SafePyramid: A Hierarchical Benchmark for In-context Policy Guardrailing


This is a companion discussion topic for the original entry at https://arxiv.org/abs/2606.29887