Soft Prompts for Evaluation: Measuring Conditional Distance of Capabilities
This provides a scalable method for automated red teaming and evaluation of potentially concerning behaviors in language models, including future powerful models, addressing a need for quantitative feedback in AI safety.
The paper tackles the problem of evaluating latent capabilities in language models by introducing an approach using optimized input embeddings (soft prompts) as a metric of conditional distance to target behaviors, with results demonstrated in natural language, chess, and pathfinding domains.
To help evaluate and understand the latent capabilities of language models, this paper introduces an approach using optimized input embeddings, or 'soft prompts,' as a metric of conditional distance between a model and a target behavior. The technique aims to facilitate latent capability discovery as a part of automated red teaming/evaluation suites and to provide quantitative feedback about the accessibility of potentially concerning behaviors in a way that may scale to powerful future models, including those which may otherwise be capable of deceptive alignment. An evaluation framework using soft prompts is demonstrated in natural language, chess, and pathfinding, and the technique is extended with generalized conditional soft prompts to aid in constructing task evaluations.