A benchmark for evaluating the robustness of LLMs and defenses to indirect prompt injection attacks. - View it on GitHub