N-GLARE: An Non-Generative Latent Representation-Efficient LLM Safety Evaluator

TL;DR

N-GLARE introduces a novel, efficient method for evaluating LLM safety by analyzing latent representations instead of generating text, enabling faster and cost-effective safety diagnostics.

Contribution

It proposes a new latent representation-based safety evaluation method, JSS metric, that reduces costs and latency compared to traditional red teaming approaches.

Findings

JSS correlates strongly with safety rankings from red teaming.

N-GLARE achieves similar discriminative results at less than 1% of token and runtime costs.

The method enables real-time safety diagnostics without text generation.

Abstract

Evaluating the safety robustness of LLMs is critical for their deployment. However, mainstream Red Teaming methods rely on online generation and black-box output analysis. These approaches are not only costly but also suffer from feedback latency, making them unsuitable for agile diagnostics after training a new model. To address this, we propose N-GLARE (A Non-Generative, Latent Representation-Efficient LLM Safety Evaluator). N-GLARE operates entirely on the model's latent representations, bypassing the need for full text generation. It characterizes hidden layer dynamics by analyzing the APT (Angular-Probabilistic Trajectory) of latent representations and introducing the JSS (Jensen-Shannon Separability) metric. Experiments on over 40 models and 20 red teaming strategies demonstrate that the JSS metric exhibits high consistency with the safety rankings derived from Red Teaming. N-GLARE reproduces the discriminative trends of large-scale red-teaming tests at less than 1\% of the token cost and the runtime cost, providing an efficient output-free evaluation proxy for real-time diagnostics.