When, and to what extent, is it okay for divergences to occur?

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

• When it is not okay, how can we prevent divergent representations from occurring?question0.833
  Third core research question motivating the CL loss approach in Section 5
• The harm of divergence is inherently claim-dependent: the same divergence can be harmless for one mechanistic claim and pernicious for anotherclaim0.780
  Important nuance that prevents a universal classification of divergence as always good or bad
• How can we produce a principled method for classifying harmful divergence for any mechanistic claim?question0.753
  Identified gap: current work lacks a general method for harmful divergence classification
• Any divergence outside of the null-space of NN layers is potentially pernicious, posing challenges for a complete mechanistic understanding of NNsclaim0.751
  Sobering conclusion about the fundamental challenge posed by divergence for mechanistic interpretability
• Representational Divergenceconcept0.749
  Core phenomenon studied: when causal interventions shift internal representations away from the natural distribution
• Divergent representations are a common, if not likely, outcome of causal interventions across a wide range of methodsclaim0.748
  Core empirical claim of the paper supported by both theoretical proof and empirical demonstration
• Harmless Divergenceconcept0.747
  Divergences that occur in the behavioral null-space and do not affect functional claims about the model
• Minimizing divergence magnitude does not guarantee elimination of hidden pathways; it only reduces the risk surfaceclaim0.744
  Important caveat to the CL loss solution, noting it is a step not a complete fix