Meta's new structured prompting technique makes LLMs significantly better at code review — boosting accuracy to 93% in some cases

This technique can be used out-of-the-box, requiring no model training or special packaging. It is code-execution free, which ...