Does VPD mechanistic faithfulness and interpretability survive at frontier model scale?

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

• Future interpretability techniques will fundamentally resemble VPDclaim0.781
  Prediction/hypothesis about the direction of the field.
• VPD scales to a 4-layer 67M-parameter model trained on The Pile.finding0.769
  Empirical demonstration of VPD on a mid-scale transformer, establishing feasibility.
• VPD is a meaningful step toward bottom-up interpretabilityclaim0.762
  Positioning of VPD as advancing the paradigm of explaining computation in the model's terms.
• The ability to make precise edits demonstrates that VPD identifies real computational machineryclaim0.759
  Claim that editing success validates VPD's decomposition.
• Interpretability features converge across different model architectures, revealing structural similarities.claim0.755
• VPD enables manual model editing through direct parameter manipulation.claim0.748
  Applied capability claim: VPD enables surgical changes to model behaviour at the parameter level.
• VPD subcomponents are sparse, interpretable, and avoid feature splitting.claim0.744
  Assertion about the qualitative advantages of VPD's rank-one decomposition.
• Lindsey 2025: frontier models can detect and report changes in their own internal activations via concept injection experiments, demonstrating functional introspective awarenessfinding0.742
  Prior finding cited as convergent evidence for LLM self-awareness capacities