Disjoint neuron clusters for Fourier periods

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

• Fourier features with period 10 contribute to base-10 sum computation in the 28-neuron clusterfinding0.830
  One of the three base-10 Fourier periods identified in the sparse neuron set
• The 28 MLP neurons at layer 18 can be partitioned into disjoint clusters each computing the sum for a Fourier feature with a different periodfinding0.781
  Structural finding showing modular organization within the sparse neuron set
• Fourier analysis of neural activationsmethod0.758
  Method used to identify the periodic features and their periods in Llama-3.1-8B's MLP neurons
• Neural Representations of Location Composed of Spatially Periodic Bands (Krupic et al., 2012)concept0.755
  Discovery of band cells; TEM-t also recapitulates these representations.
• No neuron found with Hebrew Unicode block in top dataset examples; most correlated neuron A/neurons/489 has correlation of only 0.1finding0.743
  Hebrew feature is effectively invisible in the neuron basis
• Curve detecting neurons found in every non-trivial vision model carefully examinedfinding0.738
  Empirical basis for treating curve detectors as a canonical example of meaningful, understandable features
• The Fourier feature periods (2, 5, 10) respect standard base-10 addition structure rather than cyclic concept periodicityclaim0.738
  Mechanistic claim linking identified Fourier features to base-10 arithmetic
• Superposition hypothesis: neural networks represent more features than dimensions using almost-orthogonal directions.hypothesis0.719
  Explanation for why dictionary learning can recover many more features than dimensions.