Rank-one matrix

The simple matrix form into which VPD constrains subcomponents to enforce mechanistic simplicity.

Neighborhood — ranked by edge-count

method

Rank-one matrix decomposition
implements
Constraint in VPD where each parameter subcomponent is constrained to be a rank-one matrix for simplicity.

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.

Rank-one subcomponentsconcept0.771
The atomic pieces into which weight matrices are decomposed by VPD; each rank-one component is interpretable.
Copying Matrixconcept0.754
An OV circuit matrix that maps tokens to increasing the logit of those same tokens; detectable via positive eigenvalues
Transition Matrix (B matrix)concept0.750
Prescribes transitions among hidden state factors under action; encodes policy-dependent dynamics
Rank-one matrix decomposition constraint enforcing mechanistic simplicityclaim0.746
Core design principle of VPD: each parameter subcomponent is constrained to be a simple rank-one matrix to enable isolated understanding and combination.
Orthogonal Rotation Matrixconcept0.732
Key mathematical object in DAS; transforms neural representations to alternative bases to reveal distributed structure.
class-one structureconcept0.732
A structure generated by a living process, with deep interlocking and wholeness, as described in earlier volumes.
Weight matrix decompositionconcept0.724
The core idea of decomposing weight matrices into components for interpretability.
hierarchyconcept0.711
An ordering of texts via spatial cues like indentation, size, and placement, implying importance.