VPD decomposes parameters, not activations, flipping the standard SAE / activation-patching paradigm.

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

• Decompose parameters, not activationsquote0.797
  Core slogan encapsulating the paradigm shift of VPD.
• VPD enables manual model editing through direct parameter manipulation.claim0.794
  Applied capability claim: VPD enables surgical changes to model behaviour at the parameter level.
• VPD subcomponents avoid feature splitting, improving interpretability over SAE approachclaim0.778
  Core interpretative claim that VPD's parameter-based decomposition prevents the feature fragmentation seen in activation-based methods.
• VPD (adVersarial Parameter Decomposition)concept0.774
  Core methodological framework introduced in this paper; decomposes weight matrices into rank-one interpretable subcomponents using adversarial ablations.
• VPD scales to a 4-layer 67M-parameter model trained on The Pile.finding0.772
  Empirical demonstration of VPD on a mid-scale transformer, establishing feasibility.
• VPD subcomponents are sparse, interpretable, and avoid feature splitting.claim0.769
  Assertion about the qualitative advantages of VPD's rank-one decomposition.
• The VPD-based edit has similarly low off-target effects as uninterpretable fine-tuning methodsfinding0.763
  Performance comparison showing subcomponent editing is comparable to fine-tuning in preserving off-target behavior.
• Parameter Decomposition (vs Activation)concept0.763