Linear Representation Transferability Hypothesis: Leveraging Small Models to Steer Large Models