Large-scale language models (LLMs) can generate natural sentences that sound like they were written by humans, and can even code, but the sentences and codes they generate often contain errors called ...