Towards Reliable Evaluation of Behavior Steering Interventions in LLMs
This addresses the need for more objective and standardized evaluation methods in representation engineering for researchers and practitioners, though it is incremental as it builds on existing work.
The paper tackles the problem of unreliable evaluation of behavior steering interventions in LLMs by proposing an evaluation pipeline with four criteria, and finds that some interventions are less effective than previously reported.
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.