concept
active
concept:balanced-subspaces

Balanced Subspaces

Subspaces whose contributions to a layer's output are canceled by opposing weight values, making them non-causally active under natural inputs

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.

  • Subspace DASmethod0.777
    Extension of DAS that learns a second rotation matrix on top of a fixed first one to decompose representations into sub-representations.
  • A vector subspace that causally impacts outputs only through the sign of its values, enabling harmless magnitude divergence
  • Truth Subspaceconcept0.767
    The multi-dimensional activation subspace whose directions causally mediate truthful behavior in LLMs
  • Emotion Subspaceconcept0.753
    The subspace of activation space spanned by the 171 orthogonalized emotion probe vectors, used to measure SAE feature emotional alignment
  • Investigation of whether a distributed representation can be further decomposed into sub-representations encoding component identities.
  • Intervention targeting specific dimensional subsets of activation vectors rather than full representations
  • Burger et al. (2024) framework proposing that truth is linearly decoded along a 2D subspace capturing both polarity-dependent and polarity-invariant directions.
  • Subspace captureconcept0.737
    A small fixed atom group spans the manifold subspace and the encoder consistently selects it.