English

Comparing Optimization Targets for Contrast-Consistent Search

Machine Learning 2023-11-02 v1 Computation and Language

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 attains a higher test accuracy than CCS.

Cite

@article{arxiv.2311.00488,
  title  = {Comparing Optimization Targets for Contrast-Consistent Search},
  author = {Hugo Fry and Seamus Fallows and Ian Fan and Jamie Wright and Nandi Schoots},
  journal= {arXiv preprint arXiv:2311.00488},
  year   = {2023}
}

Comments

Socially Responsible Language Modelling Research (SoLaR) NeurIPS 2023

R2 v1 2026-06-28T13:08:31.203Z