method
active
method:nearest-neighbor-features-toolNearest neighbor features tool
Retrieves similar features by decoder cosine similarity.
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Tool-using agent for open-ended model investigations.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Stringent test distinguishing a feature from its 10 nearest embedding neighbours with one true and three distractor examples, repeated ten times.
- Powerful structural force in diagrams; controls associations and semantic relations between elements.
- Interpretable concept representations hypothesized as basic units of computation.
- A hypothesized intermediate-level linearly-represented feature (e.g., Beijing and China are closely associated) that may correlate with truth in unnegated datasets but anti-correlate in negated ones
- Research thread within About Blank concerning the structure and relational properties of neural network feature representations; covariance pooling tangentially supports this thread.
- Features with similar decoder magnitude across both models.
- Dual interpretation of features: in addition to responding to inputs, features also act to increase probability of specific output tokens
- Method of optimizing input to cause a neuron to fire maximally, used to characterize what a neuron detects; establishes causal link