Conceptreadout scaling / 读出变换
Readout Scaling
The application of a diagonal scaling or linear transformation to activation vectors before feeding them into an SAE, altering the relative weight of each dimension in the reconstruction objective and thus influencing which directions the SAE prioritizes recovering.
Timeline (last 90 days)
Oct 14
Readout scaling refers to applying a diagonal scaling or linear transformation to the activation vector before feeding the activations into the SAE, thereby changing the relative weight of each dimension in the reconstruction objective.
Unverified50%