Do divergent representations change what an intervention can say about an NN's natural mechanisms?

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

• Divergent representations are a common, if not likely, outcome of causal interventions across a wide range of methodsclaim0.829
  Core empirical claim of the paper supported by both theoretical proof and empirical demonstration
• When it is not okay, how can we prevent divergent representations from occurring?question0.781
  Third core research question motivating the CL loss approach in Section 5
• Patching h[1] with a divergent representation can activate distinct, hidden pathways that result in misleadingly confirmatory behavior and/or undetected behavior.quote0.769
  Load-bearing description of the core pernicious divergence mechanism illustrated in Figure 1
• Neural representations carry rich geometric structure; but does that structure causally shape behavior?quote0.766
  Opening sentence framing the paper's core inquiry.
• Change-of-Basis for Neural Representationsconcept0.761
  Key insight that rotating a neural representation to a non-standard basis can reveal distributed causal structure invisible in standard neuron-aligned basis.
• Neural representation geometry causally shapes behavior; interventions respecting that geometry will yield natural trajectories.hypothesis0.760
  Central hypothesis tested via manifold steering experiments across language models and video world models.
• How do interventions on representations causally steer behavior?question0.755
  Core question motivating the shift from linear to geometry-aware steering; answered via manifold alignment analysis.
• Does the geometric structure of neural representations causally shape model behavior?question0.755
  The motivating research question of the paper