method
active
method:frobenius-norm-comparisonFrobenius Norm Comparison
Used to compare attention matrix similarity across recurrences and validate cyclic fixed point behavior
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Frobenius Norm Composition Measurementrelated_toMeasuring Q-, K-, V-composition between attention heads by computing the Frobenius norm of the product of relevant matrices divided by norms of individual matrices
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.
- Post-hoc metric for model exclusivity in standard crosscoders.
- Experimental method where subjects choose which of two items has more life, yielding agreement and a relative measure of life.
- Sparsity penalty weighting by sum of per-layer decoder norms, enabling comparability with per-layer SAEs.
- When task gradient norms differ greatly, large-norm tasks have not converged while small-norm tasks have nearly convergedhypothesis0.688Motivates setting αk = max norm to enable further learning on under-converged tasks
- Scaling aggregated gradient by the maximum gradient norm among tasks.
- Role of extra dimensions.
- Validates toy model.
- Result from applying the Frobenius norm composition measurement to all attention head pairs in the two-layer model