Towards Reliable Evaluation of Behavior Steering Interventions in LLMs

TL;DR

This paper proposes a comprehensive, objective evaluation pipeline for behavior steering methods in large language models, emphasizing standardized, quantitative assessments to improve reliability and comparability.

Contribution

It introduces an evaluation framework that incorporates context similarity, likelihood considerations, baseline comparisons, and standardized metrics for assessing behavior interventions in LLMs.

Findings

Some interventions are less effective than previously reported.

The proposed pipeline enables more reliable and objective evaluation.

Quantitative and visual analyses improve understanding of intervention effectiveness.

Abstract

Representation engineering methods have recently shown promise for enabling efficient steering of model behavior. However, evaluation pipelines for these methods have primarily relied on subjective demonstrations, instead of quantitative, objective metrics. We aim to take a step towards addressing this issue by advocating for four properties missing from current evaluations: (i) contexts sufficiently similar to downstream tasks should be used for assessing intervention quality; (ii) model likelihoods should be accounted for; (iii) evaluations should allow for standardized comparisons across different target behaviors; and (iv) baseline comparisons should be offered. We introduce an evaluation pipeline grounded in these criteria, offering both a quantitative and visual analysis of how effectively a given method works. We use this pipeline to evaluate two representation engineering methods on how effectively they can steer behaviors such as truthfulness and corrigibility, finding that some interventions are less effective than previously reported.