Feature ablation (zeroing feature activations)

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

• Zero Ablationmethod0.822
  Intervention type that sets activations to zero, used for interpretability analysis
• Feature attribution correlates well with ablation effects, making it an efficient proxy for causal effect.claim0.770
  Gradient-based attribution approximates ablation impact, enabling fast search for causally important features.
• Mechanism by which activation of an emotion feature sometimes leads to later suppression of that same featurequestion0.748
  Identified research gap: the paper observes anti-persistence but has no explanation for it
• L0 Norm of Feature Activationsconcept0.745
  Average number of nonzero feature entries per input; primary measure of activation sparsity in the autoencoder
• Feature steering (clamping feature activations)method0.743
  Modifying model behavior by clamping SAE feature activations to specific values during forward pass.
• Noising/Denoising Activation Patchingmethod0.741
  Methods that intentionally introduce divergent representations to test sufficiency and completeness of circuits
• Why does activation of an emotion feature sometimes lead to its later suppression?question0.736
  Open mechanistic question arising from the causal steering experiment
• Feature attribution (gradient-based) correlates 0.8 with ablation effects on the 'John' and 'Kobe' examples.finding0.733
  Validation of attribution as a fast proxy for causal importance.