ROTATE disentangles MLP neurons into faithful vocabulary channels by optimizing weight rotations to maximize vocabulary-space kurtosis, outperforming activation-based baselines for neuron descriptions.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Disentangling MLP Neuron Weights in Vocabulary Space
ROTATE disentangles MLP neurons into faithful vocabulary channels by optimizing weight rotations to maximize vocabulary-space kurtosis, outperforming activation-based baselines for neuron descriptions.