claim
active
claim:a-layer-s-jacobian-across-data-points-is-a-canonicalized-form-of-parameters-that-captures-mechanism-and-neuron-co-activationA layer's Jacobian across data points is a canonicalized form of parameters that captures mechanism and neuron co-activation
Rationale for Jacobian matching.
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Motivates Jacobian connection.
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.
- Ignition-like effect.
- Consistent but alternative explanations like amplification exist.
- Fundamental theoretical claim motivating DAS, attributed to Smolensky/Rumelhart/McClelland.
- Interpretation of eval loss result.
- Central claim motivating DAS over prior methods.
- Suggests architectural variations influence persona localization pattern
- Architectural requirement from machine learning.
- Activation patching of arithmetic intermediates is effective at layers matching J-lens emergence.finding0.745Causal accuracy.