Activation steering with contrastive vectors becomes less effective as Llama 2 models scale from 7B to 70B parameters, with peak effect at roughly 40% of the model's layers.
A toy model of universality: Reverse engineering how networks learn group operations
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Scaling laws for activation steering with Llama 2 models and refusal mechanisms
Activation steering with contrastive vectors becomes less effective as Llama 2 models scale from 7B to 70B parameters, with peak effect at roughly 40% of the model's layers.