https://fusionchat.ai/news/deciphering-the-problem-of-negation-for-ai-language-models