Comparing Optimization Targets for Contrast-Consistent Search

Hugo Fry, Seamus Fallows, Jamie Wright, Ian Fan, Nandi Schoots

Keywords: Eliciting Latent Knowledge, Contrast-Consistent Search (CCS), Optimization Targets

Abstract: We investigate the optimization target of contrast-consistent search (CCS), which aims to recover the internal representations of truth of a large language model. We present a new loss function that we call the Midpoint-Displacement (MD) loss function. We demonstrate that for a certain hyper-parameter value this MD loss function leads to a prober with very similar weights to CCS. We further show that this hyper-parameter is not optimal and that with a better hyper-parameter the MD loss function tentatively attains a higher test accuracy than CCS.

Submission Number: 78