MLP neurons

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

• Multi-layer Perceptron (MLP)method0.826
  Feed-forward neural network with hidden layers, capable of representing non-linearly separable functions.
• When and how can MLP neurons in transformers be individually interpreted, and what progress is needed to extend mechanistic interpretability to them?question0.783
  Major open problem identified in the paper; MLP layers constitute 2/3 of transformer parameters
• Rosetta Neuronsconcept0.766
  Neurons activated by the same pattern across a range of vision models, forming a common dictionary independently discovered by all models
• A sparse set of 28 MLP neurons at layer 18 (~0.2% of MLP) are reused across all cyclic tasksfinding0.746
  Quantitative finding identifying the specific neurons responsible for generic addition
• Some MLP neurons and attention heads perform memory management by reading residual stream information and writing its negative to delete itclaim0.741
  Hypothesis based on observed negative cosine similarity between input and output weights of some neurons
• The 28 MLP neurons at layer 18 can be partitioned into disjoint clusters each computing the sum for a Fourier feature with a different periodfinding0.737
  Structural finding showing modular organization within the sparse neuron set
• Dead Neuronsconcept0.732
  Autoencoder neurons that fail to activate across any datapoints during training; addressed via neuron resampling
• Neural Networksconcept0.729