institute
active
institute:transluce

Transluce

Source of LatentQA follow-ups on SAE explanations and user modeling.

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.

  • Transcodermethod0.746
    Uses activations from one layer to predict the next; special case of crosscoders.
  • Transcodersmethod0.719
    Decomposition method for activations; VPD is compared against transcoders in sparsity-reconstruction tradeoff.
  • SHIFTmethod0.717
    Method from sparse feature circuits that SCR automates.
  • Used to express MLP in monosemantic features, 4096 features with JumpReLU.
  • Dilutionconcept0.706
    Tiling form with many redundant overlapping atoms of mixed selectivity; none compact enough to capture.
  • Extenuationmethod0.698
    Attribute: any conditional refinement, a softening or complicating of a statement.
  • Tononi's measure of functional integration in neural systems, a precursor to integrated information.
  • Mask virtual weights during training targeting interference.