Benchmarks are both difficult and important. Great to see these thoughtful, open reproducible benchmarks for LLM guardrails.

→ View original post on X — @goodfellow_ian,