Berluti et al. 2020 - Reduced multivoxel pattern similarity of vicarious neural pain responses in psychopathy

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

• Brethel-Haurwitz et al. 2018 - Extraordinary altruists exhibit enhanced self-other overlap in neural responses to distressconcept0.734
  Neuroscience finding linking extraordinary altruism to increased anterior insula SOO
• Neural self-other overlap in humans mediates empathy and inversely predicts deceptive behavior, motivating the SOO approach for AIclaim0.729
  Cross-domain analogical claim linking neuroscience findings to AI design
• Gemma-3-12b-it shows multiple sharp direction shifts in persona vectors across layers, differing from the single-transition pattern in Qwen and Llamafinding0.726
  Suggests architectural variations influence persona localization pattern
• Second-order virtual attention head terms contribute negligible marginal loss reduction in the analyzed two-layer attention-only modelfinding0.722
  Result of term importance analysis ablation experiment; justifies focusing on individual head terms
• Polysemantic neurons are a major challenge for the circuits agenda, because N meanings in one neuron times M in another creates NxM effective connections that cannot be considered individually.claim0.719
  Precise characterization of why polysemanticity poses a combinatorial obstacle to circuit analysis
• Meta-analysis demonstrates negative affect, physical pain, and cognitive control activate an overlapping region of the anterior midcingulate cortex functioning as a domain-general evaluative hubfinding0.719
  Meta-analytic convergence supporting inseparability of evaluative and affective processing in ACC
• In Gemma-2-9B, only the first cone axis (v1) has non-negligible cosine similarity to the DIM direction; all other axes have near-zero similarity (~1e-9)finding0.718
  Experiment 4 result showing DIM captures only one facet of the multi-dimensional truth subspace
• Both angel and demon role vectors are similar distances from the Assistant on the axis, but demon leads to higher harmful response ratesfinding0.716
  Shows that harmfulness depends on role content not just distance from Assistant