Don't Walk the Line: Boundary Guidance for Filtered Generation

Abstract

Generative models are increasingly paired with safety classifiers that filter harmful or undesirable outputs. A common strategy is to fine-tune the generator to reduce the probability of being filtered, but this can be suboptimal: it often pushes the model toward producing samples near the classifier’s decision boundary, increasing both false positives and false negatives. We propose \emph{Boundary Guidance}, a reinforcement learning fine-tuning method that explicitly steers generation away from the classifier’s margin. On a benchmark of jailbreak, ambiguous, and long-context prompts, \emph{Boundary Guidance} improves both the safety and the utility of outputs, as judged by LLM-as-a-Judge evaluations. Comprehensive ablations across model scales and reward designs demonstrate the robustness of our approach. The code is available at https://anonymous.4open.science/r/boundary-avoidance-45G2-45FD1.