside the 171-dimensional subspace spanned by emotion probes, computed via SVD orthogonalization
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.
- Metric measuring how much of an SAE feature vector lies within the 171-dimensional subspace spanned by emotion probes, via SVD orthogonalization
- Linear probes constructed to measure 171 emotion concepts in model activations with surface semantic content removed
- Claim supporting the validity of the probe construction method via cross-validation with self-report
- Core empirical claim distinguishing emotion persistence from generic high-variance probe persistence
- 15-way classifier of emotion over five dialogue conditions.
- Vectors track operative emotion at a token position, not persistent state.
- Beliefs linearly decodable.
- Interprets the Sofroniew et al. findings within the Level 2 indicator table.