Safety Through Reasoning: An Empirical Study of Reasoning Guardrail Models

Makesh Narsimhan Sreedhar, Traian Rebedea, Christopher Parisien

Reasoning-based language models have demonstrated strong performance across various domains, with the most notable gains seen in mathematical and coding tasks. Recent research has…