An affine map trained on The Pile transfers steering vectors from Gemma-2B to Gemma-9B and reproduces much of the large model's native steering behavior.
Curve circuits
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
-
Linear Representation Transferability Hypothesis: Leveraging Small Models to Steer Large Models
An affine map trained on The Pile transfers steering vectors from Gemma-2B to Gemma-9B and reproduces much of the large model's native steering behavior.