method
active
method:highly-activating-dataset-examplesHighly activating dataset examples
Simplest method to describe causes of activation.
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- Three-step approach to identify roles of network components.
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.
- SQL databases of top-activating examples per feature for researchers lacking compute to run R1.
- Used to interpret features alongside heatmaps.
- Internal representations of the model on which probes operate; the method uses activations to rank datapoints.
- If the relationship is present in activations, better tools (e.g. trained on these datasets) could surface ithypothesis0.729Alternative explanation.
- Method of examining top-activating real images from a dataset to characterize neuron behavior
- Large activation magnitudes in the residual stream that Queipo-de Llano et al. link to causing compression behavior and stages of inference
- Finds web text maximally activating probes.
- Clamping activations along the Assistant Axis to remain above a minimum threshold (25th percentile), introduced as a stabilization method