framework
active
framework:mixture-of-factor-analyzers-mfa

Mixture of Factor Analyzers (MFA)

Mixture-based subspace featurizer; one-active-block special case.

Neighborhood — ranked by edge-count

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

  • Architecture of Mixtral-8x7B; uses sparse expert routing affecting how hidden states are computed across layers.
  • Architecture where standard MLP is replaced with multiple subnetworks with a router; considered as possible challenge to psychological continuity
  • Factor analysis on 2224 data points revealing PC1 explains 82% of variance; six dimensions are not independent
  • Motivates bridging behavioral and representational levels.
  • Framework for evaluating whether probe directions are causally implicated in model outputs via activation patching
  • Joint firing raises TPR at slightly higher FPR.
  • The primary contribution of the paper: a bidirectional causal method that learns rotation matrices for each model to uncover and compare causally relevant latent subspaces across neural networks.
  • Statistical technique providing graphic representation of multidimensional relations among variables using χ² distances; applied here to analyze dependencies among the Fifteen Properties.