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

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.

Key Contributions

• N-GLARE: a non-generative LLM safety evaluator that operates entirely on latent representations without requiring text generation, enabling output-free safety diagnostics
• Angular-Probabilistic Trajectories (APT) and the Jensen-Shannon Separability (JSS) metric that quantify geometric separation of hidden-layer dynamics under benign vs. jailbreak conditions
• Empirical validation across 40+ models and 20 red-teaming strategies showing JSS rankings are highly consistent with traditional red-teaming at less than 1% of token and runtime cost

🛡️ Threat Analysis

Details

Similar Papers