“But LLMs are great for summarizing”(No, they fundamentally fail.)https://ea.rna.nl/2024/05/27/when-chatgpt-summarises-it-actually-does-nothing-of-the-kind/