AI RESEARCH

How catastrophic is your LLM?

Amazon Science

A new framework provides a statistical method for estimating the likelihood of catastrophic failures in large language models in adversarial conversations.