Actually, LLMs aren’t that great for summarizing. It would be a boon for RAG workflows if they were.

I’m still on the lookout for a great model for this.