A benchmark for evaluating the contextual integrity of persistent memory in LLMs. - View it on GitHub