Sparse Activation Spaces

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

• Activation spaceconcept0.828
  Representation space on which linear probes operate to attribute harmful behaviors to training data.
• geometry of activation spaceconcept0.772
  Rich geometric structure carried by neural representations.
• Direction (activation space)concept0.758
  A linear combination of neurons in a layer; the general form of a neural network feature including both individual neurons and other combinations
• Sparse Autoencoderframework0.743
  Interpretability framework used to decompose layer-40 activations into sparse feature sets for studying emotional alignment and persistence
• Sparse circuitsconcept0.741
  A goal in mechanistic interpretability to identify sparse computational subgraphs; VPD promotes sparse parameter circuits.
• Behavior Spaceconcept0.740
  A geometric space of all output token probability distributions, equipped with Hellinger distance, used to visualize model behavior.
• Sparse Feature Dictionaryconcept0.733
  The extracted set of sparse interpretable features from model embeddings via SAEs
• Sparse Autoencoder Training on Layer-40 Activationsmethod0.730
  SAEs trained on 100M+ tokens to compress token layer-40 activations into 64 active features out of 100K+ for interpretability analysis