dataset

active

dataset:67m-parameter-language-model

67M-parameter language model

The target neural network on which VPD was applied and evaluated in this research.

Neighborhood — ranked by edge-count

Methods (1)

method

Adversarial Parameter Decomposition (VPD)
implements
Core technique introduced in this paper for decomposing neural network weight matrices into mechanistically simple, interpretable rank-one subcomponents.