Pith. sign in

Towards monosemanticity: Decomposing language models with dictionary learning

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

fields

cs.LG 1

years

2025 1

verdicts

CONDITIONAL 1

representative citing papers

Stochastic Parameter Decomposition

cs.LG · 2025-06-25 · conditional · novelty 6.0

SPD uses stochastic masking and a learned causal importance function to decompose neural network parameters into sparsely active rank-one subcomponents, recovering ground-truth mechanisms in toy models where APD struggled.

citing papers explorer

Showing 1 of 1 citing paper.

  • Stochastic Parameter Decomposition cs.LG · 2025-06-25 · conditional · none · ref 4

    SPD uses stochastic masking and a learned causal importance function to decompose neural network parameters into sparsely active rank-one subcomponents, recovering ground-truth mechanisms in toy models where APD struggled.