Monitor Research

Can a monitor model detect container escapes? Post-hoc evaluation of LLM monitors reviewing each agent action during live evals.