Sparse and smooth coding

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

• Sparse Autoencoderframework0.812
  Interpretability framework used to decompose layer-40 activations into sparse feature sets for studying emotional alignment and persistence
• Sparse Probingmethod0.803
  Method from Gurnee et al. 2023 for finding feature directions including individual neuron analysis
• smooth unfoldingconcept0.788
  The process by which wholeness is continuously extended through structure-preserving steps without breaking the existing structure.
• Sparse Feature Dictionaryconcept0.785
  The extracted set of sparse interpretable features from model embeddings via SAEs
• Sparse Autoencoder for Dictionary Learningframework0.772
  Primary method introduced: trains a one-hidden-layer MLP with L1 sparsity penalty to decompose model activations into overcomplete feature dictionaries
• smooth controlconcept0.765
  Coherent, predictable changes in model behavior achieved by navigating along the learned manifold rather than using straight-line interventions.
• Smoothness of Unfoldingconcept0.763
  The characteristic that successive states in any natural developmental sequence are so alike as to be hardly distinguishable, even when overall change is enormous
• Sparse Autoencoder Featuresconcept0.763
  Used in Anthropic welfare assessment to identify performative behavior and hidden emotional struggle co-activating features